GeForce 20 series to be released this year (Volta)

With luck there will be a decent card around $300 this time around. Pretty much anything sub $300 since the GTX 660ti was inadequate. Perhaps even more important is no Founders Edition, which simply inflated the price for the desirable cards by an extra $50-80. Aside from smaller form factor 1070s on sale, I don't think I've seen a single one down to the $350 MSRP.
 
The Volta equipped Drive PX with the Xavier SoC is 16nm FF+, but that doesn't necessarily mean that the same node will be used in desktop graphics. But even if they will use TSMC's 12nm node, it is still just a refinement of their 16nm FF+ process.

Shintai ninja'd :ninja:.

Yea I got Shintai on ignore thankfully I didn't see his nonsense.

Interesting so the Drive PV is 16nm......Hmmm Yea I think Volta is going to be 16nm. Unlesss TSMC pulled off a miracle.

Thanks for the info.
 
Yea I got Shintai on ignore thankfully I didn't see his nonsense.

Interesting so the Drive PV is 16nm......Hmmm Yea I think Volta is going to be 16nm. Unlesss TSMC pulled off a miracle.

Thanks for the info.
Well being 4th gen 16nm means they can launch on time without risk for their major contractual obligations; the '12nm' does not even have its own R&D project name from TSMC like true nodes/logic platform, highlighting an improvement to existing 16nm and low risk.
Rumour of Volta would be a smaller node than this was always unfortunately just that and did not fit well with the facts, this is part of what makes Volta very exciting from professionals in the HPC space as it has a very nice size jump in performance for being similar logic platform to Pascal, along with the arch improvments that are meant to be part of it as well.

Cheers
 
Last edited:
Not sure how launching sooner with higher prices is supposed to help sales.
 
Not sure how launching sooner with higher prices is supposed to help sales.

I think it is speculation; if Nvidia does confirm a small consumer launch sometime in Q4 as some of us expected or hints of it do become facts people will say a) AMD put them under pressure to do so or b) sales are weak

Cheers
 
Yeah. I'm expecting the same node, but hopefully it brings enough architectural improvements that it's a big boost over the 10x0 series. I like my 1070, but the performance isn't quite there yet for 4k.
 
The source seems to believe this is an "advance release" of Volta, meaning the timeline was moved forward. However, with the 1080 Ti's recent launch, Q3 being *way* sooner than anyone expected, and lack of competition from AMD, I'm going to predict it's actually a Pascal refresh... not Volta. I don't see how or why Nvidia would move their release schedule forward.

Once we see Vega, we should have a better idea of where Nvidia is.
 
Last edited:
Do we know anything about Volta yet, design wise, or are we stuck in speculation phase? Is it an improvement on Pascal or new arc? I'm in need of two graphics cards for two different builds but I can "settle" for a 1070 if I can expect 1080 Ti prices to come down more due to 2000 series launch.
 
Do we know anything about Volta yet, design wise, or are we stuck in speculation phase? Is it an improvement on Pascal or new arc? I'm in need of two graphics cards for two different builds but I can "settle" for a 1070 if I can expect 1080 Ti prices to come down more due to 2000 series launch.

Just lots of speculation thus far, but supposedly Volta's GPU Architecture is new and based on 12nm FinFET process, with its VRAM getting a bump to either GDDR6 (good) or possibly even HMB2 (great), the latter of which would result in significantly higher memory bandwidth (as high as 1TB/sec which would be well over twice the memory bandwidth of the 1080Ti.)
 
Impressive if it pans out. Seems pie-in-the-sky like though, having the same feel as the 580X perf rumors. "I want to believe," but Nvidia has never been that gracious with a next generation performance boost since I started building PCs nearing a decade ago.
 
Impressive if it pans out. Seems pie-in-the-sky like though, having the same feel as the 580X perf rumors. "I want to believe," but Nvidia has never been that gracious with a next generation performance boost since I started building PCs nearing a decade ago.
You're kidding right? They give massive performance improvements generation after generation.
 
I think some people are comparing a 980Ti to a 1080, when they should be comparing it to a 1080Ti since that's the direct replacement.

The 1080 was essentially a bit faster than 980 Ti at a slightly cheaper price. It was right in that area where it's not such a performance upgrade to warrant changing my 980 Ti. The 1080 Ti is a big improvement, but at the same time it's sitting at a 50-100€ higher price point and is just about enough for 4K @ 60 fps in current games. I am hoping that the first Volta cards would be the 1080 equivalent upgrade that pushes it to a very solid 4K @ 60 fps for the next few years at a price that is either slighly cheaper or the same as 1080 Ti on launch.
 
