NVIDIA ‘Ampere’ 8nm Graphics Cards

[Satchfanuk]

When are we likely to hear about new Nv cards with full specs etc, anyone think GTC or is this likely anyway?

Jensen might mention GTX/RTX consumer cards but I'm expecting any announcements to be more focused on AI/Datacentre type products.

Any mention of gaming cards will purely be to build hype and stop anyone switching sides to AMD, assuming the red team have anything to be excited about at that point.

 

[Jacky60]

Have you checked their financial news? AMD Is doing very well, despite choosing to not bother to go after the top end NVIDIA card for a few years.

In 2009, they had $5.28 Billion in debt. Today they are less than 500 million in debt. Lots of this has been fueled by console success and now the impressive Zen CPU sales. They've also massively increased their R&D.

I think AMD are well placed to overtake nvidia within the next few years, though I think it will take another two generations of GPU before they can exceed nvidia's top end offering consistently. R&D takes time, though AMD now has significantly more than before, fun things will happen

Oh I know they're doing very well, I regret not buying shares when they were down around $2-5 but such is the way of the World. I just wish they'd offer up a bit of the CPU magic in the GPU space pronto!

 

[nashathedog]

Jensen might mention GTX/RTX consumer cards but I'm expecting any announcements to be more focused on AI/Datacentre type products.

Any mention of gaming cards will purely be to build hype and stop anyone switching sides to AMD, assuming the red team have anything to be excited about at that point.

Lets hope they do ie AMD. I was hoping Intel entering the fray would shake things up but if the rumours that they're only starting off with low to mid range cards are true that's that out the window for now. IMO Intel made a big mistake hiring Mr Koduri, after how he handled RTG I thought them mad for doing so, I get that they were desperate due to having the position open for many months but they should have promoted whoever was standing in.

 

[LambChop]

Agreed. It won't be this coming generation - 'Big Navi' or whatever - but the next. AMD is getting its act together. AMD has a lock on console GPUs and is making money hand over fist on Ryzen and that translates to investment and that should translate to performance.

But then for AMD to catch up wither this gen or next, it would mean nVidia stands still for that to happen. thats not going to happen. In order to catch up they have to release something faster every time. So to do any catching up they have to go faster than the 2080Ti from their next release. I even have my doubts about that.

Roughly 1 gpu in 4 is AMD. In integrated, with all the console chips, AMD are still behind, miles behind, can you imagine if they lost that contract. ?

 

[Quartz]

it would mean nVidia stands still for that to happen.

Nvidia's rate of performance progress has slowed significantly. Not as badly as Intel. But getting there and their price / performance has gone down.

 

[KillBoY_UK]

Big Red is getting Ampere

https://hexus.net/tech/news/systems/139412-big-red-200-get-nvidia-ampere-upgrade-summer/

 

[HRL]

Nvidia's rate of performance progress has slowed significantly. Not as badly as Intel. But getting there and their price / performance has gone down.

Only because they didn't really need to try this gen. I'm sure they could have released something a little faster than the 2080Ti if they'd needed to, even if it was just higher clock speeds and VRAM.

 

[BigBANGtheory]

Big Red is getting Ampere

https://hexus.net/tech/news/systems/139412-big-red-200-get-nvidia-ampere-upgrade-summer/

Considering they used the Tesla part before this sounds more of a data centre product than a consumer gaming GPU.

 

[Quartz]

Only because they didn't really need to try this gen.

Intel fell into that trap...

 

[Kaapstad]

Only because they didn't really need to try this gen. I'm sure they could have released something a little faster than the 2080Ti if they'd needed to, even if it was just higher clock speeds and VRAM.

No one has seen a TU100 chip as NVidia were not under any pressure to release one that big.

 

[HRL]

No one has seen a TU100 chip as NVidia were not under any pressure to release one that big.

Exactly. I don't think NVidia will fall into the same trap as Intel have in the CPU market, but they certainly didn't feel forced into showing their hand this gen. I'm looking forward to Ampere, let's just hope it's a big step up in performance this time around or else my kids will not be getting a hand me down 2080Ti.

 

[Grim5]

Considering they used the Tesla part before this sounds more of a data centre product than a consumer gaming GPU.

They are buying the HPC Ampere GPU lineup

 

[BigBANGtheory]

They are buying the HPC Ampere GPU lineup

Im not really sure what that means though... As I recall (granted my memory gets a bit flakey at times) Volta came out replacing Pascal as a datacentre Tesla product with much excitement and performance gains then 12months later we got Turing in the regular consumer space and it was pants performance with a big price hike on top of Pascal which wasn't exactly cheap to begin with.

 

[Grim5]

Im not really sure what that means though... As I recall (granted my memory gets a bit flakey at times) Volta came out replacing Pascal as a datacentre Tesla product with much excitement and performance gains then 12months later we got Turing in the regular consumer space and it was pants performance with a big price hike on top of Pascal which wasn't exactly cheap to begin with.

HPC is a datacentre/workstation GPU, not for gaming. There were no HPC cards released with Turing architecture, the latest models are based on Volta. However Volta and Turing architecture is very similar when compared side by side so Volta can run games just fine.

When comparing Volta to Turing, they are not that different in gaming performance

So if someone says that Ampere is 75% faster than Volta, that makes me think Amepre is 50% faster than Turing.

 

[3t3P]

I want 10,000 CUDA cores

 

[HRL]

Unlikely as it is, if the 3080Ti is 50% faster then I’ll definitely buy one.

 

[BigBANGtheory]

When comparing Volta to Turing, they are not that different in gaming performance

but Volta and the Titan V is not optimised for graphics, even Nvidia state its optimised for machine learning and scientific work, so comparisons for graphic/gaming work loads aren't really helpful if anything it taught us in the normal retail market anything HPC and datacentre related bares no real resemblance to gaming GPUs. Volta was massively more powerful than Pascal in HPC but Turing was meh.

 

[Sargatanas2511]

Unlikely as it is, if the 3080Ti is 50% faster then I’ll definitely buy one.

Even if it's more expensive than the 2080Ti?

 

[KillBoY_UK]

Considering they used the Tesla part before this sounds more of a data centre product than a consumer gaming GPU.

True but even the high end tesla parts are not a million miles away from the graphics GPU's, This deployment shows that Ampere is not far away. Also much more tangible than the "big navi" gpu/rumour news we have had for months/over a year.

 

[GAC]

True but even the high end tesla parts are not a million miles away from the graphics GPU's, This deployment shows that Ampere is not far away. Also much more tangible than the "big navi" gpu/rumour news we have had for months/over a year.

the thing is what SPECIFIC program was it using to get this big jump, as lets be blunt nvidia is not against loading tests to show how great their next gen stuff is going to be. just because its faster by a big margin than tesla in this example dont expect it to translate to gaming.