NVIDIA Pascal Thread

Page 95 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
For what it's worth the leaked pics of the two GP104 samples both had the date stamp as 1614, or in other words the 14th week of 2016, which would be just 2 weeks ago.

So the samples might only have gotten into the hands of OEMs very recently, which might help explain the lack of leaks so far.

Do they send QA samples that were fabbed 2 weeks ago to the AIB's? Wouldn't that be something for in house or possibly Pony who make the reference boards?
 
Feb 19, 2009
10,457
10
76
GK104 and GM204 performance didn't leak until days before release. But hey, Nvidia hasn't demonstrated working silicon so they're obviously months behind AMD, right?!

I don't recall those but Titan X certainly was leaked much earlier.

It leaked 6 months earlier in charts that were spot on accurate with the power usage down to the T. But folks ignored it, like yourself, because you were expecting Titan X to be much faster than the 980 ("it's got 50% more of everything!"), more than 30% anyway and so you didn't believe it.

The pictures of the die was also leaked ~ a month before release, when we find out it was big, and the chip, isn't a QA sample but a mass production chip.

Fiji was also leaked many months earlier, with performance metrics on par with Titan X for the 22 games that it was leaked. This is what I am talking about when I mean a detailed leak, lots of game benches. Not a QA sample chip pic, with QA GDDR5X... this is something that AIBs would be testing QA for, a long way from volume production.
 
Last edited:

tviceman

Diamond Member
Mar 25, 2008
6,734
514
126
www.facebook.com
I don't recall those but Titan X certainly was leaked much earlier.

It leaked 6 months earlier in charts that were spot on accurate with the power usage down to the T. But folks ignored it, like yourself, because you were expecting Titan X to be much faster than the 980 ("it's got 50% more of everything!"), more than 30% anyway and so you didn't believe it.

The pictures of the die was also leaked ~ a month before release, when we find out it was big, and the chip, isn't a QA sample but a mass production chip.

Fiji was also leaked many months earlier, with performance metrics on par with Titan X for the 22 games that it was leaked. This is what I am talking about when I mean a detailed leak, lots of game benches. Not a QA sample chip pic, with QA GDDR5X... this is something that AIBs would be testing QA for, a long way from volume production.

Ah yes, I remember that Titan X leak. That leak was never verified or backed up. In fact, I'm pretty sure it's performance was scaled up over a 980 exactly the same as the performance of the OG Titan was over the 680. It was also accompanied with a 960 TI leak too, which turned out to be false. The point? Throw enough darts and one is bound to hit a bulls-eye. Nvidia has held close new product family launches. Fermi 1st gen didn't leak, Kepler 1st gen didn't leak, and the 900 series didn't leak. The 750 TI did leak, and if you want to call that Titan X / 960 TI chart a leak then sure, whatever.

What are you arguing about anyways? That the Pascal release can't be a month out because there hasn't been performance leaks or that Nvidia didn't demonstrate working silicon while doing jumping jacks months ago like AMD? OK.
 

casiofx

Senior member
Mar 24, 2015
369
36
61
If nvidia have it why wouldn't they launch it as titan again? $999 card but limited numbers first with lotsa memory and dp compute, and then release a $550 or $650 geforce.

Instead of building the volume for geforce launch.
 

antihelten

Golden Member
Feb 2, 2012
1,764
274
126
If nvidia have it why wouldn't they launch it as titan again? $999 card but limited numbers first with lotsa memory and dp compute, and then release a $550 or $650 geforce.

Instead of building the volume for geforce launch.

I'm assuming that you're talking about GP100 here. The simple answer is that as long as they can sell all of the GP100 dies they can make in the form of DGX-1 ($129.000 for 8 GP100s), Tesla cards ($5000+ per GP100 die) or Quadro cards (~$5000 per GP100 die), then there is no reason to start selling it as either a Titan card ($1000 per die) or Geforce card ($650-700 per die).

The question is then what Nvidia will do once yields improve and the initial demand from the HPC and professional market starts drying up. Normally at this point Nvidia could then launch the GPU as a Geforce card, but the problem is that if this point is too far out then it might not be competitive as a gaming GPU anymore (relative to the fact that it's a 600mm2 die).