Just lots of speculation thus far, but supposedly Volta's GPU Architecture is new and based on 12nm FinFET process, with its VRAM getting a bump to either GDDR6 (good) or possibly even HMB2 (great), the latter of which would result in significantly higher memory bandwidth (as high as 1TB/sec which would be well over twice the memory bandwidth of the 1080Ti.)

HBM2 is quite useless and overly expensive unless you need ECC. And its very, very unlikely that anything but GV100 will feature HBM2.
 
The source seems to believe this is an "advance release" of Volta, meaning the timeline was moved forward. However, with the 1080 Ti's recent launch, Q3 being *way* sooner than anyone expected, and lack of competition from AMD, I'm going to predict it's actually a Pascal refresh... not Volta. I don't see how or why Nvidia would move their release schedule forward.

Once we see Vega, we should have a better idea of where Nvidia is.

It comes down to perspective IMO, and separately the interpretation of product cycle with assumptions made by some sites.
No-one made these type of conclusions with the Titan Pascal when it launched in August last year, and that was massively in advance of usual.
So I think they are jumping the gun saying Nvidia are bringing it forward (the original window may shift a bit either way due to everything going smoothly or some speed bumps), remember it is not really in Nvidia's interest to have these technologies out of sync in terms of R&D and also manufacturing (Pascal showed their intent on how to align these diverse segments); the consumer GPU is based upon the Tesla and that will always follow the cadence of aggressive development and launch these days with the tech war happening in HPC/deep learning/automotive (beyond cars).

We are sort of seeing a 'refresh' now for the 1060 and 1080 involving higher speed memory.
Cheers
 
Last edited:
You're kidding right? They give massive performance improvements generation after generation.

Like for like in real world applications? Not really. They are definitely measurable, repeatable improvements (which vary by SKU), but the real world gains are less impressive. We aren't talking GTX 680 to Titan gains in most cases.
 
Like for like in real world applications? Not really. They are definitely measurable, repeatable improvements (which vary by SKU), but the real world gains are less impressive. We aren't talking GTX 680 to Titan gains in most cases.

GTX980->GTX1080 pretty much doubled my performance.

Else you can claim hardly any gains the last 15 years too.
 
I wonder if they will name these the 2000s series card 2070 TI cards get away from thei 1080P stuff confusion =)

I probably won't upgrade just got a 1080ti took alot of adjustments to my Monitor and settings just to get my eyeballs use to the different card switched from DVI to HDMI for the first time ever on PC.
 
Like for like in real world applications? Not really. They are definitely measurable, repeatable improvements (which vary by SKU), but the real world gains are less impressive. We aren't talking GTX 680 to Titan gains in most cases.
The original Titan/780 was only about 33% faster than the 680. The 980 Ti was 50% faster than the 780 Ti, and the 1080 Ti is 70% faster than the 980 Ti. In fact, most Maxwell to Pascal gains were 70% or better across the product stack. If anything, performance gains from NVIDIA have been getting better each generation.
 
The 1080 was essentially a bit faster than 980 Ti at a slightly cheaper price. It was right in that area where it's not such a performance upgrade to warrant changing my 980 Ti. The 1080 Ti is a big improvement, but at the same time it's sitting at a 50-100€ higher price point and is just about enough for 4K @ 60 fps in current games. I am hoping that the first Volta cards would be the 1080 equivalent upgrade that pushes it to a very solid 4K @ 60 fps for the next few years at a price that is either slighly cheaper or the same as 1080 Ti on launch.
No. The 1080 is much faster
 
The source seems to believe this is an "advance release" of Volta, meaning the timeline was moved forward. However, with the 1080 Ti's recent launch, Q3 being *way* sooner than anyone expected, and lack of competition from AMD, I'm going to predict it's actually a Pascal refresh... not Volta. I don't see how or why Nvidia would move their release schedule forward.

Once we see Vega, we should have a better idea of where Nvidia is.
Well like someone speculated earlier it's about the money. launch earlier get more money/profit.
 
Like for like in real world applications? Not really. They are definitely measurable, repeatable improvements (which vary by SKU), but the real world gains are less impressive. We aren't talking GTX 680 to Titan gains in most cases.

What exactly do you mean by "real world applications?" because in games, a 1080Ti is about 70% faster than a 980Ti, but lets call it 50-60% since most 980Ti's can OC pretty well, still a big number, so what is "not really" based on?

I suppose if you're gaming at 1080 the benefits are fewer. At 1440p the benefits are substantial and at 4k it's a huge difference.
 
No. The 1080 is much faster

It really isn't. Almost every review uses a 980 Ti reference model at stock clocks so the 1080 looks a lot better in that comparison. The overclocked 980 Ti comes close to a stock 1080 and an overclocked 1080 pulls ahead something like 15-20% which IMO isn't worth an upgrade. I don't know about you but I generally want something that is at least 40-50% faster than my current card and only the 1080 Ti fits the bill.

In any case not worth fighting about as this thread is about Volta.
 
  • Like
Reactions: N4CR
like this
Like for like in real world applications? Not really. They are definitely measurable, repeatable improvements (which vary by SKU), but the real world gains are less impressive. We aren't talking GTX 680 to Titan gains in most cases.

Actually the performance gain from 680 to Titan is in the same ball park as 980 to GTX Titan X and 1080 to Titan X (first one).

If there is anything else to go by, the performance difference between 1080ti to 980ti (~70%) is greater by a good margin than 980ti is to 780ti (~40%), and this is including 780ti showing signs of VRAM starvation. Hell the performance difference between 1080ti and 1080 is almost as big as 980ti is to 780ti.

On the topic of Volta, if 20 series prove to be Volta, I'll be eagerly waiting on the 2080ti or equivalent, 4k/144hz gaming can't get here fast enough.
 
I can see Nvidia launching it early but hopefully its not rushed to collect from fanboys. They probably have data on how many people are upgrading from existing nvidia base. I think the way it works is we all know. People with exisiting 1080, 1070, 1080ti will upgrade no matter what. Thats just how it is. Even if it is only 20% increase in performance. They will be sold, I really doubt Volta will be the increase we saw with pascal. Pascal was basically running on miracle clocks. Nvidia had the secret sauce. Is volta going to run at 2.5ghz? if so then yea we can expect some decent increase. But I don't see nvidia getting those clocks on similar node, but who knows they have the money to do it. But I am skeptical. They would have to totally nail it on the architecture piece to get the same perfomance bump as pascal. Pascal saw great benefit form new node. Maxwell loved high clocks and it was maxwell on steroids.
 
Volta is going to shock all of you #waitonvega people. AMD is so behind the 8-ball it isn't funny. R&D does matter as one of the folks you AMD guys just hate to read. I don't relish AMD being behind, but I certainly can acknowledge it. NV is ahead by a year in the high end market and it looks like they are going to extend that lead. GCN is way long in the tooth and it's getting longer. This doesn't bode well for competition, but I can certainly see what's happening and what will happen. #waitforvega crap is getting old.
 
Volta is going to shock all of you #waitonvega people. AMD is so behind the 8-ball it isn't funny. R&D does matter as one of the folks you AMD guys just hate to read. I don't relish AMD being behind, but I certainly can acknowledge it. NV is ahead by a year in the high end market and it looks like they are going to extend that lead. GCN is way long in the tooth and it's getting longer. This doesn't bode well for competition, but I can certainly see what's happening and what will happen. #waitforvega crap is getting old.

Understand your critcism of AMD. But this Volta is going to shock you is only true when volta does shock me. Dont count your eggs before they hatch. If Volta doesn't deliver the same performance as Pascal did due to its clock speeds then what? Are you going to still buy it because its nvidia? I don't doubt nvidia's expertise but they have dropped the ball before.
 
Understand your critcism of AMD. But this Volta is going to shock you is only true when volta does shock me. Dont count your eggs before they hatch. If Volta doesn't deliver the same performance as Pascal did due to its clock speeds then what? Are you going to still buy it because its nvidia? I don't doubt nvidia's expertise but they have dropped the ball before.

Current history says they don't drop the ball. They are executing very well and I expect them to again. With that said I don't buy failures. I buy performance no matter who sells it. With that said, if Volta (GV104 chip first out) runs at 1200mhz and is faster in games than my current Titan X(16), I will definitely buy it. Performance is the name of the game.
 
GV104, as a GP104 replacement is said to be 13.5Tflops. To compare a GP104 is 9.2Tflops with boost and GP102 around 11.4-11.8Tflops. Add some 14-16Ghz GDDR5X/GDDR6 on a 256bit bus and you are set.
 
It really isn't. Almost every review uses a 980 Ti reference model at stock clocks so the 1080 looks a lot better in that comparison. The overclocked 980 Ti comes close to a stock 1080 and an overclocked 1080 pulls ahead something like 15-20% which IMO isn't worth an upgrade. I don't know about you but I generally want something that is at least 40-50% faster than my current card and only the 1080 Ti fits the bill.

In any case not worth fighting about as this thread is about Volta.

People really need to stop comparing the 980ti to the 1080 unless the discussion is about whether to upgrade or which to buy from a consumer perspective (and even then it must be acknowledged they are different GPU tiers).
The 1080 should be compared to the 980 if talking about performance gains over previous generation.

The 980ti and 1080 are massively different in terms of core architecture, and one reason why the 980ti still does well even today.
Nvidia architecture is very powerful when it is very close to the full capability; meaning 6 GPC (with Raster Engines) and a full/near full compliment of SM in each along with the Polymorph Engine and all that entails or associated.
980ti: 6 GPC, full 4 SM in 4 of those GPC and 1 SM missing in 2 of those GPC (for 22 SM total), 96 ROPs, 2816 CUDA cores
1080: 4 GPC, full SM (20 total) in those 4 GPC, 64 ROPs, 2560 CUDA cores

Pretty clear there is absolutely no reference to comparing these two products as they are totally different spec technically.
Really should compare the 980ti to the 1080ti where the architecture is more aligned and shows how it expanded with the node shrink.
Cheers
 
Last edited:
I really should have specified 1080p...

So I guess it's clear 1080 Ti prices are safe now. With the RX 580 launch I can't see Nvidia feeling threatened by that. Probably also means 20 series is just a refresh. Dammit AMD.
 
I thought Volta was suppose to be on a new node. But if it's still on same but little improved process then it doesn't make any sense nvidia will redo their plans to put Volta on same node as pascal. I am thinking more and more that we might see Improved version of pascal first before we see Volta.
 
I thought Volta was suppose to be on a new node. But if it's still on same but little improved process then it doesn't make any sense nvidia will redo their plans to put Volta on same node as pascal. I am thinking more and more that we might see Improved version of pascal first before we see Volta.

Kepler->Maxwell didn't make any sense either, right? ;)
 
I thought Volta was suppose to be on a new node. But if it's still on same but little improved process then it doesn't make any sense nvidia will redo their plans to put Volta on same node as pascal. I am thinking more and more that we might see Improved version of pascal first before we see Volta.
It was never planned as anything else, just rumours and speculation using 2018 as a date and then linking it to nodes smaller than 16nm.
Volta brings in the next step changes to Pascal with further improved efficiency/other architecture improvements like we saw from Maxwell to Pascal (was 1st stage of a true Unified Memory and mixed precision along with early NVLink - all these take another jump), and again with another jump in performance required for the supercomputer projects kicking off this year and next.
Its performance/efficiency is meant to be pretty impressive for same node, just like Shintai mentions regarding Maxwell.

Pascal can be seen as risk management for what Volta is meant to be by splitting some of the emerging tech into two parts and further evolving them along with introducing other aspects that will be introduced with 2nd stage (Volta).
Pascal also provided some great sales but importantly is the perfect plan cycle-test for launching Volta in all of its diverse hardware and software-platform solutions, where there are several massive commitments with Volta in different technology segments.

Cheers
 
Last edited:
Now I'm conflicted. Currently have a 1070 and almost gonna jump on that 1080Ti. But if Volta comes around Q4 2017 then that makes the decision a lot more difficult.

:cry:

Though it'll probably be at least a year until consumer Volta comes out. Decisions decisions...
 
Now I'm conflicted. Currently have a 1070 and almost gonna jump on that 1080Ti. But if Volta comes around Q4 2017 then that makes the decision a lot more difficult.

:cry:

Though it'll probably be at least a year until consumer Volta comes out. Decisions decisions...

if you are already willing to jump from 1070 to 1080 ti means you are ready to upgarde within same generation of cards. Get the 1080ti and upgrade to volta. There is a reason Nvidia will release this, like I said buyers such as you will continue to upgrade. They are like apple of graphics cards. No matter what people will upgrade lol.
 
if you are already willing to jump from 1070 to 1080 ti means you are ready to upgarde within same generation of cards. Get the 1080ti and upgrade to volta. There is a reason Nvidia will release this, like I said buyers such as you will continue to upgrade. They are like apple of graphics cards. No matter what people will upgrade lol.

Well at least nVidia gives people a reason to upgrade, wish I could say the same thing for AMD.
 
Back
Top