[GPU] NVIDIA Pascal GP100 Spotted in Transit, moving to testing phase

Gattberserk

Arch-Supremacy Member
Joined
Jan 26, 2008
Messages
10,590
Reaction score
901
http://www.techpowerup.com/216203/nvidia-gp100-silicon-moves-to-testing-phase.html.


NVIDIA's next-generation flagship graphics processor, codenamed "GP100," has reportedly graduated to testing phase. That is when a limited batch of completed chips are sent from the foundry partner to NVIDIA for testing and evaluation. The chips tripped speed-traps on changeover airports, on their way to NVIDIA. 3DCenter.org predicts that the GP100, based on the company's "Pascal" GPU architecture, will feature no less than 17 billion transistors, and will be built on the 16 nm FinFET+ node at TSMC. The GP100 will feature an HBM2 memory interface. HBM2 allows you to cram up to 32 GB of memory. The flagship product based on GP100 could feature about 16 GB of memory. NVIDIA's design goal could be to squeeze out anywhere between 60-90% higher performance than the current-generation flagship GTX TITAN-X.

I hope the wait was well worth it.
 

ragnarok95

Senior Moderator
Senior Moderator
Joined
Mar 9, 2004
Messages
125,422
Reaction score
6,657
Every year same same thing. Expect more than 50% than previous. :D
 

enenyi

Supremacy Member
Joined
Nov 22, 2011
Messages
5,927
Reaction score
5
16GB!!!! :D

Hopefully Pascal will be able to do asynchronous shaders which is what killed Maxwell in DX12, compared to AMD GCN architecture.
Well expect that in Quadro first. If successful then only they may think of allowing the GP200 series GeForce cards to adopt. :D
 

Gattberserk

Arch-Supremacy Member
Joined
Jan 26, 2008
Messages
10,590
Reaction score
901
Every year same same thing. Expect more than 50% than previous. :D

This time round NV CEO mentioned 10x more performance improvement. So I believe 60-90% is more plausible.


And dont forget the biggest reason for the jump is the transition to 16nm FF+. Can pack 2x more transistors means performance can double. That is based solely on new node alone.


Of cos we can argue that NV is not going to give that the max potential of Pascal at the head start.


But 60-90% is not impossible with the combination of HBM2, new architecture, new Node. But expect the Titan series for Pascal to go beyond 999USD if they really offer that tremendous improvement


They are made for the rich of the riches.
 

MoneyFace =p

High Supremacy Member
Joined
Dec 17, 2010
Messages
35,380
Reaction score
402
This time round NV CEO mentioned 10x more performance improvement. So I believe 60-90% is more plausible.


And dont forget the biggest reason for the jump is the transition to 16nm FF+. Can pack 2x more transistors means performance can double. That is based solely on new node alone.


Of cos we can argue that NV is not going to give that the max potential of Pascal at the head start.


But 60-90% is not impossible with the combination of HBM2, new architecture, new Node. But expect the Titan series for Pascal to go beyond 999USD if they really offer that tremendous improvement


They are made for the rich of the riches.
But don't forget real-life perceivable gains are quite dismay compared to synthetic benchmarks.
 

Gattberserk

Arch-Supremacy Member
Joined
Jan 26, 2008
Messages
10,590
Reaction score
901
But don't forget real-life perceivable gains are quite dismay compared to synthetic benchmarks.

That is true, but den again there is confirm noticable performance gain. Maxwell already did that against Kepler, I would expect Pascal to do a better job in fact.
 

enenyi

Supremacy Member
Joined
Nov 22, 2011
Messages
5,927
Reaction score
5
This time round NV CEO mentioned 10x more performance improvement. So I believe 60-90% is more plausible.


And dont forget the biggest reason for the jump is the transition to 16nm FF+. Can pack 2x more transistors means performance can double. That is based solely on new node alone.


Of cos we can argue that NV is not going to give that the max potential of Pascal at the head start.


But 60-90% is not impossible with the combination of HBM2, new architecture, new Node. But expect the Titan series for Pascal to go beyond 999USD if they really offer that tremendous improvement


They are made for the rich of the riches.
I understand that u are super excited for Nvidia Pascal Titan. But if u step back and think through, hope u won't have too high hopes for a card that meets ur performance and pricing expectations, for 2 possible scenarios:

1) Nvidia ends up over-hyping their Pascal series, like wad AMD did with the Fury series. If u think about it, Nvidia's transition into 16nm chips is like Intel's transition into 14nm chips. And we are talking about silicon chips, not germanium or some other semiconductor material. If through Skylake, Intel only shows about 20% improvement over their previous 22nm chips, realistically how much performance gain can Nvidia gain with a transition to 16nm? True, double the transistor count, but with that means that more transistors are packed tightly together, which also translates to higher temps. Higher temps means GPU Boost cant really work effectively. Even with HBM2, Nvidia will not hit the advertised performance increase if their performance gained from efficiency of architecture is offset by higher temps. So we are back to square one, and like wad we are seeing with Skylake, there is only so much net improvements we can see from the 16nm chips.