If you're talking about GP104, then the answer is probably that launching a 300-350mm2 die as a titan card is a bit much, even for Nvidia. Also GP104 will probably have crappy DP performance.
 

airfathaaaaa

Senior member
Feb 12, 2016
692
12
81
seriously who cares about dp perf when it comes to games? there is no engine atm that needs that kind of power to be utilised
 

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
No one, but the original Titan card was also sold in part based upon it's DP capability as a sort of Tesla-lite card.

Not to gamers, it wasn't. We had posters on this and other boards use it to justify the price, but it's simply not used in games. The latest Titan doesn't have it and people bought it anyway.

They can't release it without HBM2. Once we have it we'll see what happens.
 

antihelten

Golden Member
Feb 2, 2012
1,764
274
126
Not to gamers, it wasn't. We had posters on this and other boards use it to justify the price, but it's simply not used in games. The latest Titan doesn't have it and people bought it anyway.

They can't release it without HBM2. Once we have it we'll see what happens.

Eerm, yeah that's what I said (the "no one" part), and obviously gamers don't buy Tesla like cards, they have no use for it.

Point is though that Titan did also have uses outside of gaming. I'm perfectly aware that some people who bought the card for gaming tried to use the DP capability as justification for said purchase even though they would never make use of it, but that doesn't mean that no one ever made use of said capability.

The Titan X is of course a different story given it's lack of DP capability, but where talking about the GP100 here, so that's not really relevant.
 

Timmah!

Golden Member
Jul 24, 2010
1,512
824
136
Eerm, yeah that's what I said (the "no one" part), and obviously gamers don't buy Tesla like cards, they have no use for it.

Point is though that Titan did also have uses outside of gaming. I'm perfectly aware that some people who bought the card for gaming tried to use the DP capability as justification for said purchase even though they would never make use of it, but that doesn't mean that no one ever made use of said capability.

The Titan X is of course a different story given it's lack of DP capability, but where talking about the GP100 here, so that's not really relevant.

Every Geforce has use outside of gaming thanks to CUDA. The DP compute tasks are not the only ones in existence.

The main advantage of Titan X over 980Ti was the double of VRAM. Thats something you definitely want to have (at least with my kind of "compute" stuff) and probably worth the price difference between the 2. If the best Pascal GPU to buy for the rest of 2016 and even after that gonna be GP104 with less VRAM than TX, only maybe 20 percent faster and at a price of 980Ti (which is 700 - 800 EUROs over here), then i guess i am keeping my money, cause i am not offered better product overall, which is something i expected to happen, when i decided not to buy Maxwell card last year.
 

antihelten

Golden Member
Feb 2, 2012
1,764
274
126
Every Geforce has use outside of gaming thanks to CUDA. The DP compute tasks are not the only ones in existence.

The main advantage of Titan X over 980Ti was the double of VRAM. Thats something you definitely want to have (at least with my kind of "compute" stuff) and probably worth the price difference between the 2. If the best Pascal GPU to buy for the rest of 2016 and even after that gonna be GP104 with less VRAM than TX, only maybe 20 percent faster and at a price of 980Ti (which is 700 - 800 EUROs over here), then i guess i am keeping my money, cause i am not offered better product overall, which is something i expected to happen, when i decided not to buy Maxwell card last year.

Of course Geforce can also be used outside of gaming, but in the case of Titan, it was mainly its DP capability that set it apart from other Geforce cards when it comes to non-gaming usage.

Of course with Titan X, this feature went down the drain. The additional memory is kinda meh in my opinion, since you can often get normal Geforce cards with double the normal amount of memory, so not really unique to Titan X.
 

Ajay

Lifer
Jan 8, 2001
16,094
8,109
136
One thing I've been thinking about is that TSMC's 16FF+ wafers are quite a bit more expensive than their 28nm wafers. Makes me think that the GP104-200 w/GDDR5 could be the new GTX 980 in terms of price and relative performance. I would think that NV would want this out in time for back to school sales, if possible. Then we have to wait until Nvidia accumulates enough GDDR5X to release a new 980Ti as early as possible for Christmas sales. It will be interesting to see when the first cut down GP100 dice make it into a Titan class GPU w/HMB2.

