Nvidia Ampere Discussion [2020-05-14]

Which goal ? 20 years old Pixar animation IQ ? Maybe. Today's ? Hell no.

 

Would be nice to get something in between.

I remember seeing the first shrek (yea its a dreamworks movie so sue me) and when they go to fight the dragon I thought wow this would be amazing for a rpg



I mean this is a 19 year old movie and I'd still love to enter a game that looked this good

and then Shrek 2 in 2004 blows that out of the water.

 

Board size is probably still the same. The cooler is what has been bulking up generation over generation

 

This cooler is actually smaller than the aftermarket coolers on currently available 2080 Ti cards from Asus and Gigabyte.
Also noteworthy that this cooler is different from what was leaked back in June - the old leak showed a 2-slot design more or less the same size as that of 2080 FE - which is likely what 3080 will get.
So I kinda don't understand the fuss about it.

 

If it's the size needed for cool&quiet, I'm ok with it.

 

I'll most likely undervolt my next gpu as low as it'll go while being able to maintain the reference boost/game clock without fluctuating.

 

After reading this, I wondered if Pixar uses Nvidia GPUs, so I googled it.

https://nvidianews.nvidia.com/news/...logy-for-accelerating-feature-film-production

 

The 2-slotter had RTX 3080 markings on it too, so yes.

 

Of course, even AMD have a dedicated product for their type of work and game engines even became viable tools.
That happened before programmable shaders on commercial GPUs and before the people from SGI made the OpenGL abstraction to make the farmwork for RenderMan viable on desktops.
https://cseweb.ucsd.edu/~ravir/6160-fall04/papers/p425-peercy.pdf

 

That much, I think, is obvious. Since Kepler they've had plenty of headroom, so what other reason is there to blow past the efficiency curve so totally with Ampere? My only question would be: What does Nvidia see as necessary? Through the lens of 10+ years ago, resorting to this kind of cooling would point to an FX-level disaster. Today perhaps, Nvidia has it in their head that it would be unacceptable for AMD to be within, say, 20% of their top card. Either way, things are going to get interesting!

 

Again, there's nothing particularly strange in such cooling. 2080Ti had bigger coolers and we're not in a water cooled as a reference FuryX situation obviously.
What it points to so far is to NV's desire to have as much headroom as possible in a reference design. Why is an unknown for now.

 

goddamit

 

There will always be some AIB that slaps a monstrous cooler on a card. Totally irrelevant.

When you resort to esoteric cooling, it's because you're consuming headroom. It's not like NV isn't going add the extra manufacturing cost out of the goodness of their hearts.

 

Why is it irrelevant? Have you noticed what happened with 20 series reference coolers and FE clocks two years ago?

And for the third time - what's so "esoteric" in the leaked photos? The fans placement?

 

It's not clear to me that the heatsink is actually larger, or more complex. It's just a unique design. It's still two fan. I'm really curious to see how it works out.

 

Somewhere between 3 slots and gigantic finstacks.

 

Ah! And Hydrox cookies actually came first, in 1908, four years before the Oreo.

 

For AMD to come close to Ampere with RDNA2, NV probably has to do worse then Turing.

 

There's also the part that they pursued sheer finstack area so aggressively that one of the fans blows air in the wrong direction, sacrificing the thermals of every other component in the system. That's was pretty esoteric before blowing the thing up to consume three slots.

 

True that, those Samsung process-related horror stories might be true after all.