Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
Dr. Video Games 0031
Jul 17, 2004

Yeah, I'm thinking gamecube/ps2-era emulation and older, old PC games, and indie games. It'll have 4 Zen 2 cores and 2 RDNA2 CUs, while the Steam Deck has 4 Zen 2 cores and 8 RDNA2 CUs. Though the power envelope of this chip is also unconfirmed currently, so those Zen 2 cores could be held back by a lower power limit. Aya Neo has also announced a thin and light handheld using that chip, and I bet they're going to enforce a pretty low TDP (15W maybe?) in order to achieve that form factor.

Adbot
ADBOT LOVES YOU

Rinkles
Oct 24, 2010

What I'm getting at is...
Do you feel the same way?
Isn’t PS2 emulation fairly demanding?

Dr. Video Games 0031
Jul 17, 2004

I mean, i used to run PCSX2 fairly okay on my Phenom II X4, so I dunno, can't be that hard.

Zero VGS
Aug 16, 2002
ASK ME ABOUT HOW HUMAN LIVES THAT MADE VIDEO GAME CONTROLLERS ARE WORTH MORE
Lipstick Apathy

Dr. Video Games 0031 posted:

And no price can sell me on the 4090 because I'm not buying a 450W GPU lmao

Yeah, do those charts have any veracity to them or just from another rumor mill site? I thought the point of these die shrinks was to get some more efficiency?

My 3080 can pull 320w but I got a very lucky chip that let me undervolt to 800mv and keep the clocks stock. Now it tops out at 250w which is good because my PSU is a 480w fanless (it's a 12v Meanwell bench power supply, so I got a 12v-only motherboard and I'm sending a super-efficient 12vdc straight to the CPU/GPU.

Those numbers don't seem to be indicating any substantial efficiency gains, but maybe it's a matter of undervolting them again.

Rinkles posted:

Isn’t PS2 emulation fairly demanding?

Not anymore it isn't, even PS3 is doable. I beat all of Demon's Souls on PS3 emulator with it upscaled to 4K and it held 60fps most of the time.

Rinkles
Oct 24, 2010

What I'm getting at is...
Do you feel the same way?

Zero VGS posted:

Not anymore it isn't, even PS3 is doable. I beat all of Demon's Souls on PS3 emulator with it upscaled to 4K and it held 60fps most of the time.

We were talking about an integrated graphics laptop chip

Bloodplay it again
Aug 25, 2003

Oh, Dee, you card. :-*

Zero VGS posted:

My 3080 can pull 320w but I got a very lucky chip that let me undervolt to 800mv and keep the clocks stock.

:eyepop: Mine starts to give me issues at lower than .9 V. Baby, tell me more about those energy savings mmmmm

hobbesmaster
Jan 28, 2008

Zero VGS posted:

Yeah, do those charts have any veracity to them or just from another rumor mill site? I thought the point of these die shrinks was to get some more efficiency?

My 3080 can pull 320w but I got a very lucky chip that let me undervolt to 800mv and keep the clocks stock. Now it tops out at 250w which is good because my PSU is a 480w fanless (it's a 12v Meanwell bench power supply, so I got a 12v-only motherboard and I'm sending a super-efficient 12vdc straight to the CPU/GPU.

Those numbers don't seem to be indicating any substantial efficiency gains, but maybe it's a matter of undervolting them again.

Not anymore it isn't, even PS3 is doable. I beat all of Demon's Souls on PS3 emulator with it upscaled to 4K and it held 60fps most of the time.

Like, “stock” base boost of 1700mhz? or whatever nvidia advertises or what the AIBS advertise up at something like 1900mhz? Either way that’s impressive.

lih
May 15, 2013

Just a friendly reminder of what it looks like.

We'll do punctuation later.

Zero VGS posted:

Yeah, do those charts have any veracity to them or just from another rumor mill site? I thought the point of these die shrinks was to get some more efficiency?
it's mostly based on reports from kopite7kimi who has inside sources but certainly doesn't know everything, with some estimates and guesses in the mix too

it's a solid idea of what's currently expected but not definitive or anything

Dr. Video Games 0031
Jul 17, 2004

Zero VGS posted:

Yeah, do those charts have any veracity to them or just from another rumor mill site? I thought the point of these die shrinks was to get some more efficiency?

My 3080 can pull 320w but I got a very lucky chip that let me undervolt to 800mv and keep the clocks stock. Now it tops out at 250w which is good because my PSU is a 480w fanless (it's a 12v Meanwell bench power supply, so I got a 12v-only motherboard and I'm sending a super-efficient 12vdc straight to the CPU/GPU.

