NVIDIA ‘Ampere’ 8nm Graphics Cards

[Shac]

@TNA
@Panos was last seen tuning in his new memory. I hope he has not fried anything!

 

[aoaaron]

Is there any chance or roadmap for DLSS to be applied to old games?

The potential for DLSS for me is clearly in a gaming laptop.

Imagine being able to pack an RTX 3060 and know you can DLSS a 540p image to 1080p and it look more than good enough for portable gameplay. It'll turn portable gaming laptops into very usable machines for most games.

I think DLSS is amazing but if its not available in most games we play.. its just a bit of a joke.

 

[TNA]

@TNA
@Panos was last seen tuning in his new memory. I hope he has not fried anything!

Hahahaha.

Maybe he upped the volts a little too much? Lol

 

[TNA]

Man, that is not enough RAM ffs. They need to be going 12 for 3070 and 3080, 16 for 3080Ti and 24 for Titan.

 

[Poneros]

Man, that is not enough RAM ffs. They need to be going 12 for 3070 and 3080, 16 for 3080Ti and 24 for Titan.

Oh I'm sure Titan's gonna pack lots of vram for the modest sum of $2499 again.

I guess at least in that respect vram might still be an ace for AMD if they do decide to go with HBM. Otherwise, yeah, it's gonna be sad all around.

 

[Ian Evey]

Nvidia have to have high prices so they can plough all those profits into thier car tech crap.

Before long i bet they wont even make competative cards for joe public.

 

[bJN]

Man, that is not enough RAM ffs. They need to be going 12 for 3070 and 3080, 16 for 3080Ti and 24 for Titan.

They don't. Better compression and Nvidia starting to leverage other means of paging data means they don't need to add more and more VRAM each generation. We're starting to be at a point where off-card storage has the bandwidth available to be a usable resource for GPU loading.

 

[NZXT30]

Nvidia have to have high prices so they can plough all those profits into thier car tech crap.

Before long i bet they wont even make competative cards for joe public.

They cant do that AMD is not THAT far behind. Unless they get together into a nice and cozy price fixing scheme.

 

[TNA]

They don't. Better compression and Nvidia starting to leverage other means of paging data means they don't need to add more and more VRAM each generation. We're starting to be at a point where off-card storage has the bandwidth available to be a usable resource for GPU loading.

Yea, I ain’t convinced about that just yet. Will believe it once I see it. A bit like AMD and their 4GB HBM Fury...

 

[bJN]

Yea, I ain’t convinced about that just yet. Will believe it once I see it. A bit like AMD and their 4GB HBM Fury...

A lot, at least from viewpoint, has changed in the storage space in 4-5 years. Better technology becoming quickly adopted by the masses (relatively speaking...) opens up a lot of potential for shared resources.

 

[TNA]

A lot, at least from viewpoint, has changed in the storage space in 4-5 years. Better technology becoming quickly adopted by the masses (relatively speaking...) opens up a lot of potential for shared resources.

Sure. But I still say we’ll see.

Unless it works perfectly I would just rather the extra 4GB of ram, not like that would add a huge amount to the cost of the card in the end.

 

[Jimmy Weirdarms]

Are we still on for >2080ti performance on a 3070 or not?

Nobody knows. At this point we're not 'on' for anything at all, the only hint was apparently Jensen saying that there was significant overlap in architecture between the data centre and consumer chips, though not in set-up.

That's it.

 

[subbytna]

Yeah, Panos convincingly disappears

If I could be asked I would dig up all his posts. Lol

Get Posting boyo

 

[subbytna]

Nvidia have to have high prices so they can plough all those profits into thier car tech crap.

Before long i bet they wont even make competative cards for joe public.

Can you imagine the Nvidia tax there will be in cars soon. Adding another 3k to the cost of a car just for the bovine excrement Nvidia are trying to force on us

 

[CAT-THE-FIFTH]

https://www.anandtech.com/show/15801/nvidia-announces-ampere-architecture-and-a100-products

The GA100 has 8192 shaders,but the A100 has 6192 shaders and a 400W TDP,which is an increase over the previous 12NM V100 which was 5120 shaders and a 300W~350W TDP. The GPU is huge at 826MM2. Poor yields a problem??

 

[Chuk_Chuk]

Man, that is not enough RAM ffs. They need to be going 12 for 3070 and 3080, 16 for 3080Ti and 24 for Titan.

Good spot, with the new consoles sporting so much RAM these may not age well if developers decide to puish things.

 

[southy1978]

Can you imagine the Nvidia tax there will be in cars soon. Adding another 3k to the cost of a car just for the bovine excrement Nvidia are trying to force on us

It already is in cars right now. The latest Mercedes A-Class has nvidia powering the dual panel display and augmented reality.

https://blogs.nvidia.com/blog/2018/02/02/mercedes-benz-a-class-mbux-ai-graphics/

 

[shankly1985]

Proud to have @AMDServer #EPYC “Rome” CPUs in the new DGX servers. Congrats @Nvidia on the new A100 launch and thanks for selecting @amd as your CPU partner.


https://twitter.com/LisaSu/status/1260980245999628289?s=19

 

[CAT-THE-FIFTH]

This chap was correct in some of his GA100 predictions earlier in the year:
https://twitter.com/kopite7kimi/status/1260897297065365504

A good news and a bad news. GA104 has been taped out. GA103 may be cancelled.

 

[Poneros]

They don't. Better compression and Nvidia starting to leverage other means of paging data means they don't need to add more and more VRAM each generation. We're starting to be at a point where off-card storage has the bandwidth available to be a usable resource for GPU loading.

Compression helps with bandwidth, but it can't substitute for vram amount requirements. As for other paging means, maybe ask Vega owners how (HBCC) that panned out.

As most of us find out, there's simply no substitute for bigger... vram amounts.