Nvidia Ampere Discussion [2020-05-14]

Turing launched couple of years ago. It's reference were 2-slot 2 fans factory OCed designs.

So it will consume more than GA100 in PCIE? Because this one is 250W.

 

https://t.bilibili.com/426631230505955919?tab=2
Google translate:

 

Sigh, my next power supply is going to have an NVidia power connection tax on top of my next motherboard's SLI tax.

SIGH.

 

And Pascal before that, 2-slot + blowers.[/QUOTE]

Maybe they feel that they need the extra headroom? 250 Watts is doable in standard 2 slot/28 cm designs.

 

None of this bodes well no matter how you spin it. Nvidia isn’t going to push the limits of power and cooling just for kicks.

 

It's been a hot topic for the past few years within the M&E industry, and traditionally the quality hasn't been the same as CPU rendering. One interesting use is using unattended GPU desktops to the add to resources available to the render farm at night.

July 24, 2020
https://expertswhogetit.ca/data-cen...-is-changing-vfx-practices-in-the-me-industry

 

There is still the 3080 with a two slot reference cooler...

 

It will be curious if AIB RTX 3090 models will use 2 and 3 slot designs, similar to Turing high-end models.

 

Doesn't bode well? The 3090 is basically, or looks like a Titan replacement, it's the enthusiast product for people that want the absolute best performance and are willing to pay for that, doubt those mind the power draw or the measurements of it.
A 3080/3080Ti will evaporate what's in consoles already, a GPU like that will suffice for most high-end gamers to last an entire generation, giving them the best version of multiplat games.

I think it will be more intresting what a 3060/3070 will be like, if they perform like a 2070S/2080/2080Ti then that's a very good thing.

 

And what are the progresses made with RT (If any. And I don't mean of course just more units)

 

... which is fine (for Nvidia) if it's competing with the fastest AMD has to offer. And which isn't (for Nvidia), if AMDs fastest competes in a 2-slot/28cm design with 3090.

 

Non-Custom 2080Ti were much to power limited. Difference from 2080Super to 2080Ti FE was pretty small because of that. It makes sense for Nvidia to go for higher TDP to increase the performance of the Enthusiast tear. Especially as they know, that AMD isn't hesitating to go for 375W TDP and Intels Xe HP Presentations could also lead to products with more than 300W TDP. SLI is dying anyway, so why not build a real ultra enthisiast graphics card? Not like the Titan, which is quite limited. I like Nvidia going all in with the 3090.

The normal 3080 is much more important as an indicator of architecture efficiency. We have rumours of 320W TDP for the 3080, but at the same time the 3080 is rumoured to use the 2 slot version of the cooler. Makes no sense for me to change to the smaller cooler design for just 30 W TDP. If the 3080 really has 320 W TDP, then something went wrong during design (process, memory power..)

 

Prior Titans also fit that description but didn’t require bizarro cooling or power. So that can’t be it.

 

The Titan RTX with 24GB and a full chip has a 20W higher TDP than the 2080TI FE: https://www.nvidia.com/de-de/deep-learning-ai/products/titan-rtx/

 

Yep for a chilly 280W. That’s a far cry from the 350W+ stuff floating around for the 3090.

 

And if you look at the progression between that through Turing you may see why it can be a 3 slot top end now, without any mystical not so good reasons.

Precisely my point.

 

Someone made a graph explaining how RTX 3090 cooling works in a case

 

This graph is wrong.

Looks like you've missed the appearance of FE as well.
It will be a fun autumn.

 

But nVidia has changed the cooler from the radial design to the two axial design. They did a "radical" decision otherwise the Titan RTX would just be as load as a jet.

 

I'm having a very hard time believing that Nvidia totally fucked their power power draw advantage over AMD while doing a node shrink. 5700xt drew almost as much power as a 2080FE, and Turing was 12nm while 5700xt was 7nm. I know people are saying Samsung 8nm isn't great, but it seems very weird to me that a 2080FE would be 210W max and suddenly the 3080 is some kind of thermal monstrosity even with a node shrink that should give some advantages.