When realistically should we see the next Nvidia video card that is faster than anything out there?

[Guru]

Its not even volta and this actually makes sense. Volta is 100% designed from the ground up for computing and AI and supercomputers. It has additional layers of tech that is useless for gaming, not to mention its close to 900 in die size, essentially almost doubling the 1080ti is size. There is no way they can downscale that and make it commercially viable.

So it only makes sense that they would create a new architecture, AMPERA for the gaming market. If they are going that route the biggest change they could make is utilize DX12/Vulkan more and lose performance in DX11. So while they are likely to do this, I don't think they will be rushing the design, especially since we don't have any clear path that developers would take, we could literally see less than 5 DX12/Vulkan games in 2018 or it might be a breakout year and have a third of games use low level API and use it good!

So anything before Q2 2018 is just wishful thinking. There is no way they can have a new generation ready, that is significantly better than the 1000 series and be positioned in alignment with the market so fast.

 

[Mopetar]

I think we're all aware of how NVIDIA launches new series. I wouldn't be surprised if we saw a late 2018 2080ti release and that's the card that those of us with 1080tis are impatiently waiting for.

2080 will be enough of a bump to get people to buy anyway. Same story as the 980 Ti and the 1080.

I’ll just be happy to see the 2070 follow the same pattern though. I wish that it would be a $300 card again, but that’s pretty unlikely.

 

[tential]

Its not even volta and this actually makes sense. Volta is 100% designed from the ground up for computing and AI and supercomputers. It has additional layers of tech that is useless for gaming, not to mention its close to 900 in die size, essentially almost doubling the 1080ti is size. There is no way they can downscale that and make it commercially viable.

So it only makes sense that they would create a new architecture, AMPERA for the gaming market. If they are going that route the biggest change they could make is utilize DX12/Vulkan more and lose performance in DX11. So while they are likely to do this, I don't think they will be rushing the design, especially since we don't have any clear path that developers would take, we could literally see less than 5 DX12/Vulkan games in 2018 or it might be a breakout year and have a third of games use low level API and use it good!

So anything before Q2 2018 is just wishful thinking. There is no way they can have a new generation ready, that is significantly better than the 1000 series and be positioned in alignment with the market so fast.

The only thing you're doing is nitpicking between whether the nextgen gaming arch will be called Volta, or now that you have new info on a brand new name, the new name.
Look, we're here to discuss the latest gaming Nvidia architecture.

You aren't disagreeing with anyone or saying anything that anyone else isn't saying.... We're all saying between April-July.

You and many others are just continuing down this Ampera/Volta debate, go make a thread for those who care about the NAME of the next gaming architecture. Otherwise, can we just stick to the topic?

 

[Cloudfire777]

The only thing you're doing is nitpicking between whether the nextgen gaming arch will be called Volta, or now that you have new info on a brand new name, the new name.
Look, we're here to discuss the latest gaming Nvidia architecture.

You aren't disagreeing with anyone or saying anything that anyone else isn't saying.... We're all saying between April-July.

You and many others are just continuing down this Ampera/Volta debate, go make a thread for those who care about the NAME of the next gaming architecture. Otherwise, can we just stick to the topic?

We care, because Volta might just be about Tesla and computing cards while the Ampere architecture is what 99.999999% of anandtech forum users will be interested in.

Question is, why a different name and what is the difference between them

 

[Eric1987]

Why would they go to 2080 when it would be better as 1180?

 

[punkyDD]

We care, because Volta might just be about Tesla and computing cards while the Ampere architecture is what 99.999999% of anandtech forum users will be interested in.

Question is, why a different name and what is the difference between them

Different name == different uarch. And we don't know that Ampere will necessarily come after Volta.

 

[Jason Ives]

Why would they go to 2080 when it would be better as 1180?

It's your opinion that it would be better I'm assuming. But 2080 is better for marketing terms. It just sounds resolutely next gen no matter if it is or isn't.

 

[Bier667]

Nvidia calls Pascal the 10 series: 10 for uArch and then the Performance Moniker aka 50/60/80 and so on.

My bet is that Nvidia changes the Numbers and will perhaps go with "something" new aka Intels i3/5/7 or AMDs R series model.


And i hope Nvidia overhauls the Driver-Frontend and removes alot of the lecagy Settings: i mean the Frontend is from what? 2004? right?

 

[Mondozei]

Everyone seems to be already assuming that Volta can't be cut-down and stripped from the Tensor cores etc. That's foolish.

NV CEO is known for being a blowhard. Saying "designed from ground-up for AI" is as much a PR statement as anything else.

A cut-down Volta w/e AI bits is still likely for consumers. Ampere goes into fast-growing AI/DL biz, at least initially. Remember: NV Is putting out new uarch EVERY year for AI/DL these days. Stagnating on Volta Is NOT an option in fast-moving & competitive markets.

AI/DL gets preference from now on. Gaming gets same uarch but +1 year later. Volta 2017. Ampére 2018. Etc.

 

[Qwertilot]

If they do Volta + 1 for compute at precisely the same time they release Volta for consumers that gap is likely to get longer than a year....
(Mind you, I do have to agree that this is very likely to happen.).

 

[Batmeat]

gddr6 will be in full production in February , and be used for the gtx2080 in May.
I think the Titan V will have HBM 2 memory 4 months later in September.

Has this been confirmed now? I thought they were going to use HBM2.

 

[tential]

https://wccftech.com/google-poaching-deep-learning-engineering-talent-from-nvidia-corporation/

Google is offering an 8 figure package (in the range of 9 – 12 Million USD over the course of 3 years) to the deep learning engineer(s) in question to shift over from NVIDIA.

Nvidia Engineers being treated like they're pro sports players.
That whole sentence reads like it should be in pro sports not engineering.

Wow, competition is fierce this is crazy.

Just thought that was crazy how this whole engineering race is turning out to be, not sure where this would go and doesn't deserve its own thread/

 

[beginner99]

There is no reason for them to rush Volta, Q1 is not even realistic, considering Volta so far is purely a professional product with a massive die size. At best early Q2 for a limited release of the 2080/2070, then the mid range and after few months the 2080ti. I actually think we are looking at late Q2 release.

Fully agree. I bet the pulled away people from the Volta project to whatever is next latest after Vega release, probably earlier. Project management triangle. Fast (time to market), Cheap (project costs), Good (the product). You can only choose 2 of them. Since time to market matters least in current scenario NV can pull people of the project while delaying the project. However later project will benefit from this which of course is very bad for AMD.