2) Even if Nvidia did manage to get GP100 chips that have 60-90% increase in performance, given their dominance in the market and their track record, they will most likely milk their customers. So we may be staring at a further segregation of the Titan line of products, with Titan X's successor as the 2nd tier product and the full GP100 chip as the top tier product. Given Titan X is already S$1600 range, with a 20% max increase in performance, Titan X successor should also range about S$1700? And knowing Nvidia's pricing habits, the top tier Titan will go for more than S$2000? So rationally speaking, would u wanna buy a top tier card that costs prob S$500 more for that perhaps 30% increase in performance over a 2nd tier card? Would u rather spend something like S$2500 on a full Titan Pascal chip or go for cheaper dual GPU options?

Just something I hope u would think about. In case ur hopes get really dashed come Q2 2016.
 

Humster

Arch-Supremacy Member
Joined
Jul 15, 2000
Messages
23,507
Reaction score
2,648
Don't bother la, the price is gonna be sooo high it is not even funny anymore.

Ain;t gonna surprise me that they are moving the prices of their enthusiast G-card closer to those professional-grade card.

Since they can always throw around the "Pro Gaming" tag.
 

Gattberserk

Arch-Supremacy Member
Joined
Jan 26, 2008
Messages
10,590
Reaction score
901
I understand that u are super excited for Nvidia Pascal Titan. But if u step back and think through, hope u won't have too high hopes for a card that meets ur performance and pricing expectations, for 2 possible scenarios:

1) Nvidia ends up over-hyping their Pascal series, like wad AMD did with the Fury series. If u think about it, Nvidia's transition into 16nm chips is like Intel's transition into 14nm chips. And we are talking about silicon chips, not germanium or some other semiconductor material. If through Skylake, Intel only shows about 20% improvement over their previous 22nm chips, realistically how much performance gain can Nvidia gain with a transition to 16nm? True, double the transistor count, but with that means that more transistors are packed tightly together, which also translates to higher temps. Higher temps means GPU Boost cant really work effectively. Even with HBM2, Nvidia will not hit the advertised performance increase if their performance gained from efficiency of architecture is offset by higher temps. So we are back to square one, and like wad we are seeing with Skylake, there is only so much net improvements we can see from the 16nm chips.

2) Even if Nvidia did manage to get GP100 chips that have 60-90% increase in performance, given their dominance in the market and their track record, they will most likely milk their customers. So we may be staring at a further segregation of the Titan line of products, with Titan X's successor as the 2nd tier product and the full GP100 chip as the top tier product. Given Titan X is already S$1600 range, with a 20% max increase in performance, Titan X successor should also range about S$1700? And knowing Nvidia's pricing habits, the top tier Titan will go for more than S$2000? So rationally speaking, would u wanna buy a top tier card that costs prob S$500 more for that perhaps 30% increase in performance over a 2nd tier card? Would u rather spend something like S$2500 on a full Titan Pascal chip or go for cheaper dual GPU options?

Just something I hope u would think about. In case ur hopes get really dashed come Q2 2016.

I doubt intel is a good eg to compare for 14nm since we suspect their mediocre performance improvement is due to lack of competition from AMD side.

Actually architecture changes is what give the performance boost. If pascal is about to deliver 20% for the same cores compare to maxwell, couple that with 33% more shader u could easily net a 60% clock for clock boost.


Of cos like u said price... Make me wonder how much higher nvidia will go this time. They came out with titan Z at ridiculous 3k usd before, so any value below that, is not remotely impossible.
 
Last edited:

Mikiberry

Supremacy Member
Joined
Jun 26, 2010
Messages
5,107
Reaction score
0
Actually wait for the actual card to come out then decide lor lol. Same goes for Fury X fiasco when launched.
 

royfrosty

High Supremacy Member
Joined
Jul 23, 2009
Messages
27,735
Reaction score
1,558
Hbm nothing new. You can put so much on memory bandwidth, its just gonna be fast moving vram. Its just more about how the new architecture is gonna perform. If its fast good. If its not then its just another gpu.

From the trend that i have seen. Nvidia always have performance boost from each gap generation is only 25%-40% increament range for each top tier cards.
 
Important Forum Advisory Note
This forum is moderated by volunteer moderators who will react only to members' feedback on posts. Moderators are not employees or representatives of HWZ Forums. Forum members and moderators are responsible for their own posts. Please refer to our Community Guidelines and Standards and Terms and Conditions for more information.
Top