• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

[Beyond3D Forums] GP102 and GP106 coming next quarter

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
They did do a lot of work to optimise Pascal 🙂 It wasn't remarkable of flashy or anything but probably ultimately more useful.

They're also doing a tick/tock thing. Volta is coming at some point of course, and seems like it that might change more things. Imagine that it takes quite a while to get really working well though.

NV did and incredible job optimizing Maxwell on 28nm. I think NV got a lot of benefits for Pascal from TSMC's excellent 16FF process. Volta will need allot of work to improve on Pascal on the same process node.

It too bad, for consumers, that AMD had to go with GF's 14 LPP node - it's really a mobile process that just doesn't serve them well. Based on Raja Koduri's comments to PCPer Vega, and I assume Zen, will be using an improved node - not sure what that is exactly yet).

PCPer Interview:
 
Last edited:
Probably hardly any faster, which is why I think they will go straight to GV104 after GP102.

Yeah, I don't imagine it would be. The one exception I could see is that if the DP parts of the GPU can be very effectively power gated to where the GPU core itself has the same power profile as GP102, and then the power savings of HBM allows GP100 to be clocked higher for a given power level than GP102. I don't see that as being really likely though.

I think a GP100 Titan could be incoming, but they maybe just skip making a Prosumer Titan again and just go with Titan being ultra-premium gaming, and put the GP100 into a Quadro for workstation use.
 
It too bad, for consumers, that AMD had to go with GF's 14 LPP node - it's really a mobile process that just doesn't serve them well. Based on Raja Koduri's comments to PCPer Vega, and I assume Zen, will be using an improved node - not sure what that is exactly yet).

Is it really that bad? It looks like AMD's been able to get good yield, and that means cheaper prices even if that's bought with extra wattage.
 
Right now, it seems yes. Those yields might be OK, but they're cutting some quite nasty corners to get them. Some fairly rubbish quality - really quite variable TDP's, some very high - chips in the first batch of 480's.

They can sort of get away with that in a budget desktop SKU, although honestly I can't see how it rates as a good idea to do so with the effects on reviews etc, but.....
 
Yeah, I don't imagine it would be. The one exception I could see is that if the DP parts of the GPU can be very effectively power gated to where the GPU core itself has the same power profile as GP102, and then the power savings of HBM allows GP100 to be clocked higher for a given power level than GP102. I don't see that as being really likely though.

I think a GP100 Titan could be incoming, but they maybe just skip making a Prosumer Titan again and just go with Titan being ultra-premium gaming, and put the GP100 into a Quadro for workstation use.

GP100 will be slower for gaming than GP102, no reason this will ever be sold as a Titan/GTX 1080 Ti.
 
Yeah, I don't imagine it would be. The one exception I could see is that if the DP parts of the GPU can be very effectively power gated to where the GPU core itself has the same power profile as GP102, and then the power savings of HBM allows GP100 to be clocked higher for a given power level than GP102. I don't see that as being really likely though.

I think a GP100 Titan could be incoming, but they maybe just skip making a Prosumer Titan again and just go with Titan being ultra-premium gaming, and put the GP100 into a Quadro for workstation use.

Being that GP100 has no ROPs, I would not expect it to ever be put into a video card. GP102 will be filling that roll.
 
Good stuff, looking forward to seeing both teams fill out their lineup. It's like the early GCN/Kepler days all over again 😀
 
I'd guess the 102 shows up when AMDs big chip shows up. Most likely next Spring.

It would be wise for NVIDIA to bring it sooner, especially since the GPU was taped out as of early May. The earlier it has it out in the market, the more it can charge for it since it will have essentially zero competition.
 
It would be wise for NVIDIA to bring it sooner, especially since the GPU was taped out as of early May. The earlier it has it out in the market, the more it can charge for it since it will have essentially zero competition.

At this point they would be competing with themselves. Bringing out a 102 only drives down their 1080 and 1070 prices. If this taped out in May(I didnt read article). Then Dec is the earliest I would expect it. But honestly I dont expect it until AMD does something. And while there are rumors of AMD brining out their big chip late this year. I will be shocked if it shows up before next spring in the march time frame.
 
Don't forget that they've got the Titan, with its enormous price tag attached as an option (which they're bound to use).