Those numbers don't seem to be indicating any substantial efficiency gains, but maybe it's a matter of undervolting them again.

Those specs come from this tweet:

https://twitter.com/kopite7kimi/status/1526135149246976001

Bear in mind that this person was a very accurate leaker for both the 20-series and 30-series, leaking the exact card configurations for those generations' initial lineups, though he's been a bit more back-and-forth on the 40-series and has bemoaned the inconsistency of the info he's getting from his sources. I think he's a legit leaker, but you still need to take the info with a grain of salt (especially performance estimates)

A 450W 4090 that's even just 50% more powerful than a 3090 would still be getting more performance per watt than a stock 3080 (21% more). A 4090 that's 2x more powerful than the 3090 would be getting 60% more performance per watt than the 3080. The truth may be somewhere in the middle, but going from samsung 8nm to tsmc 4nm would hopefully bring a pretty big perf per watt boost.

edit: forgot which gpu it's meant to be 2x of lol

Dr. Video Games 0031 fucked around with this message at 03:14 on Jun 2, 2022

hobbesmaster
Jan 28, 2008

The problem with the exact power numbers in leaks is that that’s something that can be changed anytime before they’re in stores so 100% accurate leaks can still be “wrong”.

shrike82
Jun 11, 2005

there were similar 2x rumors for Ampere and the differential between the 3090 and 2080 ti turned out to be a lot less

Dr. Video Games 0031
Jul 17, 2004

It was Nvidia themselves claiming a 2x improvement, and it turned out to be BS.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Dr. Video Games 0031 posted:

Those specs come from this tweet:

https://twitter.com/kopite7kimi/status/1526135149246976001

Bear in mind that this person was a very accurate leaker for both the 20-series and 30-series, leaking the exact card configurations for those generations' initial lineups, though he's been a bit more back-and-forth on the 40-series and has bemoaned the inconsistency of the info he's getting from his sources. I think he's a legit leaker, but you still need to take the info with a grain of salt (especially performance estimates)

A 450W 4090 that's even just 50% more powerful than a 3090 would still be getting more performance per watt than a stock 3080 (21% more). A 4090 that's 2x more powerful than the 3090 would be getting 60% more performance per watt than the 3080. The truth may be somewhere in the middle, but going from samsung 8nm to tsmc 4nm would hopefully bring a pretty big perf per watt boost.

edit: forgot which gpu it's meant to be 2x of lol

21% more perf/watt would be extremely disappointing for a ~1.5 node jump. The 4nm rumors put it at basically 2 full nodes again. The 60% more perf/watt number is very plausible/reasonable in that context.

GTX 1080 was a 67% bump in perf/w over a GTX 980 at 1440p, for context. That should be a broadly similar leap - the jump to 16nm was a big one because 20nm flopped and never happened, so you went right from 28nm to 16nm, it was probably a ~1.5x node leap, perhaps a bit short of a full 2x though.

I would not be surprised to see something in the range of 60% bump in perf/w at peak-performance. That's plausible, although of course it will strongly depend on the specifics of the SKUs - as you note, it's trivial for them to bump the TDP up if they want to, which affects perf/w, so even a number that is 100% accurate today could change before launch.

Dr. Video Games 0031 posted:

It was Nvidia themselves claiming a 2x improvement, and it turned out to be BS.

it wasn't BS tho, it was just people not understanding the numbers that were being cited. NVIDIA clearly showed iso-performance perf/watt on their slide, not peak-performance perf/watt. And yeah if you take a 3080 and clock it way down while running a 2080 ti at its max clock then sure, perf/w goes way up, that was even discussed at the time.



If you didn't understand that slide was showing iso-performance perf/w... that's kinda on you, because it's literally the entire focus of the slide. See the arrow? See where the lines are placed? Yeah, you have to actually read the chart, people - you not reading the chart doesn't make the chart BS.

I think people just kinda came out of the ampere announcement malding and then continued to mald about poor supply, and never really critically examined some of the things they were discontented about because they had new reasons to be discontented... and shrike gonna shrike.

Paul MaudDib fucked around with this message at 05:37 on Jun 2, 2022

Dr. Video Games 0031
Jul 17, 2004

I wouldn't be surprised if it's 5nm instead. but 4nm is just an offshoot node with some small optimizations apparently, so maybe it'd be 50% more perf/watt instead, which would still be great.

nvidia has announced that they're using 4nm for Hopper which is why people think lovelace will use that, but i could see them using a less expensive node for the desktop cards

Dr. Video Games 0031 fucked around with this message at 05:28 on Jun 2, 2022

MarcusSA
Sep 23, 2007

Rinkles posted:

We were talking about an integrated graphics laptop chip

I have the Odin pro which used SD845 cell phone chip and it can emulated nearly all of the ps2 library at 1x.

