Wednesday, May 25th 2016

NVIDIA GeForce GTX 1080 Ti to be Based on GP102 Silicon

It looks like NVIDIA will have not one, but two "big chips" based on the "Pascal" architecture. The first one of course is the GP100, which made its debut with the Tesla P100 HPC processor. The GP100 is an expensive chip at the outset, featuring a combination of FP32 (single-precision) and FP64 (double-precision) CUDA cores, running up to 3,840 SPFP and 1,920 DPFP, working out to a gargantuan 5,760 CUDA core count. FP64 CUDA cores are practically useless on the consumer-graphics space, particularly in the hands of gamers. The GP100 also features a swanky 4096-bit HBM2 memory interface, with stacked memory dies sitting on the GPU package, making up an expensive multi-chip module. NVIDIA also doesn't want its product development cycle to be held hostage by HBM2 market availability and yields.

NVIDIA hence thinks there's room for a middle-ground between the super-complex GP100, and the rather simple GP104, if a price-war with AMD should make it impossible to sell a GP100-based SKU at $650-ish. Enter the GP102. This ASIC will be targeted at consumer graphics, making up GeForce GTX products, including the GTX 1080 Ti. It is cost-effective, in that it does away with the FP64 CUDA cores found on the GP100, retaining just a 3,840 FP32 CUDA cores count, 33% higher than that of the GP104, just as the GM200 had 33% more CUDA cores than the GM204.

It could also not be improbable that NVIDIA could use the more readily available GDDR5X memory interface on this chip. It remains to be seen if the GTX 1080 Ti features all 3,840 CUDA cores present on this chip, or if some are disabled to improve yields. It will also be interesting to see on what chip (the GP100 or GP102) the next GTX TITAN SKU will be based on. Source: WCCFTech
Add your own comment

57 Comments on NVIDIA GeForce GTX 1080 Ti to be Based on GP102 Silicon

#1
the54thvoid
In before @medi01 but probably going to side with him. That WCCFTECH article is an editorial piece. It's not news....
Posted on Reply
#2
Ferrum Master
"the54thvoid said:
In before @medi01 but probably going to side with him. That WCCFTECH article is an editorial piece. It's not news....
Too much yellow press lately here... not welcome in my books.
Posted on Reply
#3
matar
Buy 2 GTX 1070 and SLi looks better in your case and faster and same price.
Posted on Reply
#4
the54thvoid
"Ferrum Master said:
Too much yellow press lately here... not welcome in my books.
I'll agree on this article. I got it via Google Now and the author, although sure of his sources does say implicitly it's an editorial, therefore it shouldn't be reposted as news.

And yes, as much as I like TPU and as much as I'll fight fanboy posts, some articles shouldn't be front page, rather left in the forums.
Posted on Reply
#5
Ferrum Master
"LittleFellow said:
:rockout:
Here comes the first zerg wave.
Posted on Reply
#6
dj-electric
A 384-Bit GDDR5X memory capability should bring a card like this high on the performance scale
Posted on Reply
#7
atomicus
"...if a price-war with AMD should make it impossible to sell a GP100-based SKU at $650-ish"

HA, yeah like there's a snowball's chance in hell that will happen! HBM2 cards are going to cost the moon, neither AMD or Nvidia will have any interest in selling them for $650... nearer $1000 is more likely because they know people will pay, just as they are happy to pay $700 for a basic Reference (sorry, 'Founders Edition') 1080! The 1080Ti is going to easily be an $800 product, so I don't know what crazy strength ghetto crack people are smoking if they think the GP100 based cards are going to be $650... PMSL!
Posted on Reply
#8
medi01
Standards are secondary, each site decides which to follow, I am fine with that.

But at least they should be the same across the board, it's a shame when "news" about "silhouette" of GTX card is posted and at the same time leaked benchmark results about competitor isn't worth an article.


"atomicus said:
HA, yeah like there's a snowball's chance in hell that will happen! HBM2 cards are going to cost the moon
What's the point in even producing them, then? I mean, would there be even remotely as big performance increase due to higher memory bandwidth, to justify that?

"Caring1 said:
...and still no confirmation of those supposed facts...
Very likely to be just guesstimates by some random dude on the internet.
Posted on Reply
#9
Caring1
"LittleFellow said:
:rockout:
Only seen that posted here three times now, and still no confirmation of those supposed facts.
Posted on Reply
#10
atomicus
"medi01 said:
What's the point in even producing them, then? I mean, would there be even remotely as big performance increase due to higher memory bandwidth, to justify that?
Are you kidding lol? MONEY! They will make an absolute FORTUNE from the next gen cards (not that they haven't been doing so for years now anyway). They will rev the hype machine in to overdrive and whatever astronomic price they charge, people will pay and they will fly of the shelves. It makes no difference if people need them or if there's even a big enough performance increase to justify it... Nvidia will produce the figures and charts to show otherwise, which thousands will lap up. LOL, there are people buying x2 1080's who are still gaming at 1080p... the world and half the people in it are clinically insane and Nvidia are laughing all the way to the bank with ear to ear grins on their faces.
Posted on Reply
#11
iO
While this might be pure speculation, a GP102 is a lot more plausible than the GP100. There is no need for NVLink or mixed-precision etc in a consumer card.
Posted on Reply
#12
Naito
"Caring1 said:
Only seen that posted here three times now, and still no confirmation of those supposed facts.
Well there had been hints of a GP104 variant (à la GP104-150-A1) to slot in under the GTX 1070 and that table shows it. The specs are plausible. Though, I find it hard to believe Nvidia would have a variant with half the shaders disabled unless 16nm has led to a lot of bins. You'd think it would make more sense to bring in a GP106 or something. We already know what the GP100 consists of, so it's not hard to fathom a GP102. Perhaps saved for Pascal refresh? Suppose at the end of the day anyone can make a table from speculation....
Posted on Reply
#13
P4-630
The Way It's Meant to be Played
"atomicus said:
"...if a price-war with AMD should make it impossible to sell a GP100-based SKU at $650-ish"

HA, yeah like there's a snowball's chance in hell that will happen! HBM2 cards are going to cost the moon, neither AMD or Nvidia will have any interest in selling them for $650... nearer $1000 is more likely because they know people will pay, just as they are happy to pay $700 for a basic Reference (sorry, 'Founders Edition') 1080! The 1080Ti is going to easily be an $800 product, so I don't know what crazy strength ghetto crack people are smoking if they think the GP100 based cards are going to be $650... PMSL!
A GTX1080 FE costs already around 800 Euros in my country which is $892....:shadedshu:
Posted on Reply
#14
Chaitanya
"P4-630 said:
A GTX1080 FE costs already around 800 Euros in my country which is $892....:shadedshu:
In India its already above $1000. Somehow nvidia keeps jacking prices of their gpus and it seems to work well for them as there are a lot of rearside kissers who are willing to pay those prices.
Posted on Reply
#15
P4-630
The Way It's Meant to be Played
"Chaitanya said:
In India its already above $1000. Somehow nvidia keeps jacking prices of their gpus and it seems to work well for them as there are a lot of rearside kissers who are willing to pay those prices.
WTF , it seems only Americans can get their new computer hardware cheap!!
Posted on Reply
#16
bug
Pigs could conceivably fly, but has anyone noticed there hasn't been a Gx102 or Gx202 chip like ever? Unless Nvidia theselves announce a GP102, I'm going to go out on a limb and say it won't happen.
Posted on Reply
#17
Maban
GP102 with GDDR5X would allow them to put out a new card this year. HBM2 is holding back the mass availability of the P100 until early next year. It could be conceivable that GP102 is the Ti and maybe even a Titan, then next year there could be a GP100 based Titan with the glorious HBM2.
Posted on Reply
#18
Nihilus
"P4-630 said:
WTF , it seems only Americans can get their new computer hardware cheap!!
Maybe it has something to do with the fees and taxes the corporations need to pay the Socialist Eutopias of Europe.
Posted on Reply
#19
Slizzo
"LittleFellow said:
:rockout:
Oh FFS. I've had to comment on this a lot on reddit.

This "article" is not an article as pointed out. The source for this is a POST ON A MESAGE FORUM, where the subject is "let's speculate on what a GP102 chip COULD LOOK LIKE".
Posted on Reply
#20
truth teller
"medi01 said:
But at least they should be the same across the board, it's a shame when "news" about "silhouette" of GTX card is posted and at the same time leaked benchmark results about competitor isn't worth an article.
yeah, this has been happening for quite some time now (more than a year), ever since btarunr bragged about his large news/post count and implied that he could pretty much post whatever rocks his boat as news due to his long time contributions

kinda sad really, at least have some consistency and dont put stuff up as news
Posted on Reply
#21
HisDivineOrder
So what this means is that they're reclassified the x50 Ti-x60 level product as a x70-80 product and slotted the traditional x70-80 product for 80 Ti-Titan. Well, nVidia's going to be making tons of money and I'm sure in a couple of years they'll put them back the way they originally were...

...unless they're making too much money to bother.

Net Effect: They get more for less than they used to.
Posted on Reply
#22
bug
"HisDivineOrder said:
So what this means is that they're reclassified the x50 Ti-x60 level product as a x70-80 product and slotted the traditional x70-80 product for 80 Ti-Titan. Well, nVidia's going to be making tons of money and I'm sure in a couple of years they'll put them back the way they originally were...

...unless they're making too much money to bother.

Net Effect: They get more for less than they used to.
What "this" are you referring to? All I see is (unfounded) speculation and you attempting to draw some conclusions.
Posted on Reply
#23
Slizzo
Not only that, but since the GTX 680 was launched, this is how nVidia has done things. Just because AMD can't compete as quickly as nVidia can doesn't mean that what nVidia is doing is wrong. They have a chance to stretch out their products and they're taking it.
Posted on Reply
#24
Frick
Fishfaced Nincompoop
"truth teller said:
yeah, this has been happening for quite some time now (more than a year), ever since btarunr bragged about his large news/post count and implied that he could pretty much post whatever rocks his boat as news due to his long time contributions

kinda sad really, at least have some consistency and dont put stuff up as news
At least he doesn't pass it up as "hard hitting journalism", as other contributors in the past. :laugh:
Posted on Reply
#25
Ferrum Master
"Nihilus said:
Maybe it has something to do with the fees and taxes the corporations need to pay the Socialist Eutopias of Europe.
DAFUQ?
Posted on Reply
Add your own comment