• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

NVIDIA Pascal GPU to feature 17B transistors and 32GB HBM2, coming Q2 2016 (earliest)

I don't think so, your talking about current games, not games 5 years from now tailor made for those hypothetical consoles. I expect devs to target 30fps even then, because they are going to want to push the boundary of a closed platform by default.

Even though we should not be taking Crytek's advice, they said that even if they had put Ryse on PS4, they still would have made it 900p and used the GPU cycles elsewhere, to push their graphics instead of making it 1080p with the extra GPU grunt.

Of course it hinges on whether or not devs actually want to use the power for that kind of image quality and framerate. I wasn't addressing that in my comment. I was simply stating that the power is there.
 

ShirAhava

Plays with kids toys, in the adult gaming world
Nice....In my skylake build this year I guess I'll just settle for a GTX 970 (maybe even 960) and wait for this Q4 2016/Q1 2017 (I really doubt the early 2016 date)
 
So anywhere from April-June this should be released? I guess I will hold on building my PC until then and become an early adopter.
 

mr2xxx

Banned
Node shrink, new architecture, and new memory format. To many things can happen to delay it. I'd be happy if it comes out by Q4.
 

Mirk

Member
Thank god! I want to upgrade my pc and have it mean something again. Now if intel can give me something good by the time this comes out.
 

Zoned

Actively hates charity
LMAO what if it slips to late 2016, which it probably will. My 980 Ti is like the greatest gap filler card ever.

Yeah that's what people say in every GPU rumor thread. When we were in 400 Fermi, people said Kepler is gonna be be the definite upgrade. Then in kepler, Maxwell is the way to go. And now maxwell is there, Pascal is the definite upgrade. This is a never ending trail. Better to get one than keep waiting for the "definitive one".
 
I'm primarily a console gamer so upgrading my computer hasn't been a pressing need for quite some time. But...VR has piqued my interest so I've been both trying to talk myself into getting a new computer and trying to hold out for as long as I can in order to get the best bang for my buck. Unfortunately I angered some god recently and a transformer went crazy in my neighborhood, messing up the electricity, and frying my computer. So now I pretty much have to get a computer but the thing I really want for my computer, VR, doesn't come out until next year.

I think the best plan I can come up with is to get a new computer but not buy a graphics card. I'll use the integrated GPU on the motherboard and then spring for a nice discrete GPU a year or so from now.
 
Gonna wait to see how Arctic Islands on 14nm or 16nm FF is before I grab one of these Pascal GPUs. If they're as good as it sounds right now, we'll be in for some really insane performance from both AMD and Nvidia
 
i'm not even sure if this is something to be celebrated at the moment given how quick developers are taking advantage of increased gpu memory for an excuse at poor overall optimization
 

dr_rus

Member
No way is this coming out Q2 2016. I'd be surprised if pascal hits at all in 2016.

Pascal is coming in 2016, that's for sure. But Pascal is an architecture and that big chip based on it may not show up in GeForce line in 2016 at all. I'm expecting that the first Pascal for GeForces will be "GP104" again - it should be a little bit faster than GM200 while consuming less and it will come in 2Q 2016. The "Big Pascal" may actually tape out first but its usage will be limited to HPC / supercomputers. We may or may not get a new Titan based on it by the end of 2016 but anything like 780Ti and 980Ti is a story for 2017 probably.

I reall hope this time they will close the gap:

and we get 100% same performance between the mobile and the desktop GPUs.. Seeing how the Pascal architecture is and how it can fit in a such really small die:

PASCAL-4.png


I am sure they will close the gap this time.
The die on your photo is huge, and the reason why desktop GPUs are more powerful than mobile counterparts is power - this reason won't go away any time soon if at all. Mobile PCs simply do not have these 600 watts PSUs.

There's not always a full two node transition with 4x the memory and 4x the memory bandwidth.
16FF+ is 20nm physically so there is no "two full nodes transition" and if you expect any consumer videocard from 2016 to have 32GBs of RAM on it you will be disappointed. Bandwidth is the real deal breaker though as HBM2 should actually increase it 2-3 times -- first GDDR5 generations was the last time we had something like this.
 

AmyS

Member
Holy shit.

Even if Pascal launches in Q3 2016, it'll be a most welcome upgrade from the 28nm chains we've been shackled to for years.
 

x3sphere

Member
Pascal is coming in 2016, that's for sure. But Pascal is an architecture and that big chip based on it may not show up in GeForce line in 2016 at all. I'm expecting that the first Pascal for GeForces will be "GP104" again - it should be a little bit faster than GM200 while consuming less and it will come in 2Q 2016. The "Big Pascal" may actually tape out first but its usage will be limited to HPC / supercomputers. We may or may not get a new Titan based on it by the end of 2016 but anything like 780Ti and 980Ti is a story for 2017 probably.
.

Little bit faster? Due to node shrink + HBM2, I actually expect that "mid range" Pascal will be a similar jump as 780 Ti -> 980 Ti - at least at high resolutions. If that is the case NV will probably hold off launching the big chip for consumer market until 2017.
 
I'm not super pessimistic about the prices of the mid range cards. Everyone was worried about 900 prices titan after titan that people were buying anyway, and then boom $330 cards and $500 cards.
 

ZOONAMI

Junior Member
I'm not super pessimistic about the prices of the mid range cards. Everyone was worried about 900 prices titan after titan that people were buying anyway, and then boom $330 cards and $500 cards.

Yeah for a falsely advertised gimped 970 card and a $550-$600 midrange 980 sold as high end.
 

x3sphere

Member
I'm not super pessimistic about the prices of the mid range cards. Everyone was worried about 900 prices titan after titan that people were buying anyway, and then boom $330 cards and $500 cards.

Yeah, I don't think Pascal will be launching first with a $1000 Titan.

NV may put more on of a premium on these cards, due to 16nm, but I think they'll price them reasonably enough that someone with a 980 Ti for example will want to upgrade. Then they'll make that buyer want to upgrade again when big Pascal comes :p

I guess we'll see what happens as we get closer to launch. I'm going to try and skip the mid range parts this time though, big Pascal will be worth waiting for.
 
I don't believe we will see 32 on a consumer card for quite a while. Workstation gpu maybe, but theres a lot of assumptions in this rumour. 16 for titan sounds about right, 8 for cut down version.
 

Kieli

Member
Yeah for a falsely advertised gimped 970 card and a $550-$600 midrange 980 sold as high end.

Not nVIDIA's fault AMD can't get their shit together.

And the fact that too many people are willing to pay those prices because they have too much disposable income.
 

dr_rus

Member
Little bit faster? Due to node shrink + HBM2, I actually expect that "mid range" Pascal will be a similar jump as 780 Ti -> 980 Ti - at least at high resolutions. If that is the case NV will probably hold off launching the big chip for consumer market until 2017.

The node shrink will give you 980Ti in "mid range". Add some architectural improvements and HBM2 benefits and you'll have something in range of GM200+20% on average (probably better than that in high resolutions because of HBM2). Now couple the high costs of the new production process and you have a new 680/7970 situation on your hands - a "GP104" GPU sold in higher end of the market.

I obviously don't know how it will turn out but everyone should keep their expectations in check. It's highly doubtful that costs will allow them to launch the big Pascal chip for gaming market in 2016. And it is quite possible that for 2016 they'll use this "GP104" type chip as their top offering.
 

Durante

Member
It's not out of a question that there will be such a GPU with that memory configuration next year.

But most likely in supercomputers, not in your desktop.
 

Tagyhag

Member
LMAO what if it slips to late 2016, which it probably will. My 980 Ti is like the greatest gap filler card ever.

I can wait then, not like there's a Crysis-like game right around the corner.

Star Citizen won't be until late 2016 at the earliest either way.

I'm not fortunate enough to be able to spend >$650 just for a gap filler. :x
 

magnumpy

Member
Too bad that card is not HBM1 @8GB but gddr5. And of course it would be 8GB at reference, because if it was 4GB, it's just an overclocked 290x.

sorry, got the names mixed up. I'm talking about this

v7k8CBM.jpg


this is a desktop card. if this isn't outside the realm of possibility, then nvidia better have a desktop card to answer it or they will be left behind.
 

x3sphere

Member
sorry, got the names mixed up. I'm talking about this


this is a desktop card. if this isn't outside the realm of possibility, then nvidia better have a desktop card to answer it or they will be left behind.

The Fury only has 4GB of HBM. I don't know what you're talking about in regards to "nvidia better have a desktop card to answer it." They already do in performance, albeit not in that form factor.
 

InfernoNR

Member
I'm beginning to believe I'll never upgrade my rig. Every announcement from card manufacturers leaves me wanting to wait for the next batch of card releases
 

Despera

Banned
Dammit! I was planning on finally putting together a beast of a rig by the end of this year with perhaps 2 Titan X's in SLI in preparation for VR. But I guess I'm better off waiting a few months until this comes out. Sounds like a huge leap from current tech.
 

MultiCore

Member
I reall hope this time they will close the gap:




and we get 100% same performance between the mobile and the desktop GPUs.. Seeing how the Pascal architecture is and how it can fit in a such really small die:

PASCAL-4.png


I am sure they will close the gap this time.

I think you're insane for wanting to hamstring desktop cards to mobile TDP limitations.

That's why the 900 series cards are a screw job to begin with, and you want to make it worse?

You've got tons more power and thermal dissipation available on a desktop, and you want a card that runs at 100w max for your high end desktop card?
 

AmyS

Member
Sounds neat! Depending on how big a jump it is in terms of game performance, I might consider upgrading. Otherwise I'm waiting for Volta.

Volta is probably not going to show up in GeForce cards until a year after it arrives for the HPC market, so 2018 for Volta based gaming cards.

With Pascal GPU, NVIDIA will return to the HPC market with new Tesla products. Maxwell, although great in all regards was deprived of necessary FP64 hardware and focused only on FP32 performance. This meant that the chip was going to stay away from HPC markets while NVIDIA offered their year old Kepler based cards as the only Tesla based options. Pascal will not only improve FP64 performance but also feature mixed precision that allows NVIDIA cards to compute at 16-bit at double the accuracy of FP32. This means that the cards will enable three tiers of compute at FP16, FP32 and FP64. NVIDIA’s far future Volta GPU will further leverage the compute architecture as it is already planned to be part of the SUMMIT and Sierra super computers that feature over 150 PetaFlops of compute performance and launch in 2017 which indicates the launch of Volta just a year after Pascal for the HPC market.

http://wccftech.com/nvidia-pascal-gpu-17-billion-transistors-32-gb-hbm2-vram-arrives-in-2016/
 
Top Bottom