PS2 has gotten a lot better.

Agreed
Dec 30, 2003

The price of meat has just gone up, and your old lady has just gone down

PS3 emulation seems like it barely hits my GPU, big time CPU. But I was able to run some PS3 games at 60FPS / 4K with a 4770K at 4.3GHz and a 2070, whereas Yuzu has a lot of CPU-limited stuttering in more titles for me. I dunno, emulation is weird. But awesome.

If I get a 4080 I'm gonna get one with so much RGB, you have no fuckin clue. I can't believe I care about that but it turns out I've been kind of bummed that my video card does not light up when the 3070 Ti I had did, kinda, enough to see it you know? I don't require much, maybe I ought to just go get a few xmas lights to put on the bastard I dunno

Agreed fucked around with this message at 05:56 on Jun 2, 2022

Dr. Video Games 0031
Jul 17, 2004

Dr. Video Games 0031 posted:

I wouldn't be surprised if it's 5nm instead. but 4nm is just an offshoot node with some small optimizations apparently, so maybe it'd be 50% more perf/watt instead, which would still be great.

nvidia has announced that they're using 4nm for Hopper which is why people think lovelace will use that, but i could see them using a less expensive node for the desktop cards

It's time to theory craft using random leaks and rumors. everyone quote this post in a few months to laugh at me.

Even if the 4090 is 2x the 3090, that won't mean the 4080 will be 2x the 3080. The 4090 is rumored to have 53.7% more SMs than the 3090 while the 4080 is rumored to to use the AD103 GPU which tops out at 84 SMs, so that card will maybe use 80 - 82, or 17.7% - 20.6% more than the 3080. Given that, if the 4090 has 2x perf over the 3090, then my guess is that the 4080 will be 1.5x - 1.6x the 3080. That would be a large but much more ordinary generational improvement (see the 980 -> 1080). So if these rumors are true, then the 4090's gonna be a ridiculous monster card that costs a shitload of money while the rest of the lineup will be fairly sane. All that headroom will also leave the door open for the 50-series to get a good gen-on-gen boost without having any major architectural improvements since they can just do a die shrink and move the SKUs up a bit (e.g. making the 5070 a 4080/4080 Ti that uses 20% less power or w/e). And if that's Nvidia's strategy, then that would be because they feel they're safe from any aggressive moves in that performance tier from AMD and Intel.

Comfy Fleece Sweater
Apr 2, 2013

You see, but you do not observe.

I'm not putting any component in my PC that consumes 450W lol come on

Not unless it's a VooDoo4 3dfx card

Cygni
Nov 12, 2005

raring to post

Comfy Fleece Sweater posted:

I'm not putting any component in my PC that consumes 450W lol come on

Not unless it's a VooDoo4 3dfx card

Good news! It’s only 15w!

https://www.techpowerup.com/gpu-specs/voodoo4-4500-agp.c3548

Shipon
Nov 7, 2005
i got a 1200W PSU and by god I'm using every single watt, don't want to waste money after all

lih
May 15, 2013

Just a friendly reminder of what it looks like.

We'll do punctuation later.
https://twitter.com/kopite7kimi/status/1532207096116064256

don't worry everything is still up in the air, of course

knox_harrington
Feb 18, 2011

Running no point.

Paul MaudDib posted:



If you didn't understand that slide was showing iso-performance perf/w... that's kinda on you, because it's literally the entire focus of the slide. See the arrow? See where the lines are placed? Yeah, you have to actually read the chart, people - you not reading the chart doesn't make the chart BS.

The chart may not be outright lying, but that doesn't mean it isn't BS. Nobody is running a 3080 deliberately degraded like that so it's just not clinically* meaningful data.

*bedroomily

shrike82
Jun 11, 2005

even now when people are discussing a 2x perf uplift for the 4090, what metric are we talking about? some general fp32 throughput that's meaningless in a real world setting?
most people would take it to be tied to frame rates and more specifically, the uplift in frame rate from switching in the card without BS normalized /watt numbers

hobbesmaster
Jan 28, 2008

knox_harrington posted:

The chart may not be outright lying, but that doesn't mean it isn't BS. Nobody is running a 3080 deliberately degraded like that so it's just not clinically* meaningful data.

*bedroomily

It says ampere, not 3080, so it applies to the entire line.

As for the 3080 itself, all the undervolting discussions are about the entire top 1/3 of that ampere line.

lih
May 15, 2013

Just a friendly reminder of what it looks like.

We'll do punctuation later.

shrike82 posted:

even now when people are discussing a 2x perf uplift for the 4090, what metric are we talking about? some general fp32 throughput that's meaningless in a real world setting?
most people would take it to be tied to frame rates and more specifically, the uplift in frame rate from switching in the card without BS normalized /watt numbers
usually if people are talking about fp32 throughput then they specifically say so, otherwise it's a vaguer idea that's basically just framerates yeah

mobby_6kl
Aug 9, 2009

by Fluffdaddy

I think they meant Voodo 5 6000

Kerbtree
Sep 8, 2008

BAD FALCON!
LAZY!

Agreed posted:

PS3 emulation seems like it barely hits my GPU, big time CPU. But I was able to run some PS3 games at 60FPS / 4K with a 4770K at 4.3GHz and a 2070, whereas Yuzu has a lot of CPU-limited stuttering in more titles for me. I dunno, emulation is weird. But awesome.

PS3 has some very exotic hardware going on. It’s a bit like Vulkan, amazing if you can use it to full potential, risks being meh otherwise.

https://youtu.be/lLebZyha74o

Zedsdeadbaby
Jun 14, 2008

You have been called out, in the ways of old.
PS3 in theory was faster than 360, but there isn't a single multiplatform game that ran better on it. Back in those days it was a general rule of thumb always to get the 360 version, and only get PS3 for its exclusives.

The hosed up frankenstein's monster CPU is the reason why, Sony was expecting to be used in everything from toasters to consoles, it never came to be. Ease of development is why they went for off-the-shelf x86-64 platforms for PS4 and PS5 instead

The people behind PS3 emulators are incredibly clever people and we should be thankful they aren't exerting their energies on trying to achieve world domination.

Mozi
Apr 4, 2004

Forms change so fast
Time is moving past
Memory is smoke
Gonna get wider when I die
Nap Ghost
I'm looking at undervolting my 3080 FE, CPUz says my boost clock is 1710 but Afterburner stats say it hovers around 1830-1845 and peaks slightly over 1900. Which one should I go with?

Hm - -290mHz base clock, 862mV @ 1896mHz gets an extra 1fps avg in Heaven (MoE for sure), 2C less temp, 2% less fan speed but 25-35W less power consumption. Was hoping for more decrease in temps but I think I'll just take this.

Edit: had a little instability so went with 875mV @ 1905mHz. Honestly compared to stock there's a pretty small difference, basically just saving a little wattage.

Mozi fucked around with this message at 19:05 on Jun 2, 2022

Freakazoid_
Jul 5, 2013


Buglord

Best Buy now has this set to the same price, despite having none in stock. Very curious.

Rinkles
Oct 24, 2010

What I'm getting at is...
Do you feel the same way?

Zedsdeadbaby posted:

PS3 in theory was faster than 360, but there isn't a single multiplatform game that ran better on it. Back in those days it was a general rule of thumb always to get the 360 version, and only get PS3 for its exclusives.

The hosed up frankenstein's monster CPU is the reason why, Sony was expecting to be used in everything from toasters to consoles, it never came to be. Ease of development is why they went for off-the-shelf x86-64 platforms for PS4 and PS5 instead

The people behind PS3 emulators are incredibly clever people and we should be thankful they aren't exerting their energies on trying to achieve world domination.

I think Portal 2 might have run better on PS3. Plus it had MLAA afair.

Dr. Video Games 0031
Jul 17, 2004

You can still back-order the Gigabyte 3080 12GB for $800 on Newegg: https://www.newegg.com/gigabyte-geforce-rtx-3080-gv-n3080gaming-oc-12gd/p/N82E16814932489?Item=N82E16814932489
The line is probably short since the ship date is... yesterday.

The Gigabyte 3080 10GBs are down to $770 after a $30 promo code:
https://www.newegg.com/gigabyte-geforce-rtx-3080-gv-n3080gaming-oc-10gd/p/N82E16814932459?Item=N82E16814932459
https://www.newegg.com/gigabyte-geforce-rtx-3080-gv-n3080vision-oc-10gd/p/N82E16814932460?Item=N82E16814932460

With this, the cheapest 3080 is now just 10% overpriced.The 3080 Ti, 3090, and 3090 Ti are all selling for MSRP. Below the 3080, things start to look worse.

The cheapest 3070 Ti is $699 still, two months after Asus was the first to lower a model to that price (on april 1st), a 16% markup. The cheapest 3070 on newegg is $599, though I've seen $580 models briefly come into stock on Asus' website. So that's 16 - 20% overpriced.

The 3060 Ti was available at newegg on backorder for $480 for a day or so this week and is sometimes available on EVGA's webstore for that much, and there was an asus model available for $500. But those go out of stock pretty quickly and the more permanent price tends to hover around $530 - $550, or an appalling 37.5% overpriced. This is one of those cards where you have to look at buying direct from a manufacturer because the prices at big retailers are even worse.

The 3060 pricing also hasn't moved for the last 2 months when Asus suddenly slashed their prices. The cheapest models still tend to be $390, with more permanent pricing at around $410 - $440 a lot of the time, though right now you can get a Gigabyte Eagle OC or EVGA XC Black for $400 on newegg, which is 25% overpriced. Micro Center also tends to have a ton of the single-fan Asus model for $390. My local micro center is sitting on over 100 3060s right now, and I think some of the stagnation is due to retailer reluctance to lower prices. In comparison, they have a total of 4 3060 Tis in stock. That card flies off the shelves at almost any price.

The 3050 is a joke lmao. For whatever reason, it stubbornly stays at or above $330, which is 32% overpriced. My local micro center has been sitting on over 125 3050s and has been for a while now. I'm confident they could sell these for a lower price and it would stay in stock, but they just don't want to. The most expensive 3050 is the ROG Strix at a mind boggling $430. Just why? Who is buying that?

Dr. Video Games 0031 fucked around with this message at 19:02 on Jun 2, 2022

acksplode
May 17, 2004



Rinkles posted:

I think Portal 2 might have run better on PS3. Plus it had MLAA afair.

Vanquish ran a little better on PS3, and notably right after the same studio made Bayonetta, which had a notoriously bad PS3 port. The difference was that Bayonetta targeted 360 and a different studio handled the PS3 port, while Vanquish's primary platform was PS3 and Platinum handled the 360 port in house.

Comfy Fleece Sweater
Apr 2, 2013

You see, but you do not observe.


:haw: I thought the Voodoo series ended around version 3 but there you go

repiv
Aug 13, 2009

Taima posted:

I’ve been so impressed with the Sony pc releases in general. At least, once they’re patched to work. You can’t argue with the sheer production value of their games, then you add on the extra horsepower on PC and its just a beautiful thing.

speak of the devil, someone on reddit just dropped screenshots of Sackboy and Returnals unannounced PC versions

https://imagizer.imageshack.com/img922/4348/LVzt0n.png
https://imagizer.imageshack.com/img923/2878/l4ePo0.png

https://imagizer.imageshack.com/img924/7259/ScVab1.png
https://imagizer.imageshack.com/img921/8641/TeHJF2.png
https://imagizer.imageshack.com/img922/1774/DxhK5I.png

BOOTY-ADE
Aug 30, 2006

BIG KOOL TELLIN' Y'ALL TO KEEP IT TIGHT

Comfy Fleece Sweater posted:

:haw: I thought the Voodoo series ended around version 3 but there you go

Fun fact, might've already been mentioned - there's at least one guy who's actively making (maybe refurbishing?) 3Dfx cards still :)

https://www.facebook.com/zxc64.hw

Comfy Fleece Sweater
Apr 2, 2013

You see, but you do not observe.

BOOTY-ADE posted:

Fun fact, might've already been mentioned - there's at least one guy who's actively making (maybe refurbishing?) 3Dfx cards still :)

https://www.facebook.com/zxc64.hw

Well now I have to update my 20 year old jokes!!!! :11tea:

Zedsdeadbaby
Jun 14, 2008

You have been called out, in the ways of old.
Sony games on Windows are of better build quality than Microsoft's own. You only have to compare God of War to Halo Infinite to see what I mean. I don't know what is making Sony change their minds on PC (it's probably money) but I like it. They are not half-arsing things unlike Microsoft

This is on the technical side of things to be clear

Mr. Neutron
Sep 15, 2012

~I'M THE BEST~

Zedsdeadbaby posted:

Sony games on Windows are of better build quality than Microsoft's own. You only have to compare God of War to Halo Infinite to see what I mean. I don't know what is making Sony change their minds on PC (it's probably money) but I like it. They are not half-arsing things unlike Microsoft

This is on the technical side of things to be clear

FH5 (and 4 before that) are perfect PC ports so it's not a rule.

Adbot
ADBOT LOVES YOU

FlamingLiberal
Jan 18, 2009

Would you like to play a game?



Zedsdeadbaby posted:

Sony games on Windows are of better build quality than Microsoft's own. You only have to compare God of War to Halo Infinite to see what I mean. I don't know what is making Sony change their minds on PC (it's probably money) but I like it. They are not half-arsing things unlike Microsoft

This is on the technical side of things to be clear
Isn't that more on 343 just being a bad dev

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply