NVIDIA Pascal Thread

[jpiniero]

despite only small die size increase (GK104 -> GM204 etc),

This is not true. GK104 is 294 mm2. GM204 is 398 mm2.

I really dont care how big GP104 is.If its not delivering +50% performance on brand new 16FF+ it will be epic fail(vs 970 and 980)

You have to understand, both companies are going to need to milk this node for 3-5 years. Between that and the little transistor/$ cost benefit of 16FF+ over 28 they have to be very stingy unless they want to raise prices by an awful lot suddenly. Don't think that's a good idea.

 

[el etro]

A bet a bit under 300mm^2 for GP104. Performance i don't have an idea, just think GP100 will be 2x faster and 2x performance/watt versus GM200.

 

[tential]

Where?
http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_980_Ti/31.html
1080P-21%faster
1440p-25%faster
VS GTX970
1080P-35%faster.

Did you really just try to quote a review from May 31st 2015?

Find something more relevant seriously....

 

[iiiankiii]

Lets see:
Less heat and power,
Same performance as GTX 980Ti but ~140-150W
$300-350 price vs 980Ti`s $650?
Async support?
Full DX12 support

Sign me up for 1 or two of those GTX 1070s

Historically, the GTX X70 were normally priced at $399. It was the GTX 970 that was priced at $339. It was surprising. I expect Nvidia to go back to $399 for the GTX X70 and $499 for the GTX X80. That makes more sense to me.

 

[tential]

That's in line with my expectations. A ~200-300mm² GP104 beating 601mm² GM200 would be no small feat.

At what resolution? We notice cards go further away from the 970 as resolution increases. If these results are at 1440p... It's not relevant to the "I'm a GTX 970 1080p gamer" crowd (Not my opinion, what I see regularly that hte 970 is meant for 1080p). So you can say a card is xyz faster but it depends on resolution. The GTX 980Ti is 35% fastser than the 970 at 1080p but almost 50% faster at 1440p. So we need benches lol....

 

[Kris194]

GTX 670 was about 30% faster than GTX 580, depends on game/application (40 vs 28 nm). How can anyone be happy with GTX 1070 being on par with GTX 980 Ti/Titan X?

I expect GTX 1070 to be at least 30% faster than GTX Titan X. If not, it will be the worst die shrink ever seen.

 

[Glo.]

Because it is basically the same architecture with different name.

 

[crisium]

The 670 was also barely cut back from the 680 - same ROPs, same Buswidth, just 1/8 fewer shaders. Nvidia is going for larger gaps between cards now.

770 to 780 was massive (50% more everything, although much lower clock rates), and the more recent 970 to 980 is still much larger than 670 to 680. 970 has a bigger shader gap than the 670 had and it also has fewer ROPs and cutdown memory bandwidth.

I'm expecting them to make GTX 1070 just a few percent faster than the 980 Ti / TX, single digit % maybe. GTX 1080 has a lot more room to go, as it will be the full midrange chip.

 

[ShintaiDK]

GTX 670 was about 30% faster than GTX 580, depends on game/application (40 vs 28 nm). How can anyone be happy with GTX 1070 being on par with GTX 980 Ti/Titan X?

Because of transistor cost. The key advantage now is performance/watt.

 

[Timmah!]

Historically, the GTX X70 were normally priced at $399. It was the GTX 970 that was priced at $339. It was surprising. I expect Nvidia to go back to $399 for the GTX X70 and $499 for the GTX X80. That makes more sense to me.

Historically, x70 chips were second/third fastest cards of their generation, not fourth/fifth (depending on the existence of dual-chip card in the product line-up).

 

[crisium]

By historically, you mean Fermi? That's the only time where the x70 was second fastest (well, GTX 200 series I suppose, if you count the 275 as second to 285). From Fermi naming, the GTX 980 would be the 560 Ti, and the 970 the 560. But the naming has shifted. The x70 can enjoy temporary status as the #2 at best while the big chip is still cooking in the lab.

 

[iiiankiii]

Historically, x70 chips were second/third fastest cards of their generation, not fourth/fifth (depending on the existence of dual-chip card in the product line-up).

Oh, I know. But, I was talking about pricing . In that respect, Nvidia is pretty much in line. That's why they introduced the "TI" version. It's a way to raise the price on us, the suckers.

 

[SteveGrabowski]

GTX 670 was about 30% faster than GTX 580, depends on game/application (40 vs 28 nm). How can anyone be happy with GTX 1070 being on par with GTX 980 Ti/Titan X?

I expect GTX 1070 to be at least 30% faster than GTX Titan X. If not, it will be the worst die shrink ever seen.

At launch the 680 was only 23% better than the 580 though.

But that gap was 30% 3 months later when the 670 launched. But yeah, I'd be disappointed if the X70 isn't a decent amount better than a 980 Ti.

 

[3DVagabond]

People on these boards, or similar ones. The same way they are now excited about Polaris and itching to get it, or even 2 into crossfire, despite the fact its almost given at this point it wont be really faster than the current AMD gpus, which ofc they already own (or comparable Nvidia stuff).

I think the majority is not so enthusiastic about new HW and does upgrade, only when truly needed.

I was talking about people on this board, or people like them. Then here are the people who don't upgrade every gen, or even 2nd. People running Pitcairn, Kepler, older VLIW hardware. People who need a card with a single 6pin. There is a market for improved perf/$/W.

