00:00SPEAKER- I feel like my local Chinese place
00:01SPEAKERis throwing shade at me.
00:03SPEAKER- Uh-oh.
00:04SPEAKER- Yeah, I did a big order the other day
00:06SPEAKERand then they were like, "How much rice do you need?"
00:08SPEAKERI was like, "I need it for two people."
00:10SPEAKERSo then when I got the order,
00:12SPEAKERthere were like eight sets of chopsticks
00:14SPEAKERand eight fortune cookies in there.
00:17SPEAKER- Okay.
00:18SPEAKER- Which I feel like is a judgment
00:19SPEAKERabout the amount of food that I ordered for two people.
00:22SPEAKEROkay, so I ordered enough
00:24SPEAKERso that I wouldn't have to make dinner the next day, right?
00:26SPEAKERI was like- - Sensible?
00:27SPEAKER- Yeah, I was thinking about the future.
00:30SPEAKERBut then I think the real shade comes
00:33SPEAKERbecause I ordered fairly, okay,
00:36SPEAKERmy wife likes sweet and sour pork,
00:37SPEAKERwhich I realize is an Americanized Chinese food dish.
00:42SPEAKERBut they chucked in a free order of crab rangoon.
00:45SPEAKER- Unbidden?
00:46SPEAKER- Unbidden.
00:47SPEAKERI ordered enough that I got a bonus
00:49SPEAKERand they didn't give me a bonus of like,
00:53SPEAKERI don't know, like something,
00:55SPEAKERthey gave me the American food.
00:57SPEAKER- To be clear,
00:58SPEAKERlike there was not some disclaimer on the menu.
00:59SPEAKERIt's like, "Hey, order $100 worth of food
01:01SPEAKERand get free crab ring."
01:02SPEAKERLike they literally just manifested in the bag.
01:04SPEAKER- Okay, first off, I didn't order $100 worth of food.
01:06SPEAKERThe thing that happens at this place is
01:09SPEAKERif you order toward the end of their night,
01:11SPEAKERthen often they'll chuck in a little bonus
01:13SPEAKERif you do a big order.
01:15SPEAKER- Those are the best.
01:16SPEAKERThose are the best.
01:17SPEAKER- Like, just to be clear,
01:18SPEAKERI love our local Chinese place.
01:20SPEAKERAnd if they're throwing shade at me
01:23SPEAKERby giving me the Americanized,
01:24SPEAKERfreebie Americanized Chinese food,
01:28SPEAKERI don't know how I feel about that.
01:29SPEAKERI'm a little sad.
01:30SPEAKER- Dude, that's like when I lived across the street
01:32SPEAKERfrom an escape from New York pizza in my twenties
01:34SPEAKERand would stumble in there drunk at 1.30 in the morning,
01:37SPEAKERneeding grease.
01:38SPEAKER- Yeah.
01:39SPEAKER- And they would be like, "Oh, we're about to close.
01:40SPEAKERHere you go."
01:41SPEAKER- Yeah.
01:42SPEAKER- And they would just give me some slices
01:43SPEAKER'cause they weren't about to throw it out anyway.
01:45SPEAKER- Yeah, the one, it's funny,
01:47SPEAKERI used to go to the one, the escape on hate
01:49SPEAKER'cause I used to live right up the street from there.
01:52SPEAKERAnd occasionally you'd go in and you'd order a slice
01:56SPEAKERand then they just put two more on top of it.
01:58SPEAKER- Yes.
01:59SPEAKER- It's good times, it's good times.
02:00SPEAKERBut so do you, should I be affronted
02:02SPEAKERor do you think they were just trying
02:03SPEAKERto do something nice for me?
02:04SPEAKERThat's the question.
02:05SPEAKER- No, yeah, I would say they appreciate your business.
02:08SPEAKER- But I feel like it's a little bit,
02:09SPEAKERI feel like the extra chopsticks is a judgment.
02:11SPEAKERI feel like I'm being judged now.
02:13SPEAKER- Maybe they knew.
02:14SPEAKEROh, look, they just appreciate cleanliness.
02:15SPEAKERThey knew you were gonna eat this over a period of days
02:17SPEAKERand they were like,
02:18SPEAKER"Oh, they're gonna need new chopsticks every day
02:20SPEAKER'cause this is four days worth of food."
02:22SPEAKERThat's true, I did appreciate the extra fortune cookies too.
02:25SPEAKER(upbeat music)
02:50SPEAKER(upbeat music)
02:53SPEAKER- Welcome to Brad and Will Made at TechPod.
02:56SPEAKERI'm Will.
02:56SPEAKER- I'm Brad.
02:57SPEAKER- Hello, Brad.
02:58SPEAKER- Hi, hello.
02:59SPEAKERIt's the most wonderful time of every two years.
03:03SPEAKER- It's video card season.
03:04SPEAKER- It's new video cards.
03:05SPEAKER- Well, okay.
03:06SPEAKER- Has it always been every two years
03:07SPEAKERor did the pandemic stretch or elongate things out?
03:10SPEAKER- I think they stretched six months.
03:12SPEAKERIt's usually about every 18 months.
03:13SPEAKER- Is it really?
03:14SPEAKER- But usually, so usually there's like,
03:17SPEAKERusually it's two years between new GPU
03:19SPEAKERkind of architectures
03:23SPEAKERand then, or two generations between new.
03:25SPEAKERSo there'll be like a launch one
03:27SPEAKERand then there'll be an updated one
03:28SPEAKERand then there'll be a new architecture.
03:30SPEAKERIt's typically I go,
03:31SPEAKERsometimes they stretch a third piece of hardware in there,
03:34SPEAKERa third generation,
03:36SPEAKERbut usually it's like every say three years
03:38SPEAKERfor new architectures.
03:40SPEAKER- You ever go back and well, you probably don't do this
03:41SPEAKER'cause you reviewed cards.
03:42SPEAKERSo you just touched all of them,
03:44SPEAKERbut sometimes I go back and look at the entire history
03:46SPEAKERof Nvidia GPU releases
03:48SPEAKER'cause I can kind of like clock arrows
03:50SPEAKERof my life against them.
03:51SPEAKER- Yeah.
03:52SPEAKER- It's like, oh, I remember when I scaffold that TNT2
03:56SPEAKERout of my parents' new computer
03:57SPEAKER'cause they sure as hell didn't need it
03:58SPEAKERand they could get by just fine on that Matrox millennium.
04:01SPEAKERIt probably looked better on the desktop anyway.
04:03SPEAKER- Yeah, you're doing them a favor.
04:04SPEAKER- And it's like, oh, I remember when I borrowed a,
04:07SPEAKERGod, what was the one around Doom 3?
04:10SPEAKER- Oh, Doom 3 would have been like a 5800
04:12SPEAKERor something like that. - 6800.
04:14SPEAKERMight've been a 6800 Ultra from the office.
04:16SPEAKER- Yeah, the 6800 would have been the good version of,
04:18SPEAKERlike the 5800 was the leaf blower.
04:20SPEAKER- Okay, yes, it was the one after that.
04:21SPEAKERAnyway, like there have been a ton of Nvidia cards
04:23SPEAKERand you can just go down the list and be like,
04:25SPEAKERoh, I bought that one.
04:26SPEAKEROh, I borrowed that one.
04:27SPEAKERI managed to steal, not steal,
04:31SPEAKERbut somebody gave me a review sample of that one.
04:33SPEAKERYou know what I mean?
04:34SPEAKER- Yep.
04:34SPEAKER- There've been a lot of them is the point.
04:36SPEAKER- It's interesting.
04:37SPEAKERSo for me, the thing on that is the tech demos.
04:40SPEAKERLike I like to occasionally, I'd even did it before,
04:43SPEAKERI want to say the first RTX stream when they announced
04:48SPEAKERthat I went through and loaded up all of the old,
04:51SPEAKERlike the nymphs and the bells, the musical instruments
04:56SPEAKERand the night cities and the werewolves and all that stuff.
04:59SPEAKERAnd it's fun to go back and look at those
05:02SPEAKER'cause it shows what they were interested in highlighting
05:04SPEAKERmore than anything.
05:06SPEAKER- Yeah, yeah, it's the fairy groove to me.
05:09SPEAKER- Well, yeah.
05:10SPEAKER- I think that, isn't that the,
05:10SPEAKERthe original hardware T and L, is that what that was for?
05:14SPEAKER- I am terrible at remembering which one goes with which.
05:16SPEAKERThe first fairy was reflected.
05:19SPEAKERWell, the gecko is the one that is my all-time favorite
05:21SPEAKER'cause it's the one that had like dot three bump mapping
05:24SPEAKERand reflective maps and all that kind of stuff.
05:27SPEAKERIt was really cool.
05:28SPEAKERAnd you could hit a button to switch
05:30SPEAKERbetween the different modes.
05:31SPEAKERSo you could basically get like Mario 64 liquid metal
05:35SPEAKERand then you could do translucency
05:37SPEAKERand you could do refraction and all sorts of other stuff.
05:41SPEAKER- Here it is, the G-Force 3.
05:43SPEAKER- That was a G-Force 3 one, yeah.
05:44SPEAKER- I didn't have a three, so I missed that one.
05:46SPEAKERI'm thinking about the rigid,
05:48SPEAKERmight've been like the first major tech demo
05:49SPEAKERthey ever put out.
05:50SPEAKERIt's nothing but some low poly trees
05:52SPEAKERand some colored lights flitting around them.
05:54SPEAKER- Oh, yeah, yeah, yeah.
05:55SPEAKER- I'm pretty sure that was the original like,
05:57SPEAKERhey, we do transform and lighting on the graphics card now,
05:59SPEAKERlook at this.
06:00SPEAKER- Yeah, well, and then later on they started doing like,
06:02SPEAKERstarting with the werewolf and the fairy,
06:05SPEAKERthe fairy with wings,
06:07SPEAKERthey would basically say,
06:09SPEAKERhey, what happens if we throw the entire performance budget
06:12SPEAKERof the highest end card we have at one character?
06:16SPEAKERAnd that was always fun 'cause then you'd see stuff
06:18SPEAKERthat was like three years out
06:19SPEAKERfrom actually shipping and hardware,
06:20SPEAKERlike subsurface scattering on skin and stuff like that.
06:23SPEAKER- It's gonna be two more graphics cards generations
06:25SPEAKERbefore you actually see games looking like this.
06:27SPEAKER- Exactly, but even after 20 years of covering video cards,
06:33SPEAKER20 plus, 22-ish years of covering video cards off and on,
06:37SPEAKERI think that this 4090 is maybe the most interesting card
06:41SPEAKERI've tested in this entire run across AMD, Intel, everybody.
06:45SPEAKER- That's a large claim.
06:48SPEAKER- It's a real weird card, like it's a weird launch,
06:51SPEAKERit's a weird because of the current state of the market,
06:54SPEAKERit's weird because I don't exactly know
06:58SPEAKERwho this card is for, like, I mean, look,
07:01SPEAKERrealistically this card is for people who do data science
07:03SPEAKERand machine learning stuff and then also people
07:05SPEAKERwho want the fastest possible thing.
07:07SPEAKER- That has always been my impression of the 90 cards
07:10SPEAKERis like, hey, this is, the only people who need this
07:12SPEAKERare people running giant workstations
07:16SPEAKERand people who want, who have hot rod syndrome
07:18SPEAKERand just need to say they have the fastest thing.
07:20SPEAKER- There's a scientist in one of my discords
07:23SPEAKERwho does like bioinformatics stuff
07:28SPEAKERand she has a bunch of 90 series cards in workstations
07:32SPEAKERto do data math, right?
07:34SPEAKER- That sounds awesome, data math.
07:38SPEAKER- Like everyone, like during the pandemic at one point,
07:39SPEAKERshe posted a picture of like 1230 90s
07:42SPEAKERor something stacked on top of each other
07:43SPEAKERand boxes and people hit the floor.
07:46SPEAKER- All right, you've just made me realize I have,
07:49SPEAKERthis is a longstanding life ambition
07:51SPEAKERthat I've just now identified as an ambition
07:53SPEAKERand it is to run a farm of something.
07:56SPEAKERI don't know what, probably not a farm of 40 90s.
07:59SPEAKER- I do tomatoes probably, maybe corn.
08:02SPEAKER- I mean a data farm.
08:03SPEAKER- Oh, oh, look, you could have gotten real into crypto
08:06SPEAKERlike six months ago, but that ship has sailed sadly.
08:09SPEAKER- Maybe it's just. - Or maybe Goodly, probably.
08:11SPEAKER- Should just put together a Raspberry Pi cluster
08:13SPEAKERand call it a day.
08:15SPEAKER- And a Raspberry Pi cluster is kind of,
08:16SPEAKERyou can do a lot with a Raspberry Pi cluster these days.
08:18SPEAKER- Yeah, I think people run like little web servers
08:20SPEAKERoff of them and stuff.
08:22SPEAKER- Oh yeah, oh yeah, you just get a traffic router,
08:24SPEAKERthe whole thing.
08:25SPEAKER- Get one of those little rack blade things
08:27SPEAKERthat holds like 12 Pis.
08:29SPEAKER- Yeah, it's a cluster of Pis.
08:32SPEAKER- So, okay, so today we're gonna talk about the 40 90,
08:35SPEAKERbut more we're gonna talk about the ADA series
08:39SPEAKERof ADA architecture from Nvidia, which is their new,
08:44SPEAKERit succeeds Ampere, which succeeds Turing.
08:46SPEAKERAnd those are kind of the ray tracing.
08:49SPEAKERYou know, it's the next in the line of ray tracing cores
08:52SPEAKERfrom Nvidia.
08:53SPEAKER- I thought they were gonna go with Lovelace,
08:55SPEAKERbut then we showed up for that editor's day
08:57SPEAKERand everybody was just saying ADA and I was like, oh,
09:00SPEAKERI guess it's easier to say.
09:01SPEAKER- I think ADA is easier to say.