That lets them push the big chip out without really displacing the 1080.
 
That would be an interesting approach; I wonder how they would actually go about it. GP100 is a natural Titan in the same vein as the original. It could have great applications as a prosumer card even if they gimp the DP from the full Tesla version.
For gaming through, I'm not sure how much faster (if any) an HBM2 GP100 Titan would be with even a full 3840 CCs vs a 384bit GDDR5X GP102 with 3840 CCs.

My guess is if they drop a GP100 Titan ~9 mo. prior to Volta it would be a fully unlocked one to get those last 4 SMs turned on to give it the edge + whatever benefit there is from HBM2. Either that or GP102 is not released in a fully unlocked fashion off the bat, a-la 8800 GT

I do really think though that they've got contigency plans in waiting for a GP100 Titan if Vega is faster than anticipated and can beat gp102. nVidia seems too well prepared to not have considered it

I'm w/ Sweepr though if Vega does not top GP102 then they will likely go straight to GV 104
 
Last edited:
GP102 Titan won't cause 1070/1080 price drop. Price $1099 for ~30% performance increase over 1080. Problem solved for NV, barring any threat from AMD.
 
At this point they would be competing with themselves. Bringing out a 102 only drives down their 1080 and 1070 prices.

I dont think so. They will just charge $899 and $1299 for for 1080ti and TITAN XXX. Why not? Enough people will probably pay it. They could even try $999 and $1499.
 
I dont think so. They will just charge $899 and $1299 for for 1080ti and TITAN XXX. Why not? Enough people will probably pay it. They could even try $999 and $1499.

So what do they bring out when AMD brings out their new chip?
 
So what do they bring out when AMD brings out their new chip?

Last gen it was AMD who was left answer to NV, 980 Ti v Fury X. No one knows what timeline each will follow. AMD isn't rushing to match 1070/1080 perf, probably safe to assume they won't rush to market with a GP102 competitor either (if these Q4 rumors are viable). NV has been running the high-end show for years now, theres no secret there.
 
Interesting that you bring Jen-Hsun up. When somebody asked him about Polaris on NVIDIA's last earnings call, he said this:



I look forward to seeing how his claims hold up when GP106 goes up against Polaris 10 and later Vega against GP104/GP102. Exciting times for GPU buyers from both camps. We all benefit from companies working tirelessly to build excellent products. :thumbsup:

He seems very confidant and I don't blame him. As of late, nvidia is beating AMD to the punch and counter-punch. It's almost like they have a mole working at AMD.
 
GP102 Titan won't cause 1070/1080 price drop. Price $1099 for ~30% performance increase over 1080. Problem solved for NV, barring any threat from AMD.

Probably closer to 35%. The 1080Ti will need to be 25-30% to be able to justify upgrading for 1080 owners. And the Titan needs to still be a tad ahead performance wise when 1080Ti is released a few months later for the Titan brand to continue to be hot stuff.
 
Probably closer to 35%. The 1080Ti will need to be 25-30% to be able to justify upgrading for 1080 owners. And the Titan needs to still be a tad ahead performance wise when 1080Ti is released a few months later for the Titan brand to continue to be hot stuff.

Thus the tilde. And who knows how long it will be until a Ti variant comes to market. If AMD has no answer for the Titan within the next year, NV wont have a reason to release a cut down Titan enthusiast chip.

We all know NV loves all that is green.
 
Thus the tilde. And who knows how long it will be until a Ti variant comes to market. If AMD has no answer for the Titan within the next year, NV wont have a reason to release a cut down Titan enthusiast chip.

We all know NV loves all that is green.

Same with every other corporation. nVidia just happens to be in the position to best exploit that level of greed. It's always the "underdog" that's viewed like they are "for the people" it's all a mirage until they're in a favorable position.
 
Same with every other corporation. nVidia just happens to be in the position to best exploit that level of greed. It's always the "underdog" that's viewed like they are "for the people" it's all a mirage until they're in a favorable position.

This is true. People probably dont remember or never went through the 3DFX vs Nvidia days. Where 3DFX was the king of discrete graphics cards. One Nvidia eliminated 3DFX that under dog title moved to ATI.
 
Back
Top