 

[3DVagabond]

Historically, x70 chips were second/third fastest cards of their generation, not fourth/fifth (depending on the existence of dual-chip card in the product line-up).

Yes. X70 should have been X60 but the performance increase was large enough the last 2 times to up it a notch in the pricing hierarchy. In the past the 780/980 ti would have been the X70.

 

[nenforcer]

Historically, the GTX X70 were normally priced at $399. It was the GTX 970 that was priced at $339. It was surprising. I expect Nvidia to go back to $399 for the GTX X70 and $499 for the GTX X80. That makes more sense to me.

Your exactly right - they will want all of the people with GTX 970 to upgrade to a faster card - and the X70 at $399 will provide the performance increase along with a price increase.

 

[moonbogg]

I need Pascal GPU's that are big and hot as hell. None of this power efficient crap. I want something to make my loop work hard.

 

[Silverforce11]

I need Pascal GPU's that are big and hot as hell. None of this power efficient crap. I want something to make my loop work hard.

You could always Quad SLI, similar power profile to SLI, but extra performance!

 

[mysticjbyrd]

Any word on whether pascal will properly support async compute?

 

[Silverforce11]

Any word on whether pascal will properly support async compute?

You won't know until its in the hands of reviewers for testing. Since we obviously can't trust the specs released even if it's directly from NVIDIA, re: NV's 3.5 + 0.5GB and Async Compute claims.

But if they do have it (which I think Pascal will have hardware support), I would expect major Async Compute usage for new GameWorks features, it's gonna really reinforce the point for Kepler/Maxwell owners to upgrade to the latest & greatest.

I mean some folks here aren't happy with mid-range Pascal beating 980Ti/Titan X by ~20%, it could easily be 980Ti/Titan X +50% in DX12 if GameWorks pushes Async (which tanks on GPUs without support). Upgrade-worthy then? You bet.

 

[garagisti]

You could always Quad SLI, similar power profile to SLI, but extra performance!

Meh, just get two, and sell and upgrade later, to two big chips. The scaling is a questionmark, so why spend money for little gain?

Crossfire scales better when it works, and specifically gimpworks is not involved. However, i doubt this user intends to go that route. Eitherways, its all middle-tier stuff for now, smart money should wait.

 

[Kris194]

You won't know until its in the hands of reviewers for testing. Since we obviously can't trust the specs released even if it's directly from NVIDIA, re: NV's 3.5 + 0.5GB and Async Compute claims.

But if they do have it (which I think Pascal will have hardware support), I would expect major Async Compute usage for new GameWorks features, it's gonna really reinforce the point for Kepler/Maxwell owners to upgrade to the latest & greatest.

I mean some folks here aren't happy with mid-range Pascal beating 980Ti/Titan X by ~20%, it could easily be 980Ti/Titan X +50% in DX12 if GameWorks pushes Async (which tanks on GPUs without support). Upgrade-worthy then? You bet.

Single GTX 980 Ti struggles to reach constant 60 FPS on Ultra at 1080p in some games and that's why some people wouldn't be happy with mid-range Pascal beating Titan X by ~20%.

 

[Mondozei]

Single GTX 980 Ti struggles to reach constant 60 FPS on Ultra at 1080p in some games and that's why some people wouldn't be happy with mid-range Pascal beating Titan X by ~20%.

Unrealistic but quite telling how far away from reality people's expectations are if they seriously think a 20% adv for a X70 over a 980 Ti is now expected.

NV has been quite clear, as has AMD, that the biggest gains from the 16 nm node will be from power efficiency, not performance, even with the 16 FF+ node.

This is why false memes start, because people build up laughable expectations. I remember the "HBM will make Fury crush 980 Ti" hype, just last summer. People who said, wait a minute, GDDR5 probably is enough for most situations today, got attacked by the people on the hype train.

Then, when Fury came out, it was branded a "failure" by a lot of people who fell for the hype in the first place. Fury's failure was lack of OC, because it was competitve on normal clock speeds within a few months. It was nothing about HBM. Now we're seeing the exact same thing by the same suspects. It's kind of funny how some never learn.

 

[3DVagabond]

Unrealistic but quite telling how far away from reality people's expectations are if they seriously think a 20% adv for a X70 over a 980 Ti is now expected.

NV has been quite clear, as has AMD, that the biggest gains from the 16 nm node will be from power efficiency, not performance, even with the 16 FF+ node.

This is why false memes start, because people build up laughable expectations. I remember the "HBM will make Fury crush 980 Ti" hype, just last summer. People who said, wait a minute, GDDR5 probably is enough for most situations today, got attacked by the people on the hype train.

Then, when Fury came out, it was branded a "failure" by a lot of people who fell for the hype in the first place. Fury's failure was lack of OC, because it was competitve on normal clock speeds within a few months. It was nothing about HBM. Now we're seeing the exact same thing by the same suspects. It's kind of funny how some never learn.

Actually I see far more attacks by GDDR5 proponents than those for HBM. I'm responding to one right now.

 

[airfathaaaaa]

the first gen of those cards on the nodes will surely be more about power eff instead of perf but it would be stupid not to expect at least a 10-15% jump on perf from a flagship to another

also why this thread keep saying the cards will arrive on may? does OP have clairvoyance ?