09:04SPEAKERSo the interesting thing, so, okay,
09:07SPEAKERthis card is interesting, this 40 90 is interesting.
09:10SPEAKERAnd rather than do, so I figured this time
09:13SPEAKERwe would do this in the reverse order than we usually do.
09:15SPEAKERWe're gonna tell you whether the current card,
09:16SPEAKERtell you about the card and whether it's worth buying
09:18SPEAKERand then jump to the stuff that's about
09:20SPEAKERwhy it's interesting for people who care.
09:22SPEAKERSo like we can front load this with,
09:25SPEAKERno, you probably don't need to buy a $1,600 video card.
09:29SPEAKER- Is that your assessment?
09:30SPEAKER- My assessment is, okay, yes,
09:33SPEAKERthat is my short-term assessment.
09:34SPEAKERI'm gonna tell you why now.
09:39SPEAKERIt's stupid fast on everything I tested it.
09:42SPEAKERI put it in-
09:42SPEAKER- Hang on.
09:43SPEAKER- Yeah.
09:45SPEAKER- I mean, go on.
09:46SPEAKER- It's so-
09:47SPEAKER- When you come out and say it's, I don't know man,
09:50SPEAKERyou're making a pretty good case for it.
09:51SPEAKER- It's so stupid fast that you can't get full,
09:56SPEAKERokay, there's some hardware that is coming out very soon
09:59SPEAKERthat we don't know how it's gonna perform from Intel
10:02SPEAKER'cause there's still embargoes on the new Intel CPUs
10:05SPEAKERthat were announced a couple of weeks ago.
10:07SPEAKERIt seems like it is CPU bound on basically all the hardware
10:11SPEAKERthat's available right now
10:12SPEAKERunless you do some crazy overclocking stuff.
10:14SPEAKER- Including the Ryzons that are less than a month old.
10:16SPEAKER- Including the month old Ryzons.
10:18SPEAKERI don't know about the next gen Intel stuff,
10:20SPEAKERbut judging by how it performed from people who tested
10:22SPEAKERon the current gen Intel stuff seems unlikely
10:27SPEAKERthat they're gonna blow through any performance barriers.
10:30SPEAKER- Yeah, I don't think Raptor Lake is that different from,
10:33SPEAKERwe'll have Gordon soon, right?
10:34SPEAKER- Yeah, I think Gordon's gonna come by next week
10:36SPEAKERis the plan.
10:37SPEAKER- I think the early info is that I don't think Raptor
10:39SPEAKERis a huge, huge like architectural change from older.
10:43SPEAKER- Well, and what CPU bound means is that the CPU
10:46SPEAKERis the bottleneck for most performance on this GPU
10:50SPEAKERbecause the GPU is so massively overpowered for games.
10:54SPEAKERThis is why when people say,
10:55SPEAKERhey, this isn't for people who play games,
10:57SPEAKERthis is what they mean.
11:00SPEAKERIt only really makes sense if you're gaming,
11:02SPEAKERif you're playing games at 4K.
11:03SPEAKERIf you're competitive at 1440,
11:05SPEAKERlike Nvidia made a case that they think
11:07SPEAKERthat competitive gamers are gonna move from 1080p to 1440.
11:10SPEAKERAnd by competitive, I mean, I don't mean like,
11:13SPEAKERyo, I play Call of Duty every night.
11:14SPEAKERI mean like I'm on an esports team and I play for money
11:17SPEAKERand stuff like that.
11:20SPEAKERI'm skeptical of that because 1080 has a lot of benefits
11:24SPEAKERboth for people who are playing and streaming
11:27SPEAKERand also like frame dips are death in a competitive shooter
11:33SPEAKERwhere like flicks or pixel precision matter.
11:37SPEAKERSo my guess is that people are gonna go
11:38SPEAKERfor higher frame rate versus slightly more resolution
11:42SPEAKERand lower frame rate.
11:44SPEAKER- Here's a sidebar.
11:46SPEAKERDo you think anybody in the esports world
11:48SPEAKERis using any kind of reconstruction DLSS, AMD, FSR,
11:52SPEAKERor do they only want native resolution
11:54SPEAKERbecause they can't afford any like wayward artifacts,
11:57SPEAKERpixel artifacts from the reconstruction
11:58SPEAKERlike they only want native rendering?
12:00SPEAKER- So that is a really good question.
12:01SPEAKERI actually don't have any idea what the answer to that is.
12:03SPEAKERI know of the competitive games.
12:06SPEAKERI don't think...
12:08SPEAKERI think actually Call of Duty supports DLSS too.
12:12SPEAKERI think Apex supports DLSS too.
12:15SPEAKER- Yeah, I would assume everything does.
12:17SPEAKER- The CPU bound stuff like Counter Strike and Fortnite
12:20SPEAKERand those games mostly flipped on reflex.
12:25SPEAKERSo I'm wondering, I don't know what the thought is
12:28SPEAKERon DLSS and competitive games.
12:29SPEAKERI'd love to hear what people know
12:31SPEAKERif there's word out there on that.
12:33SPEAKER- Yeah, I mean, the use case to me is like trying to land
12:35SPEAKERa sniper rifle headshot from a very long distance
12:38SPEAKERwhere the head you're aiming at
12:39SPEAKERis probably only a few pixels high.
12:41SPEAKER- So my experience playing Fortnite with DLSS on
12:44SPEAKERis that DLSS doesn't, the artifacting in DLSS
12:48SPEAKERdoesn't impact players.
12:49SPEAKERIt impacts scenery that's moving around you.
12:52SPEAKERSo things like chain link fences and stuff with that,
12:57SPEAKERlike that kind of pattern is where the artifacting pops up,
12:59SPEAKERnot necessarily on stuff with high contrast,
13:01SPEAKERlike a person standing on a ridge line
13:02SPEAKERor something like that.
13:04SPEAKERBut I'm definitely not an expert on that.
13:08SPEAKERSo it really only makes sense at 4K.
13:10SPEAKERThat said, everything I tested at 1080p
13:14SPEAKERbasically gave me double the performance of a 3080 Ti
13:18SPEAKERin both ray tracing and raster benchmarks.
13:22SPEAKER- Double.
13:22SPEAKER- Double.
13:23SPEAKERAnd when you turn, that's without turning DLSS on.
13:25SPEAKERWhen you turn DLSS on, you're looking at four times
13:28SPEAKERthe performance in a lot of cases.
13:30SPEAKER- For ray tracing as well?
13:31SPEAKER- For ray tracing as well.
13:33SPEAKER- Man.
13:33SPEAKER- I was playing, I was able to play, so there's a beta.
13:36SPEAKERSo, okay, first off disclosure,
13:38SPEAKEReverything that I'm talking about testing
13:39SPEAKERwas on a beta branch that was set up by NVIDIA for this
13:44SPEAKERbecause the DLSS3 implementations
13:46SPEAKERaren't in shipping products yet.
13:47SPEAKER- Branch of drivers?
13:48SPEAKER- On Steam, on Steam.
13:50SPEAKERSo we were using pre-release drivers provided
13:52SPEAKERas part of the review kits.
13:54SPEAKERAnd we were, all of these games,
13:57SPEAKERincluding Cyberpunk, Plague Tale Requiem,
14:00SPEAKERFlight Sim, and F1 2022, or F122 rather,
14:05SPEAKERwere all tested with beta branches.
14:08SPEAKERSo presumably DLSS3 will be coming
14:11SPEAKERin the not too distant future.
14:12SPEAKERIs not shipping on anything as far as I can tell, yeah.
14:17SPEAKERDLSS3 fucking kicks ass.
14:20SPEAKER- Okay, that's interesting to hear
14:22SPEAKER'cause I have seen a lot of skepticism out there
14:24SPEAKERand I don't necessarily disagree
14:26SPEAKERabout the frame insertion in particular.
14:28SPEAKER- I'm going to tell you, I can't tell a difference.
14:31SPEAKERI am not super sensitive to artifacting generally,
14:35SPEAKERbut I do notice DLSS artifacts in DLSS2 games
14:38SPEAKERwhere you have like noise around chain link fences
14:41SPEAKERand stuff like that.
14:42SPEAKER- Yeah, I think in my case, and I think for most people,
14:45SPEAKERit's more latency that was the concern with frame insertion.
14:48SPEAKER- So the latency, the thing I will tell you is that
14:51SPEAKEReven when the DLSS3 stuff doesn't improve the frame rate
14:55SPEAKERas dramatically as it doesn't say Cyberpunk at 1080p,
14:58SPEAKERyou're still seeing an enormous reduction in microstutters.
15:03SPEAKERSo they target the 1%.
15:05SPEAKERSo there's something NVIDIA started paying attention to
15:07SPEAKERa few years ago when they, prior to Reflex coming out,
15:10SPEAKERbut around that time, which is like, what's the,
15:13SPEAKERwhat frame rate, what's the average frame rate
15:15SPEAKERfor the slowest 1% of frames?
15:18SPEAKERAnd that's a pretty good, like if there's a big delta
15:21SPEAKERbetween that and your average frame rate,
15:22SPEAKERthat's a pretty good indicator
15:23SPEAKERthat you have a lot of microstutters.
15:26SPEAKERSo what the DLSS3 stuff does is smooths out
15:29SPEAKERthose slowest 1% of frames and brings them back up
15:33SPEAKERcloser to the mean, which makes the whole game feel
15:36SPEAKERreally, really buttery smooth in a way that is difficult to,
15:40SPEAKERlike it doesn't necessarily show in benchmarks
15:42SPEAKERthe tools that we typically use to benchmark
15:45SPEAKERdon't register those inserted frames
15:47SPEAKER'cause they happen at a later in the pipeline
15:48SPEAKERthan the benchmarks register frame rates.
15:51SPEAKERAnd you have to use some special tests,
15:52SPEAKERspecial tools to see the difference,
15:56SPEAKERbut it feels noticeably different when you turn it on.
15:58SPEAKERLike-
15:59SPEAKER- That sounds pretty good.
16:00SPEAKER- For example, Cyberpunk, Cyberpunk, Cyberplunk.
16:04SPEAKER- Oh no, it's no, it's Cyberplunk from now on.
16:07SPEAKER- Cyberplunk now, when you're playing Cyberplunk
16:11SPEAKERat either 4K or 1080p with DLSS3 on
16:14SPEAKERand the ray tracing turned all the way up,
16:16SPEAKERit felt like you were not, it felt smooth in a way
16:20SPEAKERthat I've never felt a ray tracing game to play.
16:22SPEAKER- I was gonna say all of what you're describing,
16:24SPEAKERnot even just DLSS3, but even just the general perf
16:28SPEAKERover the 3080 Ti, like this is all sounding to me
16:30SPEAKERlike the first time, and hopefully this is also the case
16:33SPEAKERwith 4080s because the 4090 again is maybe not the card
16:36SPEAKEReverybody's going to get, but it's sounding like
16:39SPEAKERthe first time that you might actually turn on ray tracing
16:41SPEAKERas just a matter of course and just say, of course,
16:43SPEAKERI'm going to play the ray tracing mode in this game
16:45SPEAKERand not worry about the performance hit
16:47SPEAKERbecause like they've just been on the cusp for a while
16:51SPEAKERof like 50 FPS, I can't quite do it, it's almost there.
16:55SPEAKERLike if you could keep a locked 60, for example,
16:58SPEAKERon a given card, like, okay, I could do ray tracing,
17:00SPEAKERbut when you're dipping a little bit below that,
17:02SPEAKEReven with, you know, adaptive sync,
17:06SPEAKERthat have to refresh and all that stuff,
17:07SPEAKERlike it's just not quite there.
17:09SPEAKERSo if you're getting into those numbers with ray tracing,
17:12SPEAKERit's like, oh yes, I'll just use ray tracing then.
17:15SPEAKER- On the i9 at 4K, like the average frame rate
17:19SPEAKERin the benchmark was like mid fifties, 52, 50,
17:24SPEAKERsomething like that on my test with the DLSS3 turned on
17:27SPEAKERand on the performance setting.
17:29SPEAKER- Okay.
17:31SPEAKER- But it was solid and--
17:32SPEAKER- Wait, at what resolution?
17:33SPEAKER- At 4K.
17:34SPEAKER- Oh, okay, so I guess I'm--
17:36SPEAKER- But 4K with DLSS3 is an internal resolution
17:38SPEAKERof like 1080p scaled up, right, so.
17:41SPEAKER- I guess for me, I'm not trying to play 4K anyway,
17:44SPEAKERso that's--
17:45SPEAKER- Right.
17:46SPEAKER- I wonder what the market segmentation is at this point.
17:48SPEAKERHow many people are actually trying to game at 4K?
17:51SPEAKERI don't feel like 4K monitors
17:53SPEAKERare quite where they should be for me.
17:56SPEAKER- I mean, the ones that are, the ones,
17:57SPEAKERso for me, I would take higher refresh rate over 4K
18:01SPEAKERand to get 144 hertz 4K monitor,
18:04SPEAKERyou have to spend like 850 bucks right now.
18:06SPEAKER- Yeah, pricing and also like, depends on the OS,
18:09SPEAKERbut like UI scaling,
18:10SPEAKERI feel like it's still also not quite there universally.
18:13SPEAKER- Windows 11 is pretty, like, so I,
18:15SPEAKERmy setup is that I have a 4K monitor up top
18:18SPEAKERand a 1080p monitor down low.
18:20SPEAKERI play games on the 1080p monitor most of the time,
18:22SPEAKERbut occasionally I'll switch up for something single player.
18:25SPEAKERI'm pretty comfortable playing,
18:27SPEAKERlike I was comfortable playing Cyberpunk at 50,
18:29SPEAKERat the frame rate that I was getting
18:31SPEAKERbecause it was stable and it wasn't up and down.
18:33SPEAKERLike because there weren't those microstutters,
18:35SPEAKERit felt a lot smoother than,
18:39SPEAKERit felt smoother than what I associate
18:41SPEAKERa 50 frame per second average feels like.
18:43SPEAKER- Okay, and just again, to reiterate,
18:45SPEAKERbecause I don't have the numbers in front of me,
18:46SPEAKERthis was at 4K with DLSS3.
18:49SPEAKER- With DLSS3 and ray tracing cranked up all the way.
18:52SPEAKER- Okay, so not native 4K.
18:53SPEAKER- Not native 4K.
18:54SPEAKERI don't think anything ray tracing
18:56SPEAKERis gonna run native 4K for the time being.
18:58SPEAKER- Yeah.
18:59SPEAKER- So the takeaway, so, okay, so there were one, two, three,
19:05SPEAKERfour, five, four or five titles supported,
19:09SPEAKERsome tech demos supported, and a Korean RPG,
19:13SPEAKERor a Chinese MMORPG, I think was the fifth,
19:17SPEAKERwas the last game that was in the,
19:20SPEAKER"Hey, we can help you figure out how to test this stuff."
19:25SPEAKERMy assumption is that the number of titles
19:27SPEAKERthat support DLSS3 will explode
19:31SPEAKERin the same way that the DLSS support has exploded
19:33SPEAKERbecause the plugin as a developer
19:35SPEAKERis relatively simple to implement.
19:36SPEAKERYou basically, it's a drop-in for Unreal and Unity,
19:40SPEAKERand you generally don't have to do too much work
19:42SPEAKERto get it up and running.
19:43SPEAKERI think when we put it in the Anacrusis,
19:45SPEAKERit was a half day work for an engineer,
19:47SPEAKERand then a little bit of testing and tuning.
19:49SPEAKER- Jeez, that's pretty fast.
19:50SPEAKERAnd I assume that it's just three as a superset of two,
19:54SPEAKERlike it would just gracefully fall back,
19:55SPEAKERor more to the point, you do the work once,
19:57SPEAKERand it supports all versions based on the hardware.
20:00SPEAKER- That's it.
20:00SPEAKERIf you use the DLSS3 plugin,
20:03SPEAKERand somebody's using two series or three series GPUs,
20:07SPEAKERthey'll get DLSS2 support.
20:10SPEAKERIt also, DLSS3, we'll talk about how DLSS3 works later on,
20:15SPEAKERbut it also, that comes with reflex support now,
20:17SPEAKERbasically, because of the way the frame insertion works.
20:22SPEAKERSo DLSS3, the short version of DLSS3 is it inserts frames
20:27SPEAKERto fill in the lowest frame rate gaps,
20:30SPEAKERand it makes the whole thing feel
20:31SPEAKERlike you're running a lot faster.
20:33SPEAKERAnd in games that are like pure CPU bound game,
20:35SPEAKERlike games like Flight Sim,
20:37SPEAKERyou literally get a 2X performance boost
20:39SPEAKERfor turning on DLSS3.
20:41SPEAKER- That's pretty substantial.
20:41SPEAKERI mean, the concern, you tell me,
20:43SPEAKERat least my concern is, you're not sampling input
20:46SPEAKERon artificially created frames, right?
20:48SPEAKERYou're only sampling input for native frames,
20:50SPEAKERand so that seemed like it might cause some kind of weirdness
20:55SPEAKERwith just feel, control feel that you seem to-
20:57SPEAKER- I didn't notice it.
20:59SPEAKERI didn't play as much Flight Sim,
21:00SPEAKERwhich is a lower frame rate game
21:01SPEAKERin stuff like playing cyberpunk with a game pad
21:06SPEAKERor with a mouse and keyboard.
21:07SPEAKERI didn't notice any input lag,
21:09SPEAKERpartly because of, and part of the reason
21:12SPEAKERis they use the reflex, no screen buffer technology
21:17SPEAKERto make it, to reduce latency on the pipeline end
21:23SPEAKERas the CPU is feeding the video card.
21:28SPEAKERSo yeah, it's a trip, like the whole thing is a trip.
21:32SPEAKERIt's really weird talking about this card
21:35SPEAKERbecause it's a really, really fast card.
21:38SPEAKERI don't know that any hardware exists right now
21:40SPEAKERthat can take full advantage of this card in games,
21:42SPEAKERunless, like if you're playing at 8K, I guess,
21:44SPEAKERI don't know how many monitors are available.
21:45SPEAKERIt seems like not many.
21:47SPEAKER- Is there a game on the market
21:48SPEAKERthat has built-in support for running at 8K?
21:52SPEAKER- I mean, I think everything will run.
21:53SPEAKERI think you can set the resolution.
21:55SPEAKERIt's just, it'll probably, the games will just poll
21:58SPEAKERif they're set up right, we'll just poll the-
21:59SPEAKER- Well, very game dependent.
22:01SPEAKERSome of them are artificially constrained.
22:02SPEAKERSome of them, I don't feel like a lot of games
22:04SPEAKERjust pull from available OS resolutions
22:07SPEAKERat this point, do they?
22:08SPEAKER- I feel like they, mostly they have a subset
22:10SPEAKERof resolutions they support out of that polled list,
22:14SPEAKERbut I don't know.
22:16SPEAKER- I'm sorry, I found a page on Nvidia support.
22:19SPEAKER- Oh, how do you play games at 8K?
22:20SPEAKER- Entitled, recommended 8K titles.
22:22SPEAKER- I bet "Watch Dogs Legion" is on there.
22:25SPEAKERIs it? No, it is not.
22:28SPEAKER- Hmm.
22:29SPEAKER- "Dude Maternal," "Apex Legends," "Control,"
22:34SPEAKER"Genshin Impact," "Warframe," "Smite," "Left 4 Dead 2."
22:38SPEAKER- The "Left 4 Dead 2" will support every resolution.
22:40SPEAKERWhen you pull down source engine games,
22:43SPEAKERit gives you every resolution that the computer supports,
22:47SPEAKERthat Windows reports.
22:48SPEAKER- I want to play "World of Warcraft Classic" at 8K.
22:51SPEAKER- God.
22:52SPEAKER- Go back to my 2004 self and say, "Just look at this."
22:54SPEAKERJust put four 4K TVs in a quad,
22:57SPEAKERand then frame combine those, and you're good to go.
23:04SPEAKERI think even with new CPUs,
23:06SPEAKERraster games with these cards are gonna be CPU bound.
23:12SPEAKERThrough a comedy of errors,
23:14SPEAKERI had some 12th gen Intel hardware lined up to test with,
23:18SPEAKERand it just hasn't gotten here in time, so.
23:20SPEAKERUnfortunately, all of my testing was with the i9,
23:23SPEAKERwhich I think is actually kind of interesting
23:25SPEAKERbecause the i9 that I built was a 2080 Ti machine,
23:29SPEAKERI think, originally.
23:33SPEAKERThat's where the upgrade sweet spot is for this, I think.
23:36SPEAKERIf you're looking at these cards,
23:38SPEAKERif you're looking to upgrade your 2000 series
23:40SPEAKERor 1000 series NVIDIA cards,
23:42SPEAKERthis is the generation you want to jump to, right?
23:45SPEAKER'Cause there's actually gonna be
23:46SPEAKERsubstantial performance improvements.
23:48SPEAKERThe DLSS3 stuff is presumably gonna kick down
23:51SPEAKERto the lower end cards from the 4090,
23:53SPEAKERso it'll work the same way.
23:56SPEAKERAnd frankly, this card is kind of a pain in the ass.
23:58SPEAKERIt uses an enormous amount of power.
24:01SPEAKERI think its TDP is 450 something watts.
24:03SPEAKER- Yes, that's correct.
24:05SPEAKERSomething they said in the,
24:06SPEAKERI don't know if you can speak to this or not.
24:07SPEAKERI think other reviewers have tested this,
24:10SPEAKERthat they said in those sessions we were in
24:12SPEAKERthat they have worked to eliminate transient spikes
24:16SPEAKERon the power draw.
24:17SPEAKER- I am not set up to test transient spikes
24:19SPEAKERon the power draw.
24:20SPEAKERI feel like the 30 series cards were real bad
24:22SPEAKERabout exceeding TDP by a lot.
24:25SPEAKER- Yeah.
24:26SPEAKER- To the point that it's-
24:27SPEAKER- For instantaneous spikes, basically, yeah.
24:29SPEAKER- Just very short, but it doesn't matter how short it is.
24:32SPEAKERIf it's enough to put too much load on the power supply,
24:34SPEAKERthen you're screwed.
24:36SPEAKER- It's why they started shipping 3080 TIs
24:38SPEAKERwith three or four PCI express connectors.
24:46SPEAKERBut yeah, I didn't,
24:48SPEAKERso the people that I trust that did reviews
24:50SPEAKERdid not find that it was doing the transient spikes.
24:53SPEAKERI didn't see anything that would indicate that.
24:55SPEAKER- I haven't looked too closely at other reviews,
24:57SPEAKERbut the chatter I saw was that that stuff
24:59SPEAKERhas gotten smoothed out quite a bit, which is good.
25:01SPEAKER- It seems to be resolved.
25:02SPEAKERIt also is entirely possible that we're not hitting,
25:04SPEAKERthat no one is able to put a full load on this thing.
25:07SPEAKER- That does actually seem possible.
25:09SPEAKERI don't know how much you mentioned it.
25:11SPEAKER- The min spec for PSUs is 850 watts.
25:16SPEAKERIf you have a big Intel CPU,
25:19SPEAKERI'm gonna say you probably want 1,000 watts just to be safe.
25:22SPEAKER- Boy, as someone who just bought a power supply
25:24SPEAKERlike two, three months ago,
25:26SPEAKERI originally bought an 850 watt
25:28SPEAKERand B&H shipped me an open box, which I sent back.
25:32SPEAKERAnd because they didn't have another 850 watt in stock,
25:36SPEAKERI bought 1,000.
25:36SPEAKERAnd now I'm feeling like even that 1,000
25:38SPEAKERwas maybe low balling it.
25:40SPEAKER- I think the 1,000 is safe.
25:42SPEAKERLike the theoretical maximum that you're gonna pull
25:44SPEAKERfrom a 15 amp 120 volt circuit in the US is 1,600 watts.
25:49SPEAKER- Yeah.
25:50SPEAKER- And like that's gonna reduce by 20%
25:51SPEAKERbecause of the kind of overhead and heat
25:53SPEAKERand stuff like that.
25:54SPEAKER- Sure.
25:55SPEAKERWell, the other thing I'm guessing,
25:57SPEAKERyou're not living this life yet, right?
25:59SPEAKERBut I am pretty committed to undervolting hardware
26:02SPEAKERat this point.
26:03SPEAKER- I am not doing that on the current machine,
26:05SPEAKERbut I think if I build a Ryzen next,
26:06SPEAKERthat's probably in my future.
26:08SPEAKER- I mean, they built that right into the BIOS.
26:10SPEAKERHow many Ryzens have that eco mode
26:11SPEAKERwhere you can just straight up say like,
26:12SPEAKERhey, run this at 65 watts 'cause it's still gonna get you
26:15SPEAKERalmost all the, like it's crazy.
26:17SPEAKERThis race to the top that we've been seeing
26:19SPEAKERacross both CPU's and GPU's is like kind of unfortunate
26:22SPEAKER'cause I don't have the exact numbers,
26:24SPEAKERbut a little bit of the review stuff I was looking at,
26:26SPEAKERit seems like they're pumping triple digit watts
26:29SPEAKERinto these cards to achieve an extra like three to 5%
26:31SPEAKERjust for benchmark purposes.
26:34SPEAKERAnd like, who needs that?
26:35SPEAKER- That's the game, man.
26:36SPEAKER- Who needs that?
26:37SPEAKER- Well, so.
26:38SPEAKER- I mean, maybe this is my work from home talking,
26:40SPEAKERmy small business owner talking that I have to pay
26:44SPEAKERfor all of the electricity for my work myself now,
26:47SPEAKERtrying to keep it somewhat sensible.
26:49SPEAKER- It turns out that your power bill goes way up
26:51SPEAKERwhen you work from home and have a giant ass PC
26:53SPEAKERthat you play games and work on all day.
26:56SPEAKERThe thing that's interesting about that though,
26:59SPEAKERis that I think that the more than overclocking the GPU
27:04SPEAKERon this go round, being able to overclock the CPU
27:08SPEAKERand memory, system memory are probably gonna be
27:10SPEAKERthe performance bottlenecks on these cards,
27:14SPEAKERat least for the current gen hardware.
27:16SPEAKER- Just in terms of being able to feed the card fast enough.
27:18SPEAKER- Like if you want the card to run it,
27:19SPEAKERlike if you wanna get better performance out of a machine
27:22SPEAKERthat has one of these cards,
27:23SPEAKERyou're probably gonna be better off overclocking your CPU
27:25SPEAKERand your RAM than overclocking the GPU.
27:30SPEAKERThe card's a big chunky boy.
27:32SPEAKER- Yeah, I wanted to ask about physical installation as well.
27:35SPEAKER- It's three slots wide instead of two,
27:38SPEAKERwhich is different than the 30 90s were two slots wide,
27:41SPEAKERbut they were very long and very kind of tall.
27:44SPEAKER- When you say wide, you mean tall?
27:47SPEAKER- No, I mean, it's three, like it fills three slots.
27:50SPEAKER- That's very cool.
27:53SPEAKER- Well, but-
27:53SPEAKER- I'm being very pedantic here.
27:55SPEAKER- Yeah, no, I mean, the width is how many slots
27:57SPEAKERit takes up in the case.
28:00SPEAKERThe height is how far it is from the motherboard.
28:02SPEAKER- I know, I mean, in a tower case.
28:02SPEAKER- Just because you're living in a tower world
28:04SPEAKERdoesn't mean that that's the way the world works.
28:06SPEAKER- Who's not using a tower car or-
28:08SPEAKER- Yeah, but anyway, okay, so it's not as,
28:10SPEAKERit is shorter than the 30 90s by,
28:13SPEAKERlike it fits in an H510,
28:15SPEAKERwhich the 30 90s theoretically don't.
28:17SPEAKER- Oh, that is the opposite of what I expected to hear.
28:19SPEAKER'Cause usually length is usually the real problem, right?
28:23SPEAKER'Cause people have got drive cages
28:25SPEAKERin various cases and stuff.
28:26SPEAKERLike that's usually the thing that stops you
28:27SPEAKERfrom installing a gigantic card.
28:29SPEAKER- Yeah, it's not a problem.
28:32SPEAKERIf it fit in my H510 nicely,
28:34SPEAKERI was kind of surprised.
28:35SPEAKERI had to, I did turn up the amount of air
28:38SPEAKERthat comes in on the intakes on both the front
28:40SPEAKERand the side though, at the top.
28:43SPEAKER- It's probably too early to ask about sag.
28:46SPEAKER- I-
28:47SPEAKER- Oh, he's rolling under the desk.
28:49SPEAKER- I'm looking.
28:50SPEAKERIt hasn't, a friend of the pod Wedge
28:54SPEAKERasked about this the other day,
28:56SPEAKERand I have not seen any sagging yet.
28:58SPEAKERI do not have anything bracing it,
28:59SPEAKERbut I think that's probably in the future.
29:03SPEAKERI mean, that's straight up-
29:04SPEAKER- Ultimately, I'll do a vertical install.
29:06SPEAKER- Yeah, that's straight up a product category now.
29:08SPEAKERYou can just get PCIe slot bracers
29:12SPEAKERfor 15 bucks to hold up your goddamn 20 pound graphics card.
29:17SPEAKER- Yeah, it's like the box and stuff for it is ridiculous.
29:20SPEAKERIt comes into like this big ridiculous
29:22SPEAKERunfolding presentation case that has,
29:25SPEAKERlike it's mounted diagonally in between this big cube.
29:27SPEAKERSo it takes up like three times
29:28SPEAKERas much space as a box than you need.
29:32SPEAKERAny differences in the cooler?
29:33SPEAKERI mean, it's pretty much the same as the Ampere cards, right?
29:37SPEAKER- It's the blow through cooler
29:38SPEAKERwith the V cutout on the actual card
29:40SPEAKERand like fans on both sides,
29:42SPEAKERone blowing through and one blowing across,
29:44SPEAKERor one blowing out.
29:47SPEAKERThe cooler seems quieter.
29:49SPEAKERThe 3080 Ti Founders Edition cards
29:53SPEAKERwere not particularly loud.
29:55SPEAKERLike even under full load, it didn't blow that hard.
30:01SPEAKERSo this seems comparable to me.
30:03SPEAKERI mean, under the desk, I don't hear it,
30:06SPEAKEReven when I'm playing a ray trace game at full bore.
30:11SPEAKER- Since we're talking about card form factors,
30:12SPEAKERbefore we move on,
30:13SPEAKERany thoughts about the AIB cards
30:15SPEAKERthat I assume you've seen photos of out there?
30:17SPEAKER- The big third party ones?
30:19SPEAKER- Have you seen that ASUS one in particular,
30:21SPEAKERthe side-by-side with the Founders Edition
30:23SPEAKERthat is like three times the size?
30:25SPEAKER- It is really big.
30:26SPEAKER- What is it at?
30:27SPEAKERWho has a case that could fit that?
30:29SPEAKER- I mean, full towers, man.
30:31SPEAKEREven that, I don't know. - People do full towers.
30:33SPEAKER- The thing is, it is terrifying.
30:35SPEAKER- I was gonna say, so the other thing about this
30:38SPEAKERis that there's a new power connector.
30:40SPEAKER- Yeah, that's another thing that like,
30:42SPEAKERI mean, it's not a big deal,
30:45SPEAKERbut they went and announced that PCIe 5 power connector
30:48SPEAKERlike the fucking month after, I gotta take that back.
30:52SPEAKERI went back and looked.
30:52SPEAKERThey announced that connector in like February.
30:55SPEAKERBut Seasonic went and announced a new model of power supply
30:58SPEAKERwith that connector on it,
30:59SPEAKERlike a week after I bought my new power supply.
31:01SPEAKER- Well, I mean, so if you have a modular supply,
31:04SPEAKERit's not a big deal,
31:05SPEAKER'cause you can get a modular connector from your vendor.
31:08SPEAKERLike, Coursera already has one listed.
31:10SPEAKERI don't think they're shipping them yet.
31:11SPEAKER- It's, what is it, 12-pin, I wanna say?
31:14SPEAKER- It's, I actually don't remember this.
31:17SPEAKERThe thing is, it's smart.
31:18SPEAKERSo it has load, it has power load monitoring
31:20SPEAKERthat comes back into the power supply.
31:22SPEAKER- And that's built into the spec?
31:23SPEAKER- And that's built into the spec.
31:24SPEAKER- Wow, that's crazy.
31:26SPEAKERI didn't know about that.
31:27SPEAKER- The one, Nvidia ships an adapter with the card
31:30SPEAKERthat plugs into four standard PCIe power connectors.
31:33SPEAKERYou know, the-
31:34SPEAKER- Four?
31:35SPEAKER- Four.
31:36SPEAKER- The 8-pin, there's a-
31:37SPEAKER- The 8-pin one.
31:38SPEAKERIt's the standard 8-pin one.
31:40SPEAKERIt's annoying because they literally just rolled out
31:43SPEAKERa new power connector in the last generation
31:46SPEAKERthat was no longer sufficient for this thing.
31:49SPEAKER- Like, I went and got, I mean, it was free.
31:52SPEAKERIt was free from my power supply vendor,
31:53SPEAKERbut I went and got one of those 12-pin adapters for,
31:56SPEAKERI assumed that was the new Nvidia connector of record-
31:58SPEAKER- Me too.
31:59SPEAKER- For some generations to come and used it for one.
32:04SPEAKERIt's fine, it's fine.
32:05SPEAKER- I spent money on a cable.
32:06SPEAKERI regret it.
32:10SPEAKERSo, okay, is it worth buying?
32:13SPEAKERWe said this before, the founder's edition is 1,600 bucks.
32:16SPEAKERThe add-on board are even more expensive in a lot of cases.
32:20SPEAKERThis is profoundly not a card
32:22SPEAKERthat is for people who play games,
32:23SPEAKERat least not stuff that's currently available.
32:25SPEAKERBut maybe if you really are into ray-traced games.
32:27SPEAKER- Sure, sure.
32:28SPEAKERI mean, I'm not sitting here running the numbers.
32:30SPEAKERIt sounds like performance per dollar
32:32SPEAKERis pretty damn good on this thing
32:34SPEAKERcompared to the previous generation.
32:35SPEAKERBut the bigger question is,
32:36SPEAKERdo you need that much performance?
32:38SPEAKERAnd it sounds like most people don't.
32:39SPEAKER- I was gonna say, there's a couple of gotchas.
32:41SPEAKERThe card only supports DisplayPort 1.4, DisplayPort 1.4,
32:45SPEAKERwhich gives you like 8K at 60 Hertz,
32:49SPEAKERor anything over 4K is gonna be a 60 Hertz refresh rate.
32:53SPEAKER- That's, you know-
32:54SPEAKER- Without compression.
32:56SPEAKER- I'm not sweating that so much.
32:58SPEAKER- I mean, I went and looked,
32:59SPEAKERI couldn't find DisplayPort 2.0 monitors
33:01SPEAKERon the market right now.
33:03SPEAKERIt seems like it probably matters more
33:05SPEAKERfor like high-end VR headsets.
33:07SPEAKER- Sure.
33:08SPEAKER- Because they actually, high refresh rates,
33:09SPEAKERhigh resolution, multiple panels will do more bandwidth
33:13SPEAKERthan like a 4K at 144 Hertz monitor.
33:16SPEAKER- Are there VR headsets plugging into DisplayPort?
33:19SPEAKER- Pimax and, yeah, some of them are, yes.
33:21SPEAKER- Oh, wow.
33:22SPEAKER- The index plugs into, most of the PC,
33:27SPEAKERexcept for the Windows Mixed Reality headsets,
33:31SPEAKERmost of them are DisplayPort.
33:32SPEAKER- Oh, I can tell you how long it's been
33:34SPEAKERsince I saw a new headset in person.
33:36SPEAKER- Yeah.
33:38SPEAKERThey, it's a problem for the future.
33:41SPEAKERIt feels like an omission on a card
33:44SPEAKERthat's like top of the line, super expensive,
33:46SPEAKERespecially since I think AMD is chipping DisplayPort 2.0
33:49SPEAKERon hardware. - Yeah, I can see it.
33:50SPEAKERI can see it.
33:52SPEAKER- The big question I have after this whole thing
33:54SPEAKERis like, why is Nvidia not leading
33:56SPEAKERwith the more reasonably priced cards
33:58SPEAKERand why are they kicking the big boy out first?
34:01SPEAKER- Yeah.
34:02SPEAKER- Do we need to-
34:03SPEAKER- I have some theories.
34:04SPEAKERWe can theory craft this. - We need to run those down.
34:06SPEAKERI mean, do we need to run those down?
34:07SPEAKERBecause like they're, they announced three SKUs.
34:09SPEAKERI'm sure everybody knows at this point.
34:11SPEAKERThere are two 4080s,
34:12SPEAKERwhich I think it's the first time that's ever happened
34:14SPEAKERat launch.
34:15SPEAKER- I mean, they've,
34:16SPEAKERso there's always been variation in the 4080s
34:18SPEAKER'cause the AIB partners will like clock the memory
34:21SPEAKERdifferently and stuff like that.
34:22SPEAKER- Oh yeah, sure.
34:23SPEAKERBut I mean, I mean, coming straight from Nvidia,
34:24SPEAKERalthough Nvidia selling its own cards is in and of itself,
34:27SPEAKERa relatively new phenomenon.
34:28SPEAKERSo that's-
34:29SPEAKER- New recent generations, yeah.
34:30SPEAKER- And also like this sure does dovetail
34:32SPEAKERwith that EVGA episode we did a few weeks ago.
34:35SPEAKER- What a weird coincidence.
34:37SPEAKER- That these founders' cards for the beta generation
34:40SPEAKERare quite nice.
34:41SPEAKERAnd then you've got this like monstrosity coming from
34:44SPEAKERa major third party partner.
34:46SPEAKERYou know what I mean?
34:47SPEAKERLike where are we headed with this?
34:49SPEAKERLike, are we actually gonna get to the point where,
34:51SPEAKERare they just gonna buy up a couple of AIBs
34:53SPEAKERand just start selling all the cards themselves?
34:55SPEAKER- I don't think they need to buy the AIBs.
34:57SPEAKER- Well, no, but then they have to do all the work
34:59SPEAKERto establish the supply chain themselves
35:00SPEAKERis what I'm saying.
35:01SPEAKER- I think they're probably,
35:02SPEAKERI think they've been gradually working toward that
35:04SPEAKERthis whole time.
35:05SPEAKER- They may well be.
35:06SPEAKER- I think that's why EVGA bailed out of the market
35:07SPEAKERprobably in retrospect.
35:09SPEAKER- Well, sure, but you know,
35:10SPEAKERbuying a couple of partners with established assembly lines
35:13SPEAKERcertainly speeds that process.
35:15SPEAKER- Maybe.
35:17SPEAKERSo yeah, the 24080s are the only other cards
35:20SPEAKERthat they've announced.
35:20SPEAKERThey're not gonna be available until late this year
35:22SPEAKERor early next year, I think.
35:25SPEAKER- Those come out next month.
35:26SPEAKER- Don't do it, okay.
35:27SPEAKERBut I mean, realistically you're not gonna get them
35:30SPEAKERuntil next year.
35:31SPEAKER- Oh, I wonder if you could, yeah, hang on.
35:33SPEAKER- So 4090s sold out pretty fast.
35:34SPEAKERThey went on sale two days ago.
35:36SPEAKER- I was gonna see.
35:37SPEAKER- And like even the local screwdriver shops don't have,
35:41SPEAKERthe usually central computer would have something in stock
35:43SPEAKERand they'd just mark it up real high.
35:44SPEAKERThey didn't have anything 4090ish in stock
35:48SPEAKERthat I saw this morning or yesterday.
35:50SPEAKERBut they didn't sell out instantly.
35:52SPEAKERYou could go to Micro Center and get one
35:54SPEAKERat the end of the day on Wednesday in some places.
35:57SPEAKERFriend of the Stream JP did that, I believe.
36:00SPEAKER- Oh, interesting.
36:02SPEAKER- He also is the only streamer I know
36:04SPEAKERwho plays games at 4K, so.
36:05SPEAKER- Sure, he's a tech guy for sure.
36:08SPEAKER- Yeah, so he's living large right now.
36:12SPEAKER- Were you not in the rarefied position
36:14SPEAKERto be getting a video card straight from Nvidia?
36:18SPEAKER- Yeah.
36:19SPEAKER- Or were you out there buying one?
36:20SPEAKERDo you think you'd be at the point yet
36:21SPEAKERwhere it would be Founders Edition or nothing?
36:23SPEAKERWould you be like trying to get that one
36:25SPEAKERbecause it's the--
36:26SPEAKER- I tend to value reliability over other things.
36:32SPEAKER- Yeah.
36:33SPEAKER- So I would probably do the,
36:34SPEAKERI mean, look, the Founders Edition cards have had problems.
36:36SPEAKERI wanna say the 1000 series Founders Edition cards
36:39SPEAKERhad issues where they had to--
36:41SPEAKER- Were there 1000 series?
36:44SPEAKERI thought--
36:45SPEAKER- Maybe it was 2000.
36:46SPEAKERThe first couple of generations of Founders Edition cards
36:48SPEAKERboth had defects that required them to fix.
36:51SPEAKER- Interesting.
36:53SPEAKER- Not like massive, it wasn't like everything.
36:55SPEAKERIt wasn't like an Xbox 360 where everything died eventually.
36:57SPEAKERIt was like some noticeably large percentage had failures.
37:02SPEAKER- Okay, I'm still using a founder,
37:04SPEAKERa launch Founders Edition 3080,
37:06SPEAKERand it seems to have held up.
37:08SPEAKERI don't know of any--
37:09SPEAKER- I would expect that to be fine.
37:11SPEAKER- Any widespread issues with that generation?
37:13SPEAKER- Yeah, I would be shocked if that had problems.
37:14SPEAKER- I'm fascinated by this transition
37:16SPEAKERthat appears to be going on
37:17SPEAKERwhere Nvidia seems to be trying to
37:19SPEAKERrest some more control of the market from its partners.
37:22SPEAKER- I mean, it's a weird situation.
37:25SPEAKEROkay, so that's a good segue though.
37:28SPEAKER'Cause like, I think that that's one of the reasons
37:30SPEAKERthat I think they launched with the,
37:31SPEAKERlike I've been thinking about this,
37:34SPEAKERlike this is the thing I've been thinking about a lot.
37:35SPEAKERIt's why did they lead with the 4090?
37:36SPEAKER'Cause like leading with the $1600 card
37:39SPEAKERis gonna piss off a lot of people.
37:42SPEAKERThey're not gonna sell a massive volume in this,
37:44SPEAKERbut the margin on this card is probably pretty good.
37:46SPEAKERIt, I don't understand,
37:51SPEAKERthey usually lead with the lower priced cards
37:53SPEAKERbecause that's where the fat part of the market is
37:54SPEAKERand they wanna get those out as soon as possible.
37:56SPEAKERSo they didn't launch with,
37:58SPEAKEReither they still have a lot of 3080s in the supply chain
38:01SPEAKERand they don't wanna hose all the ad and board partners
38:03SPEAKERand have to buy back a bunch of chips
38:05SPEAKERor bail people out or whatever,
38:06SPEAKERwhich is I think probably part of it, right?
38:08SPEAKER'Cause everybody overproduced 3080s
38:10SPEAKERbecause of the crypto glut.
38:12SPEAKERAnd then when Ethereum switched to proof of stake
38:16SPEAKERrather than proof of work,
38:17SPEAKERthe market for high-end GPUs
38:20SPEAKERto do crypto shit evaporated overnight.
38:25SPEAKERI, one of the things that you and I talked about before
38:28SPEAKERthat I've kind of theorized
38:29SPEAKERis that maybe the 4080 will actually be faster
38:31SPEAKERbecause they're clocking things higher,
38:35SPEAKERclocking CPU and memory higher
38:36SPEAKERbecause they're, the 4090,
38:40SPEAKERthe 90 series cards are typically about width and not speed.
38:43SPEAKERSo it's about doing massive parallel math,
38:46SPEAKERmaybe at a slightly slower speed.
38:48SPEAKERSo it's the equivalent of having 32 cores in your CPU
38:52SPEAKERthat run at 75% of the speed
38:55SPEAKERversus having 16 cores in your CPU
38:56SPEAKERthat run at 100% of the speed.
38:59SPEAKER- Right.
39:00SPEAKER- Looking at the specs on paper,
39:02SPEAKERassuming they shipped the specs that they told us,
39:05SPEAKERI don't think that's the case
39:06SPEAKERbecause the 4080 has reduced both RT
39:13SPEAKERand raster workload performance
39:15SPEAKERbased on stuff like the tensor and RT cores,
39:18SPEAKERthe number of tensor and RT cores for ray tracing workloads
39:21SPEAKERand the memory bandwidth and texel fill rates
39:25SPEAKERand pixel fill rates based on the specs that they gave.
39:30SPEAKERI think it's possible maybe that DLSS3 is good enough
39:34SPEAKERthat it makes the low end cards too good.
39:37SPEAKERSo like if these cards,
39:39SPEAKEReven at the lower specs are still CPU bound
39:42SPEAKERon all existing hardware, most existing hardware,
39:46SPEAKERthen DLSS3 comes into play
39:47SPEAKERand that all of a sudden means
39:48SPEAKERthat you're gonna see roughly the same performance
39:52SPEAKERon the same CPU out of all the ray tracing benchmarks,
39:56SPEAKERwhich is great for people who are buying lower end cards,
39:59SPEAKERbut doesn't make this card
40:00SPEAKERlook particularly appetizing at $1,600.
40:02SPEAKER- Okay, that sounds like a plausible reason
40:05SPEAKERto put this thing out and nothing else next to it.
40:07SPEAKER- Yeah.
40:08SPEAKER- For a month and try to sell these while you can
40:10SPEAKERbefore people realize, oh, I should just get one of those.
40:13SPEAKER- The other options are that it's competitive stuff.
40:16SPEAKERSo maybe AMD and Intel,
40:18SPEAKERmaybe they were worried about either what AMD or Intel are.
40:21SPEAKERI mean, Intel is launching new GPUs
40:23SPEAKERfor the first time in 10 years, 15 years.
40:27SPEAKERAMD has new hardware coming.
40:29SPEAKERMaybe they were worried that they're not gonna be,
40:30SPEAKEReither won't be competitive at one of the price points
40:32SPEAKERor just wanted to cut the knees out
40:33SPEAKERfrom under whatever those offerings are,
40:36SPEAKERwhich is a very Nvidia move.
40:38SPEAKER- Hard to read the tea leaves right now.
40:39SPEAKERLike it sounds like RDNA3 is going to be pretty competitive
40:42SPEAKERin everything except maybe ray tracing.
40:44SPEAKER- Yeah, and ray tracing is really a tricky spot right now
40:47SPEAKER'cause there's what, a half dozen games that are coming out,
40:52SPEAKERright, that are ray traced games probably this year.
40:55SPEAKERI think we're gonna start to see more
40:56SPEAKERas Unreal Engine 5 games start to ship later this year.
41:00SPEAKER- Probably not wrong.
41:01SPEAKER- Sorry, later in 2023, not later in 2022.
41:03SPEAKER- Do you agree that the consoles launching when they did
41:06SPEAKERis probably going to hamstring ray tracing adoption
41:09SPEAKERfor kind of the entirety of this generation?
41:11SPEAKERI mean, those consoles barely do ray tracing at all.
41:13SPEAKER- I think it, hmm.
41:14SPEAKER- My feeling since like the 360 era
41:16SPEAKERis that kind of graphics and tech features
41:19SPEAKERlargely trickle down from the consoles
41:21SPEAKER'cause that's where the mass market is.
41:22SPEAKERI mean, like obviously you can do extra work
41:24SPEAKERon your PC ports, but the core of your pipeline,
41:28SPEAKERof your like asset pipeline tech,
41:30SPEAKEReverything is going to be kind of,
41:32SPEAKERin a cross-platform game is gonna be defined
41:34SPEAKERby the lowest common denominator that sells the most, right?
41:37SPEAKER- I think the gate is if the current gen PS5
41:40SPEAKERand Xbox series are fast enough
41:43SPEAKERto run global illumination stuff
41:44SPEAKERversus the pre-rendered lighting.
41:48SPEAKER- So my gut feeling is they are not.
41:51SPEAKER- I don't know.
41:52SPEAKERI think it's gonna require,
41:53SPEAKERI think that we're going to see
41:56SPEAKERa fairly typical console adoption curve on that stuff
41:59SPEAKER'cause it's gonna require, it's like,
42:02SPEAKERso global illumination,
42:03SPEAKERinstead of using pre-rendered light maps
42:04SPEAKERto light the vast majority of the map
42:05SPEAKERand having a handful of dynamic map lights in the map,
42:08SPEAKERyou just basically do everything dynamically
42:10SPEAKERand have base lighting for the whole level established.
42:13SPEAKER- It's essentially using ray tracing
42:14SPEAKERfor your entire generalized lighting model.
42:17SPEAKER- Kind of.
42:18SPEAKER- And I don't think, I could be wrong,
42:20SPEAKERmaybe as development matures on the console hardware,
42:25SPEAKERthey'll get there, but the early indications
42:27SPEAKERwere that that performance is not strong.
42:29SPEAKER- Well, it requires you starting the game,
42:32SPEAKERlike if you're doing a game
42:33SPEAKERthat takes place entirely outside,
42:34SPEAKERit's less of a problem
42:35SPEAKER'cause the sun is a fairly fixed lighting source
42:38SPEAKERand you're dealing with a base level of brightness.
42:40SPEAKERThis is why Call of Duty
42:41SPEAKERis able to put shadows on in ray tracing.
42:43SPEAKERIt doesn't really impact performance all that much, right?
42:46SPEAKERI think it's,
42:48SPEAKERmy guess is somebody is going to figure something out
42:52SPEAKERbut in order to do that,
42:53SPEAKERyou're gonna have to start the game with,
42:55SPEAKERhey, this is a global illumination game from day one
42:58SPEAKERand then every choice that you make is based on that
43:01SPEAKERand the performance you're gonna get on that
43:02SPEAKER'cause it's gonna change everything,
43:04SPEAKERevery aspect of the art creation
43:05SPEAKERthroughout the entire process of making the game.
43:07SPEAKER- That's entirely possible.
43:08SPEAKERYeah, we might very well see the Doom 3 of these consoles
43:13SPEAKERin terms of a game that just pushes
43:15SPEAKERthe state of the art forward in a dramatic way.
43:17SPEAKER- Yeah, it's not,
43:19SPEAKERthe thing about global changing your lighting model
43:22SPEAKERlike that is it's not,
43:24SPEAKERlike I said, DLSS was a half day of work for an engineer
43:26SPEAKERand then a little bit of testing.
43:28SPEAKERGlobal, switching to global illumination
43:30SPEAKERrequires a complete redo of your art pipeline
43:33SPEAKERand your rendering pipeline for the game.
43:35SPEAKERSo it is a foundational decision
43:39SPEAKERthat you make at the start of a project,
43:40SPEAKERnot something you tack on at the end.
43:41SPEAKER- Right, also I think like UE5
43:43SPEAKERthrows a whole wrench into this thing, right?
43:45SPEAKERBecause Lumen is not ray tracing based at all.
43:47SPEAKER- Lumen is buck wild and it's, who knows what,
43:51SPEAKERlike Lumen is weirdly CPU based.
43:53SPEAKERLike there's a bunch of weird choices there.
43:55SPEAKERSo, and then the last thought
43:59SPEAKERis that if it's not competitiveness,
44:02SPEAKERit's not that DLSS is too good
44:03SPEAKERand the perf on the lower end cards will be too good.
44:06SPEAKERIf it's not that there's too many 3080s,
44:08SPEAKERit's that they think that there's pent up demand
44:10SPEAKERbecause people haven't been able to buy GPUs
44:12SPEAKERand they can hit people for $1,600,
44:16SPEAKERa bunch of $1,600 cards,
44:17SPEAKER'cause that's what you were paying for a 3080
44:19SPEAKERlike six months ago.
44:20SPEAKER- Yeah.
44:23SPEAKER- And if like, if that was your choice,
44:24SPEAKERif you wanna spend $1,600 on a video card
44:27SPEAKERand you waited to buy a 3080 and you're buying this,
44:30SPEAKERyou did pretty well 'cause this card fucking rips
44:32SPEAKERcompared to a 3080.
44:33SPEAKER- Sure.
44:35SPEAKERAny thoughts on the 24080 SKUs having different GPUs on them?
44:39SPEAKERI mean, like by product code, they are AD103 and AD102.
44:44SPEAKERAnd when you have two different cards,
44:46SPEAKERthey're both called 4080 and on the box,
44:49SPEAKERthe only distinguishing factor is the amount of RAM.
44:51SPEAKERYou kind of assume like,
44:51SPEAKERoh, this is probably the same card otherwise,
44:53SPEAKERbut that is extremely not the case.
44:55SPEAKER- So remember when we talked about, okay,
44:58SPEAKERI think we'll find out when benchmarks land.
45:01SPEAKER- Yeah.
45:02SPEAKER- Like there's a little bit of a dip in RT cores
45:05SPEAKERbetween the 12 gig and the 16 gig model.
45:08SPEAKEROh, hold on, I forgot one other person.
45:10SPEAKERIf you do video editing
45:11SPEAKERand you're working with AV1 especially,
45:13SPEAKERthis card is a no brainer.
45:15SPEAKER- Yeah, or even streaming to Twitch.
45:18SPEAKERI really wish I had jumped in and asked them
45:20SPEAKERin those editor sessions
45:22SPEAKER'cause they spent quite a bit of time
45:23SPEAKERtalking about how great AV1 was going to be for streaming.
45:27SPEAKER- There's no support on Twitch yet for AV1.
45:29SPEAKER- And nobody supports AV1 ingest yet.
45:32SPEAKERI mean, supports anything, but H.264 still.
45:35SPEAKER- OBS doesn't support output yet.
45:36SPEAKERSo like the meta support is out now, I think,
45:39SPEAKERbut it's not in mainline until later this year.
45:41SPEAKER- Is that right?
45:42SPEAKEROkay. - Yeah.
45:43SPEAKER- Well, so, but also the ARC cards support AV1 encoding,
45:46SPEAKERlike so do the new AMD cards that are coming,
45:49SPEAKERlike everything that's coming out this year.
45:51SPEAKERIf you buy any of those cards, I believe,
45:53SPEAKERunless maybe some of the low end models don't have it,
45:55SPEAKERbut it sounds like the industry finally
45:59SPEAKERis moving on to a new codec.
46:01SPEAKER- But it's time.
46:02SPEAKER- Yeah, it's way past time.
46:04SPEAKERI found, I guess I won't name a shame.
46:08SPEAKERI found a test stream on Twitch.
46:11SPEAKERThe person who owns the channel
46:12SPEAKERis one of Twitch's like top video engineering,
46:16SPEAKERvideo codec engineers.
46:18SPEAKERI mean, it says it right in his bio.
46:19SPEAKERHe's like, "Hey, I'm Twitch's video codec guy."
46:21SPEAKER- Yeah.
46:22SPEAKER- And it's an AV1 stream.
46:24SPEAKERIt's from like two years ago, it's just sitting there.
46:26SPEAKER- Well, that's not-
46:26SPEAKER- No, no, it's fucked up.
46:27SPEAKER- But he gets to keep stuff for two years?
46:29SPEAKER- I guess so, I'm sure there are some,
46:31SPEAKERprobably some special flags enabled on this account,
46:33SPEAKERbut like the archive of the stream is sitting there
46:36SPEAKERand you can go into the settings on this stream
46:37SPEAKERand he plays like "Call of Duty" and "Fortnite" and "Apex"
46:42SPEAKERand maybe one or two other things, "Tomb Raider," I think.
46:45SPEAKERBut you can go into the settings
46:46SPEAKERand like in the typical settings pop up on the Twitch player,
46:51SPEAKERthere are settings like 1440p 120 AV1.
46:56SPEAKER- Wow.
46:57SPEAKER- So you can go see the-
46:58SPEAKER- So the player's there, they just don't have to code it.
46:59SPEAKERThey don't have the transcode backend for it, I bet, yeah.
47:02SPEAKER- The player supports it, you can go,
47:03SPEAKERbut you can get a sense of what it's gonna be like
47:06SPEAKER'cause you can go watch a 1440p 120 stream
47:09SPEAKERand it looks pretty damn good for Twitch bit rates.
47:11SPEAKER- Well, so that's the benefit of AV1 is that you get about,
47:15SPEAKERit's about 40% more efficient
47:18SPEAKERand it's not encumbered by patents.
47:19SPEAKERSo you can use it without having to pay anybody.
47:22SPEAKERSo the upshot is you get 1440p at current 1080p bit rates
47:26SPEAKERand equivalent image quality,
47:28SPEAKERor you do 1080p with much higher IQ at the current bit rates.
47:31SPEAKER- Massively better quality.
47:32SPEAKER- Which sounds great, I'm all for that.
47:34SPEAKER- Yes.
47:35SPEAKER- Okay, so we were talking about
47:36SPEAKERwhether you should buy this.
47:40SPEAKERLike it's a weird card.
47:44SPEAKERBefore I spent $1,600 on this,
47:46SPEAKERI would wait to see what the 4080s look like, honestly.
47:49SPEAKER- For sure.
47:50SPEAKER- 'Cause the 4080 price points,
47:51SPEAKEReven with the weirdness of the two different dies.
47:54SPEAKERSo you'd asked about the two different dies,
47:57SPEAKERit's weird to have the two different dies
48:00SPEAKERwith different numbers of shader cores and tensor cores
48:03SPEAKERand stuff like that.
48:05SPEAKER- 'Cause this is essentially the segmentation
48:07SPEAKERyou'd normally see between either a 4080 and a TI
48:11SPEAKERor a 4080 and a 4070.
48:13SPEAKERI don't know which direction that's going here,
48:15SPEAKERbut which version of that do you think this is?
48:18SPEAKER- Well, usually the TI is a mid-cycle bump.
48:21SPEAKER- Oh, I know, yeah.
48:22SPEAKER- My guess is that this would be the 4070 and the 4080,
48:26SPEAKERbut it also may be that at the resolutions
48:29SPEAKERpeople play games at, there's no difference in perf
48:31SPEAKERwith the changes in these dies.
48:33SPEAKERSo this may be a binning issue where they're just like,
48:36SPEAKERwhere if they're having yield problems
48:37SPEAKERand there's a certain number of RT cores or whatever
48:39SPEAKERper die that don't work, they just shut those off
48:42SPEAKERand they sell it as the 12 gig part.
48:44SPEAKERMy guess is that this 12 gig, 16 gig segmentation
48:47SPEAKERhas to do with memory controllers.
48:49SPEAKERAnd this is probably the reason EVGA
48:52SPEAKERisn't building these cards.
48:54SPEAKER- Interesting. - Because it no longer,
48:55SPEAKERit doesn't give them the ability
48:56SPEAKERto change memory configuration.
48:59SPEAKERAnd the fact that they're on GDDR6
49:02SPEAKERand there's not a lot of overhead
49:03SPEAKERbeyond what they're shipping on the founder's editions
49:05SPEAKERmeans that they can't make a better card.
49:07SPEAKERThey couldn't make a better card, you know.
49:10SPEAKER- So presumably if they're building an AD-104,
49:16SPEAKERwhich is the 12 gig 4080 die,
49:19SPEAKERthey can't make a 16 gig version of that
49:20SPEAKER'cause of the way the memory controllers work.
49:22SPEAKER- Yes, 102, AD-102 is the 12 gig 4080.
49:29SPEAKER- Oh, see, no, AD-102 is the 4090.
49:31SPEAKERAD-104 is the 12 gig 4080. - Oh, I'm sorry.
49:34SPEAKERThis image you put in the show notes is so scrunched.
49:36SPEAKER- It's backwards and small. - Eight looks like a nine.
49:37SPEAKERI was trying to, okay, these numbers are very hard to read.
49:39SPEAKER- Yeah, it's all good.
49:43SPEAKER- One of the things I thought was the most interesting
49:44SPEAKERin the presentation that we sat in on
49:46SPEAKERwas that they went out of their way to emphasize
49:48SPEAKERthat the 3080 is going to remain
49:50SPEAKERpart of their product offering.
49:51SPEAKER- And my guess is that the 3080 is going to be
49:54SPEAKERthe 4070 equivalent for the current run.
49:58SPEAKER- That's exactly what it, that's how that read to me.
50:00SPEAKERAnd again, that's largely probably
50:02SPEAKERbecause they overproduced so many
50:04SPEAKERand they need to sell them.
50:04SPEAKERBut also, but it's still notable
50:07SPEAKERbecause I don't think they've ever done that before, right?
50:08SPEAKERDon't they typically do the Apple style clean sweep
50:11SPEAKERof the old product line and just say, okay,
50:13SPEAKERhere's all the new cards.
50:14SPEAKERThe old ones are going to sell through and then they're gone.
50:17SPEAKERBut they were, they very much seem to be emphasizing,
50:20SPEAKERyou know, the low end 4080 is 900.
50:22SPEAKER- Somebody asked in the Q and A,
50:27SPEAKERwhether they were still producing 3080s
50:29SPEAKERand they did not answer that.
50:30SPEAKER- So that, I think that's the real key is not,
50:33SPEAKERdo they have a glut of cards they need to sell through?
50:36SPEAKERIt's are they going to keep making them
50:37SPEAKERand keep that in the product line?
50:38SPEAKERBecause again, so the low end 4080 is 900
50:42SPEAKERand I want to say the 3080 is what, still seven?
50:47SPEAKER- The 3080, they didn't give 3080 pricing,
50:50SPEAKERupdated 3080 pricing in the new regime.
50:52SPEAKER- It's hard to find, it's hard to find MSRP
50:55SPEAKERbecause prices are so all over the place
50:56SPEAKERfor cards right now.
50:57SPEAKERI think most of the 3080s I'm seeing are around 700,
51:00SPEAKER750, 760.
51:02SPEAKER- I would guess that when the 4080s come out,
51:04SPEAKERthe 3080s will drop to five to $600.
51:06SPEAKER- Right, but again, they're going to continue
51:08SPEAKERto be available for a while and those are not,
51:10SPEAKERthose are still perfectly fine cards for 1080 or 1440.
51:14SPEAKER- Well, so here's the thing though.
51:15SPEAKERI don't know, I want to see what the DLSS3 performance
51:18SPEAKERlooks like on these other things.
51:20SPEAKERI want to see how many games support it
51:21SPEAKERin the next three to six months.
51:24SPEAKER- For sure, for sure.
51:25SPEAKERI just mean a 900 dollar card is not exactly entry level
51:28SPEAKERfor a lot of people.
51:29SPEAKER- No, and I think if you can get a 30,
51:32SPEAKERI think if you can get a 4070 with DLSS3
51:35SPEAKERand it performs the same way that it does on the 4090,
51:40SPEAKERit's much better to spend 600 bucks on that
51:41SPEAKERthan 600 bucks on a last gen card.
51:43SPEAKER- Oh, for sure, for sure.
51:44SPEAKERI mean, that relies on the existence of a 4070 though.
51:48SPEAKERThat's what I was getting at is there's just not
51:49SPEAKERthat much room in the pricing strata right now.
51:53SPEAKERThere's nowhere for a 4070 to slot pricing wise
51:56SPEAKERbetween below in 4080 and a 3080.
51:59SPEAKER- Well, I think at some point, I think my assumption
52:01SPEAKERis that when the 3080s run out,
52:03SPEAKERthat's when the 4070 will roll out.
52:05SPEAKER- Yeah, you're probably right.
52:06SPEAKER- Typically the 3070, the 70 series cards
52:09SPEAKERare a little bit later, so.
52:11SPEAKER- Yeah.
52:12SPEAKER- So that's the kind of the TLDR.
52:13SPEAKERDo you want to dig into some of the hardcore,
52:15SPEAKERlike the more detailed stuff
52:17SPEAKERon how these cards actually work?
52:18SPEAKER'Cause it did some neat stuff.
52:20SPEAKER- Yeah.
52:21SPEAKER- It's on the four nanometer TSMC process,
52:24SPEAKERwhich I thought was interesting,
52:25SPEAKERwhich is why the cards are faster
52:27SPEAKERand run the same temperature as the old cards.
52:34SPEAKERThey updated the RT cores, the ray tracing cores.
52:38SPEAKERSo they have roughly 2X the triangle of Val power,
52:42SPEAKERwhich means they can, like that's the number of bounces
52:44SPEAKERthey can evaluate on triangles per cycle.
52:47SPEAKER- Yeah, I'm sitting here trying to read through these notes
52:49SPEAKERI wrote down at six in the morning about these GPU's
52:53SPEAKERand trying to make sense of them.
52:54SPEAKER- Well, so the upshot of the triangle Val power
52:59SPEAKERis that they're 2X faster at the raw ray tracing math.
53:02SPEAKERAnd then they did some other stuff,
53:04SPEAKERlike there's dedicated hardware on the board
53:06SPEAKERthat kind of tessellates high complexity models
53:08SPEAKERdown to lower complexity models
53:10SPEAKERfor the lighting calculations,
53:12SPEAKERthe ray tracing lighting calculations.
53:13SPEAKER- Yeah, it's basically the same stuff
53:14SPEAKERthey've been doing with raster rendering
53:17SPEAKERfor 20 years or whatever.
53:19SPEAKER- It's like they rez it down
53:21SPEAKERand then they do the math on the lighting
53:22SPEAKERand then they rez it back up
53:23SPEAKERand they apply the lighting math
53:25SPEAKERto the up rezzed triangles.
53:29SPEAKERAnd the results are indistinguishable
53:31SPEAKERfrom running the math on the full ones
53:35SPEAKERat a fraction of the frame rate.
53:37SPEAKER- Yeah, they've also done some reordering
53:38SPEAKERwith the way that the ray bounces are processed
53:41SPEAKERbecause they basically said in the old model,
53:44SPEAKERcertain units were kind of sitting semi-idle.
53:47SPEAKER- Yeah.
53:48SPEAKER- Because of the unpredictable and variable way
53:50SPEAKERthat scenes are kind of laid out.
53:52SPEAKER- Yeah, so I think they specifically mentioned
53:54SPEAKERbrushed metal in one of the white papers,
53:55SPEAKERlike brushed metal surfaces are computationally expensive
53:59SPEAKERfor ray tracing to figure out.
54:00SPEAKER- Right.
54:01SPEAKER- 'Cause there's a diffusion and reflection there.
54:03SPEAKERAnd so when you hit a brushed metal surface,
54:06SPEAKERit'll shift load from one core to the other
54:10SPEAKERto kind of like balance that out.
54:11SPEAKER- Basically just trying to make sure
54:12SPEAKERall the units are working efficiently at all times.
54:15SPEAKER- Yeah, and then they also added dedicated hardware
54:18SPEAKERto handle opacity better in ray tracing workloads.
54:21SPEAKERAnd the specific examples here were like leaves.
54:24SPEAKERSo previously, foliage was computationally really expensive
54:29SPEAKERon two series cards and they'll say,
54:31SPEAKERthey said you'll see some substantial performance increase
54:34SPEAKERin areas where that comes into play.
54:37SPEAKER- Yeah, those sessions are fun to sit in on
54:39SPEAKERjust if you wanna get a sense of what's going on internally
54:41SPEAKERand how they're able to claim things
54:43SPEAKERlike 2X the ray tracing performance.
54:44SPEAKERI mean, they don't matter much to people
54:46SPEAKERwho just want to buy a card and run games faster,
54:48SPEAKERbut it's really interesting from an academic standpoint.
54:50SPEAKER- It's absolutely nuts.
54:52SPEAKERYeah, like the level of detail,
54:55SPEAKERlike when you look at a 30 page digital foundry article
54:59SPEAKERor a 30 minute digital foundry video
55:01SPEAKERthat goes into the detail
55:02SPEAKERabout how every single aspect of this work works.
55:04SPEAKERThis is where they share that information.
55:06SPEAKERYou get to talk to a lot of engineers
55:08SPEAKERand ask questions and stuff, which is fun.
55:09SPEAKER- Yes, we didn't mention, you mentioned earlier
55:12SPEAKERthat the cards are limited to display port 1.4.
55:16SPEAKER- Yes.
55:17SPEAKER- They're also PCI express four, not five,
55:19SPEAKERwhich somebody raised as a question
55:23SPEAKERat the end of the session and their response was,
55:25SPEAKERwe're not even close to saturating PCI E4.
55:28SPEAKER- Yeah, for what it's worth.
55:30SPEAKER- Which is interesting 'cause like,
55:32SPEAKERit seems like the CPU feeding the GPU is a big,
55:36SPEAKERlike it seems like the CPU perf is the gate,
55:38SPEAKERnot the GPU perf for once.
55:41SPEAKERLet's see, they did a deep dive into how DLSS3 works.
55:44SPEAKERI spent a lot of time testing DLSS3 stuff
55:46SPEAKER'cause I think it's the most interesting part
55:48SPEAKERof this card, oddly.
55:51SPEAKER'Cause like, hey, the ray tracing stuff is fast enough
55:53SPEAKERto really crank up the ray tracing settings
55:55SPEAKERon existing games is a great place to start.
55:58SPEAKERBut hey, all of your games will run two times faster
56:02SPEAKERif the developers implement this new library is pretty cool.
56:06SPEAKERThe way it works is by taking the lower res images
56:09SPEAKERand using motion vectors and depth maps
56:12SPEAKERto look for motion.
56:15SPEAKERLike now there's a different part of this whole thing
56:18SPEAKERthat uses, it's called an optical flow accelerator
56:20SPEAKERand it uses, it looks at motion differences between frames
56:23SPEAKERto generate vectors, to do in between frames,
56:28SPEAKERit's like part of this works because the cards
56:31SPEAKERare really fast and they know what's coming
56:33SPEAKERbefore you can see what's coming, right?
56:35SPEAKERSo it'll put a frame in between two frames that are slower.
56:38SPEAKEROne is running slower than it needs to
56:40SPEAKERand you get just a smoothed out frame rate.
56:45SPEAKERThe DLSS3 stuff only works on the 4,000 series cards,
56:48SPEAKERbut like I said, it does fall back gracefully
56:50SPEAKERon the older cards.
56:52SPEAKERSo 2,000 and 3,000 series cards will get DLSS2 perf on it.
56:57SPEAKERAnd they're trying to, the whole point of this,
57:01SPEAKERlike NVIDIA's quest to destroy microstutter
57:04SPEAKERcontinues with this.
57:05SPEAKER- I salute them, it is a bane.
57:09SPEAKER- I will say from working on the Anacrusis,
57:12SPEAKERwhen we found stuff that was doing like lots of draw calls,
57:16SPEAKERwhen we changed the materials on the horde enemies
57:18SPEAKERand stuff like that, the actual average frame rate increase
57:22SPEAKERwas basically nothing, but the perceived frame rate increase
57:26SPEAKERwas enormous even on like Xbox Series S and Xbox One
57:31SPEAKERand stuff like that.
57:32SPEAKER- It's like you were saying earlier,
57:33SPEAKERit's all about bringing up that 1% low.
57:35SPEAKER- Yeah, yeah, 'cause even if you don't,
57:38SPEAKEReven if you as a player don't notice
57:40SPEAKERthat you're getting the stutters,
57:42SPEAKERyou notice it in the smoothness of the motion
57:44SPEAKERwhen you're like moving your mouse across the scene
57:45SPEAKERand stuff like that.
57:46SPEAKER- It drives me nuts.
57:47SPEAKERLike it is, it's way worse than just a, well, whatever.
57:51SPEAKERI mean, this is basically what we've already said,
57:52SPEAKERbut like it's worse
57:54SPEAKERthan just a consistently low frame rate to me.
57:56SPEAKER- Yeah.
57:57SPEAKERNow this requires reflex.
58:00SPEAKERSo you have to implement reflex for this to work
58:02SPEAKERbecause those frames skip the render queue.
58:06SPEAKERBut when reflex lets everything skip the render,
58:08SPEAKERreflex basically works by working very closely with the CPU
58:12SPEAKERto make sure that the CPU is delivering new information
58:15SPEAKERto the GPU just in time for the frame to be rendered.
58:17SPEAKERSo there's no buffer which adds latency
58:20SPEAKERto the whole pipeline.
58:21SPEAKERAnd it was one of the places that micro stutter happens.
58:26SPEAKERSo that's that.
58:26SPEAKERThe other big thing we touched on this a little bit
58:28SPEAKERis the eighth gen NVENC encoder,
58:33SPEAKERwhich adds AV1 support.
58:34SPEAKERAnd the 40, let's see,
58:36SPEAKERI think it's cards with greater than 12 gigs have two,
58:41SPEAKER12 gigs or greater can do two AV1 streams at once.
58:45SPEAKER- Yeah, they've got an NVENC matrix on their site
58:50SPEAKERthat has been around since the Turing cards
58:52SPEAKERthat you can go look at
58:52SPEAKERthat tells you literally every card they have
58:54SPEAKERand like how many simultaneous encodes
58:57SPEAKERcan this one do for a given codec?
59:00SPEAKER- It's pretty nifty.
59:02SPEAKER- Yeah.
59:03SPEAKER- I guess nobody implemented H.265
59:05SPEAKERbecause it's encumbered by patents, right?
59:07SPEAKER- Yeah, if you listen to the VLC episode of the FOSPOD.
59:11SPEAKER- Oh, interesting.
59:12SPEAKER- You can hear Jean-Baptiste.
59:14SPEAKER- Is that at FOSPOD.content.town?
59:16SPEAKER- It is.
59:16SPEAKER- Wow.
59:17SPEAKER- You can hear him talk fairly bluntly
59:19SPEAKERabout why nobody adopted H.265.
59:21SPEAKERI mean, so AV1 is more efficient
59:23SPEAKERthan that codec also on a technical level.
59:28SPEAKERIt's not just a royalties situation,
59:32SPEAKERbut it's largely a royalty situation.
59:36SPEAKER- So the AV1-
59:37SPEAKER- Dude, I'm so excited for fucking the world to move on.
59:41SPEAKER- I mean, look, my brain has filtered
59:43SPEAKERout H.264 artifacts for long enough.
59:45SPEAKERI'm ready to teach it some new artifacting.
59:47SPEAKER- Yeah.
59:49SPEAKER- The AV1 stuff is like the other,
59:54SPEAKERlike the other bits and pieces we've talked about are,
59:59SPEAKERit's not widely supported yet.
1:00:01SPEAKER- Definitely not, but if you care at all about streaming,
1:00:05SPEAKERyou want to buy hardware that can encode AV1 at this point.
1:00:08SPEAKER- Yeah, it's weird 'cause it's like,
1:00:09SPEAKERit means in a two-piece stream setup,
1:00:11SPEAKERI'm gonna be looking at, you know, a second,
1:00:16SPEAKERa second, another AV1 capable GPU, I guess.
1:00:20SPEAKER- Well, that's why it's so great
1:00:21SPEAKERthat every card launching this year has it.
1:00:22SPEAKERLike you can go pick up a mid-range Intel Arc
1:00:25SPEAKERand get AV1 for a streaming machine.
1:00:28SPEAKER- That is true.
1:00:29SPEAKERI'm trying to see, I think that the Premiere update
1:00:32SPEAKERmight've landed the other night.
1:00:36SPEAKERThe stuff that I was using for AV1
1:00:40SPEAKERwas early access, I believe, for DaVinci.
1:00:45SPEAKERAlso, to be clear, as far as I know,
1:00:47SPEAKERthe Arc cards use the same encoding engine
1:00:49SPEAKERas Quick Sync on the Intel CPUs.
1:00:51SPEAKERAnd I believe that is also coming.
1:00:54SPEAKERIf Wikipedia is to be believed,
1:00:56SPEAKERthe next generation of Intel CPU is not Raptor
1:00:58SPEAKERthat's coming out right now.
1:00:59SPEAKER- Okay. - But next year,
1:01:00SPEAKERMeteor Lake will have AV1 hardware encode in Quick Sync.
1:01:05SPEAKER- Oh, that's interesting.
1:01:06SPEAKER- So I don't know where AMD is at with that stuff.
1:01:08SPEAKERI think AMD's hardware encoders
1:01:10SPEAKERhave not been super well thought of.
1:01:12SPEAKER- The Quick Sync quality is often not production level,
1:01:16SPEAKERis my understanding, at least in the old days.
1:01:18SPEAKERI don't know where it is now.
1:01:19SPEAKER- I think it's better.
1:01:21SPEAKER- Okay. - My impression is it's decent.
1:01:24SPEAKER- Well then, okay, so then there's also,
1:01:26SPEAKERthat's kind of the hardware.
1:01:27SPEAKERThose are the big hardware beats.
1:01:28SPEAKERI mean, we can talk about number of cores.
1:01:30SPEAKERLike the 4090, when you're looking at shader cores,
1:01:36SPEAKERit's almost double what the 4080 cards are gonna have
1:01:39SPEAKERwhen they're available.
1:01:40SPEAKERThe ray tracing cores are almost double again.
1:01:43SPEAKERIt's like 113 for the big 4080 and 191 for the 4090
1:01:48SPEAKERand 320 tensor teraflops,
1:01:54SPEAKERwhich is, this is maybe, sorry, this is,
1:01:58SPEAKERit's like, it's bonkers.
1:01:59SPEAKERIt's a big chunky card.
1:02:02SPEAKER- Yeah.
1:02:03SPEAKER- The other stuff that they showed,
1:02:05SPEAKERa bunch of other tech demos that I thought were kind of cool
1:02:07SPEAKERlike the modding tool.
1:02:09SPEAKERIf we want to talk about that a little bit.
1:02:12SPEAKERI think you talked about on Nextlander.
1:02:13SPEAKER- Oh, I was gonna say,
1:02:14SPEAKERI feel like I've gushed about that a lot already.
1:02:16SPEAKERYeah, the RTX remix looks amazing
1:02:18SPEAKERand I am very excited to see what people do with that.
1:02:21SPEAKER- It's not available yet, but it will work with every,
1:02:23SPEAKERwith all RTX cards, I believe.
1:02:26SPEAKER- I believe it goes into open beta next month.
1:02:28SPEAKER- Yeah.
1:02:29SPEAKERIt has to be games with DX9 or DX8 fixed function pipelines.
1:02:33SPEAKER- Hang on, does the editor work on all RTX cards?
1:02:35SPEAKERI know the output, like a modded game,
1:02:38SPEAKERan RTX modded, sorry,
1:02:40SPEAKERan RTX remix modded game will run
1:02:42SPEAKERon any ray tracing capable cards.
1:02:44SPEAKERI don't know if the editor does.
1:02:46SPEAKER- I think I'm gonna look now,
1:02:47SPEAKERbut I, so you should explain what it is though, first.
1:02:51SPEAKER- It is a compatibility layer for getting DirectX 8
1:02:54SPEAKERand 9 games running with a modern renderer
1:02:56SPEAKERthat supports ray tracing and a bunch of other stuff.
1:03:00SPEAKERI really, there's a video you can go watch of this process,
1:03:04SPEAKERbut it's really, it's the way that you hook into the game
1:03:06SPEAKERthat it seems so magical to me.
1:03:08SPEAKERBecause you literally run one of these old games
1:03:11SPEAKERand then run RTX remix next to it,
1:03:13SPEAKERand essentially kind of hit a capture button.
1:03:15SPEAKER- It grabs stuff from the pipeline, the graphics pipeline.
1:03:18SPEAKER- Right, it basically peers into this rickety old
1:03:20SPEAKERrendering pipeline and sees what's going on
1:03:22SPEAKERand then inserts itself.
1:03:25SPEAKERThere's a slide, I think, that's in that trailer
1:03:28SPEAKERthat I'm talking about that basically shows
1:03:30SPEAKERlike they're doing, I believe, kind of dynamic,
1:03:34SPEAKERI don't know if it's like code recompilation or whatever,
1:03:36SPEAKERbut you know what I mean?
1:03:37SPEAKERThese are games that were written for hardware
1:03:40SPEAKERthat is 20 years old at this point.
1:03:42SPEAKER- It requires a GeForce RTX GPU to create RTX mods.
1:03:46SPEAKER- Oh, okay, so any RTX.
1:03:47SPEAKER- So any 2000 series card or newer will work.
1:03:50SPEAKER- Okay, that's great.
1:03:53SPEAKERThey're using, and it's not that it's super relevant,
1:03:55SPEAKERthey're using DXVK for this,
1:03:57SPEAKERwhich you might know from Proton and some other things.
1:04:00SPEAKER- That's interesting.
1:04:01SPEAKER- They're basically, that's basically a DirectX 2
1:04:03SPEAKERVulkan translation layer.
1:04:05SPEAKERIs part of how they're achieving that.
1:04:07SPEAKER- And the mods are playable on any hardware
1:04:10SPEAKERthat can run Vulkan ray traced games.
1:04:12SPEAKERSo that means it's not limited to Nvidia stuff.
1:04:14SPEAKERIt'll also work on AMD and presumably Intel hardware as well.
1:04:17SPEAKER- And they basically made it sound like these mods
1:04:20SPEAKERwill plug in to an old game just like any existing mod does.
1:04:23SPEAKERLike they essentially said like,
1:04:25SPEAKERhey, these will be hosted on Nexus mods,
1:04:26SPEAKERjust like the mods you're used to.
1:04:28SPEAKER- Yeah, and ModDB as well.
1:04:31SPEAKERThe demos were really cool 'cause like the,
1:04:33SPEAKERoh, and the assets pipe out in a USD format.
1:04:37SPEAKERSo you can edit them in pretty much any,
1:04:39SPEAKERlike you don't need the normal pipeline of 30 tools
1:04:42SPEAKERto do this, to do the uprising.
1:04:46SPEAKERSo I assume that means we'll see the Half-Life 2
1:04:50SPEAKER- Yes, for sure.
1:04:51SPEAKER- remaster sooner than it took.
1:04:53SPEAKERI mean, I guess the Black Mesa people
1:04:55SPEAKERjust rolled straight over to that, right?
1:04:56SPEAKER- Well, so watching them do a,
1:04:59SPEAKERwell, it's in the trailer, you can go watch it.
1:05:00SPEAKERLike watching them do the modifications themselves,
1:05:02SPEAKERit seems like it's going to take a lot of work per scene.
1:05:05SPEAKER- Yeah, there were some questions.
1:05:07SPEAKERWe didn't get to ask questions at the end of this,
1:05:09SPEAKERI don't think, 'cause we had to roll to the next session.
1:05:11SPEAKERBut I was curious if, you know,
1:05:13SPEAKERif they demoed this with Morrowind,
1:05:15SPEAKERand I was curious if, you know,
1:05:17SPEAKERsay you replaced the chair,
1:05:18SPEAKERthat's the chair in every fancy person's house in Morrowind.
1:05:22SPEAKER- That's in that demo.
1:05:23SPEAKERThey show you can replace the assets globally.
1:05:25SPEAKER- Oh, and it replaces globally?
1:05:26SPEAKEROkay, cool.
1:05:27SPEAKER- If you do the work on like a chair or a bowl or whatever,
1:05:29SPEAKERyou can say, hey, replace this model across the entire game.
1:05:32SPEAKERBut I'm talking more like bespoke light sources
1:05:35SPEAKERin a given scene have to be massaged quite a bit.
1:05:39SPEAKER- But I wonder if when you replace one,
1:05:41SPEAKERif you can replace like light sources globally,
1:05:43SPEAKERso if it's the same fixture in every house,
1:05:45SPEAKERif you can do that,
1:05:46SPEAKERI mean, there'll be performance issues
1:05:47SPEAKERthat come with that, I bet.
1:05:48SPEAKER- There's like that paper lamp they showed in there
1:05:50SPEAKERwhere it was like, hey,
1:05:51SPEAKERwe had to define a translucent paper material on this globe
1:05:54SPEAKERto make the light pass through.
1:05:55SPEAKERLike you could probably do that globally, I would assume.
1:05:57SPEAKER- Yeah.
1:05:58SPEAKER- I want to see Mass Effect.
1:06:01SPEAKER- Are there, is Mass Effect a fixed function, I guess?
1:06:04SPEAKER- All three, that I don't know.
1:06:07SPEAKERAll three Mass Effect games are DX9 on PC.
1:06:10SPEAKERI went and checked, but actually I don't know about,
1:06:13SPEAKERI don't know what their use of shaders is.
1:06:16SPEAKER- Yeah, I'm trying to think.
1:06:17SPEAKERI feel like they use shaders for,
1:06:19SPEAKERthey used vertex shaders for a lot of stuff in that game,
1:06:22SPEAKERI think.
1:06:23SPEAKER- I forgot that they'd also had to be a fixed function game.
1:06:25SPEAKERThat may be more limiting.
1:06:27SPEAKER- The thing is there's a lot of,
1:06:30SPEAKERthere's a lot of vertex shaders in Xbox 360 era hardware,
1:06:34SPEAKERbut not a lot of pixel shaders.
1:06:35SPEAKERSo I don't know.
1:06:37SPEAKERLike we might just not see grass.
1:06:40SPEAKERBut anyway, I think it's the most interesting,
1:06:43SPEAKERthis is one of the more interesting demos that they've seen.
1:06:45SPEAKERThey've apparently done a whole version of Portal 1
1:06:49SPEAKERthat'll be coming out soon.
1:06:51SPEAKERThat is Ray Traced Portal 1 next month.
1:06:54SPEAKER- Yes.
1:06:55SPEAKER- I'm very excited to see what this looks like.
1:06:58SPEAKERI haven't played Portal a very long time.
1:07:00SPEAKERThis seems like a very good reason.
1:07:01SPEAKERIt's short, you know.
1:07:02SPEAKER- Yeah, it's like one good session.
1:07:04SPEAKER- Three hours or something, you could,
1:07:06SPEAKERthis would be a great excuse to play Portal again.
1:07:08SPEAKER- I mean, look,
1:07:09SPEAKERdo you ever need an excuse to play Portal again?
1:07:11SPEAKER- Ah, you know, got a lot of games to play, all right.
1:07:14SPEAKER- Fair, there's never more.
1:07:16SPEAKERI will say going back into Cyberpunk made me wanna go back
1:07:20SPEAKERand play Cyberpunk for real this time.
1:07:21SPEAKER- Yeah, I never played it
1:07:22SPEAKER'cause it was such a mess at launch.
1:07:23SPEAKERAnd now I think I wanna try to get one of these new cards
1:07:26SPEAKERbefore I play it.
1:07:27SPEAKER- It's pretty good, but anyway.
1:07:31SPEAKER- Yeah, the only other thing I'll mention here
1:07:32SPEAKERbefore we wrap up is in the slide in our notes here.
1:07:37SPEAKERApparently no founder's edition of the 12 gigabyte 4080.
1:07:41SPEAKER- That's interesting, I missed that.
1:07:43SPEAKER- It looks like they're only doing founder's editions
1:07:45SPEAKERof the 16 gig, the 12 gig will only be AIB.
1:07:49SPEAKER- Interesting, yeah, that's a weird choice.
1:07:56SPEAKER- Not a lot of weird choices.
1:07:57SPEAKER- I mean, like I said, I would hesitate mightily
1:08:01SPEAKERbefore buying one of these cards if--
1:08:04SPEAKER- The 90 in particular, you mean?
1:08:06SPEAKER- I would wanna know what's going on with the,
1:08:07SPEAKERlike why did they launch 90s first and not launch 80s?
1:08:11SPEAKER- Well, especially 'cause it's a month, you know,
1:08:12SPEAKERlike we can wait a month and see what happens next month
1:08:14SPEAKERwith the 4080s.
1:08:15SPEAKER- Yeah, and it may be that like the 4080
1:08:19SPEAKERis demonstrably slower even with DLSS and all that,
1:08:22SPEAKERbut I like, I would not be surprised if,
1:08:24SPEAKERespecially if you're playing Raster games, you know,
1:08:26SPEAKERif you're playing your Fortnite's and your PUBG's
1:08:27SPEAKERand your Apex Legends and all that,
1:08:29SPEAKERif you see roughly equivalent performance with DLSS on,
1:08:33SPEAKERespecially once people start supporting DLSS3.
1:08:35SPEAKER- Sure, are you gonna undervolt?
1:08:39SPEAKERYou should undervolt.
1:08:42SPEAKER- The CPU or the GPU?
1:08:43SPEAKER- Well, all of the above, man.
1:08:45SPEAKERSomebody literally said it on our Discord a day or two ago,
1:08:48SPEAKERundervolting is the new overclocking.
1:08:50SPEAKER- I do like quiet.
1:08:52SPEAKERI don't know, I'm about ready to build a new computer.
1:08:53SPEAKERI was trying to get the new computer together
1:08:57SPEAKERin order to benchmark this.
1:08:58SPEAKERAnd it just, for a multitude of reasons, didn't work out,
1:09:01SPEAKERbut we'll have something cool to talk about
1:09:02SPEAKERon that front soon.
1:09:04SPEAKER- AMD or Intel?
1:09:05SPEAKER- I haven't decided yet.
1:09:06SPEAKERI want to see what the, I want to see what the 13th,
1:09:07SPEAKERwhat the Intel numbers look like next on the embargo list.
1:09:10SPEAKER- That's sensible.
1:09:11SPEAKERHey, with all of these things,
1:09:13SPEAKERif you're not absolutely dying to upgrade
1:09:14SPEAKERor to own the fastest thing in the market,
1:09:16SPEAKERmaybe a lot of hardware coming out
1:09:20SPEAKERin the next three, four months.
1:09:21SPEAKER- Yep, yep.
1:09:23SPEAKER- Before you commit to anything.
1:09:25SPEAKER- And also it seems like the supply chain stuff
1:09:28SPEAKERis a little bit easier, at least right now,
1:09:29SPEAKER'cause without the crypto pressure,
1:09:32SPEAKERit's getting a lot easier.
1:09:34SPEAKER- Yes, thank God.
1:09:36SPEAKERI'm glad that's over.
1:09:37SPEAKER- Yeah, the other thing is I want to see
1:09:39SPEAKERwhen the DLSS3 enabled stuff lands.
1:09:42SPEAKER'Cause like none of the games that I,
1:09:44SPEAKERI went and checked before we recorded this morning,
1:09:46SPEAKERI haven't seen patches that flip on DLSS3
1:09:48SPEAKERfor any of the games that were listed.
1:09:50SPEAKERI don't think Plague Tale's out yet.
1:09:52SPEAKER- No, it's out in the next week, I think.
1:09:55SPEAKERI'm surprised nothing was ready in time for launch.
1:09:58SPEAKER- But like Flight Sim I don't think is updated yet.
1:10:01SPEAKERThat was the one that I couldn't find.
1:10:02SPEAKERCyberpunk doesn't have any patch notes.
1:10:06SPEAKERIt is non-optimal for these things to not,
1:10:10SPEAKERlike for this much of a marquee feature
1:10:12SPEAKERto not be supported on day one in any games.
1:10:14SPEAKER- Yeah.
1:10:15SPEAKER- So, but with that, it's time, Brad.
1:10:19SPEAKERAnd if people have questions about this,
1:10:20SPEAKERI have had some personal stuff this week.
1:10:23SPEAKERSo my intent was to stream a lot this week with 40 90s.
1:10:26SPEAKERI am going to pick that up.
1:10:27SPEAKERI think things are improving this week.
1:10:30SPEAKERSo maybe this weekend and into next week,
1:10:32SPEAKERwe'll do some 40 90 streams.
1:10:33SPEAKERSo if people want to see specific games, we can run them.
1:10:35SPEAKERI don't think I'm allowed to stream Plague Tale,
1:10:37SPEAKERbut Cyberpunk is fair game.
1:10:39SPEAKERAnd I will stream that anytime people want.
1:10:41SPEAKERSo if people want to see games,
1:10:43SPEAKERor if you have questions send in the normal,
1:10:46SPEAKERyeah, either post in the queues seeking A's channel
1:10:48SPEAKERor email us at tech bod@content.town.
1:10:51SPEAKERAnd we will answer them in the,
1:10:54SPEAKERin the upcoming questions and answers episode.
1:10:56SPEAKER(imitating air horn)
1:11:00SPEAKER- Brad here from the future.
1:11:02SPEAKERThat's my little breaking news sound for you there.
1:11:05SPEAKERDropping a little fill in here that's recorded later
1:11:08SPEAKERbecause literal moments
1:11:10SPEAKERafter we finished recording this episode,
1:11:11SPEAKERNvidia put up a very, very short blog post entitled
1:11:16SPEAKERUnlaunching the 12 gigabyte 4080.
1:11:20SPEAKERI'm just going to read the entire thing here
1:11:21SPEAKERbecause it's extremely short.
1:11:23SPEAKERThe RTX 4080 12 gigabyte is a fantastic graphics card,
1:11:27SPEAKERbut it's not named right.
1:11:29SPEAKERHaving two GPU's with the 4080 designation is confusing.
1:11:33SPEAKERSo we're pressing the unlaunch button
1:11:35SPEAKERon the 4080 12 gigabyte.
1:11:37SPEAKERThe RTX 4080 16 gigabyte is amazing and on track
1:11:40SPEAKERto delight gamers everywhere on November 16th.
1:11:43SPEAKERIf the lines around the block and enthusiasm for the 4090
1:11:46SPEAKERis in the indication,
1:11:47SPEAKERthe reception for the 4080 will be awesome.
1:11:51SPEAKERSo there you have it.
1:11:53SPEAKERDespite all the stuff you just heard in this episode,
1:11:55SPEAKERthere will actually only be one 4080 coming out in November
1:11:58SPEAKERon the 16th, which might be the first time they've mentioned
1:12:02SPEAKERthat specific date actually.
1:12:04SPEAKERIt is the 16 gigabyte model, which is $1,200.
1:12:07SPEAKERThe $900 4080 12 gigabyte, which is, as we said,
1:12:10SPEAKERa reduced GPU in terms of capabilities
1:12:14SPEAKERwill not be coming out in November.
1:12:16SPEAKERNo other information about when it will be coming out
1:12:19SPEAKERor what it will be named or what it will be priced at.
1:12:23SPEAKERBut wanted to make sure
1:12:24SPEAKERthat that late breaking information got into this episode.
1:12:27SPEAKERSo there you go.
1:12:28SPEAKER- Brad, I think it's time that we thank our patrons.
1:12:32SPEAKER- Thank you, patrons.
1:12:33SPEAKER- Thank you, patrons.
1:12:34SPEAKERAs always, Brad and Will Made at TechPod
1:12:36SPEAKERis a 100% listener supported show.
1:12:38SPEAKERIf you would like to support the show
1:12:39SPEAKERwith as little as $2 a month,
1:12:41SPEAKERyou can join the fabulous Discord
1:12:43SPEAKERfilled with beautiful nerds
1:12:44SPEAKERwho talk about things like undervolting
1:12:47SPEAKERby going to patreon.com/techpod.
1:12:49SPEAKERAgain, that's patreon.com/techpod and signing up there.
1:12:54SPEAKERAs always, we would like to thank our executive producer
1:12:57SPEAKERtier patrons, including Paddle Creek Games,
1:12:59SPEAKERmakers of "Fractured Veil,"
1:13:00SPEAKERAndrew Slawski, #bunnyslimes,
1:13:04SPEAKERReluctant Meme Participant Wedge,
1:13:08SPEAKERJoel Krauska, Twinkle Twinkie, and James Kamik.
1:13:12SPEAKERAnd I think the gift swap is starting up soon.
1:13:14SPEAKERI got a message last night about,
1:13:16SPEAKER"Hey, can you make a channel for the gift swap?"
1:13:18SPEAKERSo we have a well-trafficked gift swap
1:13:22SPEAKERfor the TechPod last year.
1:13:23SPEAKERSo if you wanna get some nerdy stuff
1:13:25SPEAKERfrom other delightful nerds,
1:13:27SPEAKERthat's a good thing to get involved with.
1:13:28SPEAKERAnd I like that we started early enough
1:13:30SPEAKERthat it's not like into the holiday crush.
1:13:34SPEAKERSo anyway, I guess that'll do it for us this week.
1:13:37SPEAKERWe'll be back next week with another episode of the TechPod.
1:13:41SPEAKER- True, a true statement.
1:13:43SPEAKERAlways there.
1:13:44SPEAKERSee you all next week.
1:13:47SPEAKER(upbeat music)