This would give NV plenty of room to bump GPU/Mem clocks and/or Cuda cores for the next generation AIBs on 16FF+. Source - my gut + Nvidia's profit motives.
 

Timmah!

Golden Member
Jul 24, 2010
1,512
824
136
Of course Geforce can also be used outside of gaming, but in the case of Titan, it was mainly its DP capability that set it apart from other Geforce cards when it comes to non-gaming usage.

Of course with Titan X, this feature went down the drain. The additional memory is kinda meh in my opinion, since you can often get normal Geforce cards with double the normal amount of memory, so not really unique to Titan X.

Maybe in the past. Nowadays, the next best thing is 980Ti at 6GB. Thats a massive difference.
 

jpiniero

Lifer
Oct 1, 2010
15,176
5,717
136
I would think that NV would want this out in time for back to school sales, if possible. Then we have to wait until Nvidia accumulates enough GDDR5X to release a new 980Ti as early as possible for Christmas sales.

It appears that nVidia is not holding back announcing a GDDR5X model but will be in short supply obviously until the end of the year.

It will be interesting to see when the first cut down GP100 dice make it into a Titan class GPU w/HMB2.

This isn't happening. A Titan with HBM2 is what GP102 is for and should be enough to compete with Vega. I would not be surprised if Vega 10 is like GP100 in that it won't be sold as a consumer product, only for HPC and workstations.
 

Mopetar

Diamond Member
Jan 31, 2011
8,110
6,754
136
This probably won't be happening.

Eventually it might, but the more I think about it I doubt it's anytime soon. Part of the reason is that we already know that NVidia has started P100 production, and that thing is going to eat wafers in order to produce chips even if they're cut. GP104 is supposed to launch relatively soon so either it's currently in production or needs to production needs to ramp soon if they want to have a hard launch.

NVidia can rest on their mobile laurels a little longer because AMDs recent performance there has been utterly terrible, but eventually they'll need to get GP106 or GP108 out to compete with Polaris 11, which even if it's not architecturally great or doesn't close the existing gap with Maxwell, it will still have a huge process node advantage.

TSMC is likely going to keep adding capacity as time goes on, but right now NVidia probably doesn't have access to enough wafers to start on GP100 anytime soon. They're probably using P100 as a good barometer for when to start GP100 or what design choices should be made to get better chips as a result, but any big chip they're going to try to make right now needs the $12,000 price tag to justify the wafers needed to produce it.

My best guess is that we don't see GP100 in any form until late 2017.
 
Feb 19, 2009
10,457
10
76
This isn't happening. A Titan with HBM2 is what GP102 is for and should be enough to compete with Vega. I would not be surprised if Vega 10 is like GP100 in that it won't be sold as a consumer product, only for HPC and workstations.

If there's excess wafers for NV, they can release big Pascal as a GTX. Otherwise it makes no sense to sell GP100 for 10% of the price as a GTX. They don't need to respond until 2017 anyway since Vega isn't due this year.

AMD's FP16 & FP64 implementation differs to NV. Look at Hawaii.
 

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
Eerm, yeah that's what I said (the "no one" part), and obviously gamers don't buy Tesla like cards, they have no use for it.

Point is though that Titan did also have uses outside of gaming. I'm perfectly aware that some people who bought the card for gaming tried to use the DP capability as justification for said purchase even though they would never make use of it, but that doesn't mean that no one ever made use of said capability.

The Titan X is of course a different story given it's lack of DP capability, but where talking about the GP100 here, so that's not really relevant.

I was referencing it to Titan's DP capability. I understand that some used it, but not gamers. It's like Tahiti's DP capability. It had 1/3, which decimated any other gaming card (until Titan, which I believe was the same)and I'm sure some took advantage of it. It was never a big marketing point though nor justification to choose it over GK104 for users on this board. Mining and OpenCL were touted as useful for some here, but many still gave it no value for some reason.
 
Mar 10, 2006
11,715
2,012
126
It will if nVidia needs it to compete against Vega and has nothing else. Maybe the GP102 is a real thing though, and that will be the next Titan.

Listen to NVIDIA's analyst day comments. GP100 was built specifically for HPC. Those FP64 units are totally wasteful on a GPU intended to be used in gaming.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |