Volta could be released before Vega (Q3 2017)

[Alantor]

The source is Chinese site MyDrivers, which has proven to be a reliable source before. Do you believe this? If this is true Vega will be DOA.

[Fascinate]

I dont see that possible, the amount of R and D that went into pascal was tremendous. Unless its just an iteration on that (die shrink etc), the article is bunk.

[Nellah]

I think it's plausible if not probable, since Pascal was released sixteen months after Maxwell and Pascal has now been out for eleven. I'm betting on a Christmas launch though.

[Gaidax]

I dont see that possible, the amount of R and D that went into pascal was tremendous. Unless its just an iteration on that (die shrink etc), the article is bunk.

Pascal is around for a year now, surely by now they are pretty advanced in Volta development.

Time flies.

But dayum, if that happens? lulz Vega will be DOA, it's already on a thin ice vs beastly 1080Ti, Volta will rip it.

[Deleted]

The source is Chinese site MyDrivers, which has proven to be a reliable source before. Do you believe this? If this is true Vega will be DOA.

But dayum, if that happens? lulz Vega will be DOA, it's already on a thin ice vs beastly 1080Ti, Volta will rip it.

I don't think at this point anyone is in position to judge that. We do not know what gains will be seen from volta, we also are not clear on how vega will perform. The way I see it AMD is expecting to be competing against volta - based on their jab in vega teaser "poor volta".

[Fascinate]

Pascal is around for a year now, surely by now they are pretty advanced in Volta development.

Time flies.

But dayum, if that happens? lulz Vega will be DOA, it's already on a thin ice vs beastly 1080Ti, Volta will rip it.

Not sure you understand the scale of pascals development, if volta is a thing this year its not going be a leap like maxwell>pascal was. My guess is volta isnt a new architecture, its pascal on a different processing node. If it is a brand new architecture its only going to be slightly faster than pascal.

[Denpepe]

Not sure you understand the scale of pascals development, if volta is a thing this year its not going be a leap like maxwell>pascal was. My guess is volta isnt a new architecture, its pascal on a different processing node. If it is a brand new architecture its only going to be slightly faster than pascal.

I was under the impression that volta was pushed back for reasons and pascal (according to some maxwell 2.0) introduced to fill the gap so that would make it possible but it would be pretty fast after launching the 1080ti and the other ones with speedier memory.

Then again, the AMD RX 400 series is already beeing replaced by the 500 series and was launched after pascal.

[Alantor]

I don't think at this point anyone is in position to judge that. We do not know what gains will be seen from volta, we also are not clear on how vega will perform. The way I see it AMD is expecting to be competing against volta - based on their jab in vega teaser "poor volta".

I thought Vega was supposed to compete against Pascal, my mistake I guess :P

[Gaidax]

Just got a call that my 1080Ti is ready for pickup. Guess I'll get to enjoy it for a couple of months before it becomes pleb card, lulz.

Also yeah, I thought year ago Vega DoA edition was supposed to come out in April, but I just got tired of waiting, my 7 years red streak is going to end today.

[Vash The Stampede]

I dont see that possible, the amount of R and D that went into pascal was tremendous. Unless its just an iteration on that (die shrink etc), the article is bunk.

From what I remember Pascal was a minor upgrade over Maxwell. Other than some minor tweaks and 16nm manufacturing, it's mostly a refresh of Maxwell. So Nvidia has had much time to do R&D while AMD is still trying to catch up to Maxwell.

[Fascinate]

From what I remember Pascal was a minor upgrade over Maxwell. Other than some minor tweaks and 16nm manufacturing, it's mostly a refresh of Maxwell. So Nvidia has had much time to do R&D while AMD is still trying to catch up to Maxwell.

Not sure where you are getting that, Gtx 1080 is twice as fast as a 980, it was the largest performance leap by far from any generation. I distinctly remember from the 1080 press conference the CEO saying pascal was years in development and some ridiculous amount of money invested. That is why i say volta isnt going to be much of anything, probably just a refresh on a different node.

[Lathais]

Not sure where you are getting that, Gtx 1080 is twice as fast as a 980, it was the largest performance leap by far from any generation. I distinctly remember from the 1080 press conference the CEO saying pascal was years in development and some ridiculous amount of money invested. That is why i say volta isnt going to be much of anything, probably just a refresh on a different node.

Yeah, it's faster because of the manufacturing process change. The architecture itself was almost identical, just smaller. They went from 28nm manufacturing process, skipped a few and went all the way down to 16nm. Of course there will be huge gains from that alone. And that's about all it was. There were a couple minor tweaks, I don't recall what they were, but nothing major.

[Vash The Stampede]

Pascal was something Nvidia farted out. Have to admit, they didn't need to redesign a GPU when a small refresh of Maxwell was enough to still dominate the market. The RX 480 is barely equal to GTX 1060 in DX11 games, and only edges out in DX12/Vulkan games. AMD could have made a more powerful Polaris chip to compete with the 1070 and 1080, but AMD knows that people would rather buy Nvidia. In order for AMD to put a dent in the market, they need something powerful. Hence Vega. Nvidia meanwhile is working on Volta while their Pascal chip just prints money.

Can't say Nvidia wasn't smart in how they went about it.

[Lathais]

Pascal was something Nvidia farted out. Have to admit, they didn't need to redesign a GPU when a small refresh of Maxwell was enough to still dominate the market. The RX 480 is barely equal to GTX 1060 in DX11 games, and only edges out in DX12/Vulkan games. AMD could have made a more powerful Polaris chip to compete with the 1070 and 1080, but AMD knows that people would rather buy Nvidia. In order for AMD to put a dent in the market, they need something powerful. Hence Vega. Nvidia meanwhile is working on Volta while their Pascal chip just prints money.

Can't say Nvidia wasn't smart in how they went about it.

Well, to be a little fair to nvidia, even if they don't really deserve it, when using a new manufacturing process for the first time, and such a large jump down too, probably a good idea to go with a more mature architecture. More things to go wrong if using a new architecture along with a new manufacturing process. So Pascal was kind of just a test of the process, now they have some experience with it and can try out a new architecture on it.

[Fascinate]

Yeah, it's faster because of the manufacturing process change. The architecture itself was almost identical, just smaller. They went from 28nm manufacturing process, skipped a few and went all the way down to 16nm. Of course there will be huge gains from that alone. And that's about all it was. There were a couple minor tweaks, I don't recall what they were, but nothing major.

Pascal was more than just a die shrink, how else do you explain it being the single largest performance bump from any generation in nvidias history?

The point here is, dont expect volta to be a greater leap in performance than maxwell>pascal, cause that aint happening.....at least not this year as the article indicates.

[Lathais]

Pascal was more than just a die shrink, how else do you explain it being the single largest performance bump from any generation in nvidias history?

The point here is, dont expect volta to be a greater leap in performance than maxwell>pascal, cause that aint happening.....at least not this year as the article indicates.

There were some very minor architecture improvements, but yes, that large of a shrink in the manufacturing process can absolutely explain the large performance bump.

[Gaidax]

https://imgur.com/nyZ6fp2

This gonna be good, time to put my old 290X horse to rest.

[Alantor]

The point here is, dont expect volta to be a greater leap in performance than maxwell>pascal, cause that aint happening.....at least not this year as the article indicates.

Yeah, those "I think we'll see..." numbers are just made up. The interesting part, at least for me, is that Nvidia might not launch Volta in 2018. So we have Pascal uncontested right now, AMD launch Vega this summer (or whenever) and right after we get the new green cards. Graphically it would be something like this: Veeegaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa.

[DeltrusDisc]

Pascal was something Nvidia farted out. Have to admit, they didn't need to redesign a GPU when a small refresh of Maxwell was enough to still dominate the market. The RX 480 is barely equal to GTX 1060 in DX11 games, and only edges out in DX12/Vulkan games. AMD could have made a more powerful Polaris chip to compete with the 1070 and 1080, but AMD knows that people would rather buy Nvidia. In order for AMD to put a dent in the market, they need something powerful. Hence Vega. Nvidia meanwhile is working on Volta while their Pascal chip just prints money.

Can't say Nvidia wasn't smart in how they went about it.

A die shrink wouldn't get you upgrades like this between xx80 Ti iterations. It just wouldn't.

http://www.anandtech.com/bench/product/1856?vs=1715

A bit bigger than the jump from the 780 Ti to 980 Ti, even...

[Life-Binder]

dont think any gaming Voltas will hit in 2017

even if they do - probably GV106, not even GV104


but expecting the first Volta Titan by mid-2018 seems reasonable to me, maybe August-September 2018 at latest





P.S. Latest rumors are also saying that Intel has moved up Skylake-X to June (at least some of the models) and Coffee Lake mainstream to August 2017

- - - Updated - - -

Pascal has improvements specifically for DX12/Vulkan over Maxwell

on DX11 its generally identical clock for clock core for core