• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Only AMD has true Async Compute - Doom Devs

Jono8 said:
What i mean is the 1070 = 980ti = Fury x in relative gaming performance so the 1070 is showing that amd dont have the dx12 advantage they had against maxwell.
Click to expand...

The Fury X is slower overall than the GTX 1070 despite the tflops number. The gains we are seeing in DX12 are what most people think we will continue to see with more advanced architectures like Vega (even bigger gains in DX12 considering it is built specifically to take advantage of DX12 features), so AMD are still going to hold on to their edge. Also most sites like Guru3D aren't using msaa in AotS which causes a considerable drop in performance on nvidia's architecture. Pascal does gain some performance improvements thanks to its added hardware scheduler, and nvidia engineers are awfully good at developing software, so it could very well be that Pascal ends up being the overall champion this time around. Vega cannot come here soon enough! ;)
 
GoogalyMoogaly said:
And some people were saying that DX12 and Vulkan are just Mantle with the name changed...

Also, "Overclockers Dream"...
Click to expand...

DX12 maybe not but I 100% feel AMD helped Microsoft with DX12 same has they give Mantle to help build Vulkan.

Vulkan is closer to Mantle than mantle is to DX12 that's for sure. It still prove the point people was saying DX12 was in the making long before Mantle ETC ETC!
 
AlamoX said:
i think this was clear from some time now, nvidia explained that the improvement they made in pascal is related to context switching and better pre-emption, they never claimed hardware implementation.
the war on async will happen first at the studios of game developers, AMD will try to push as many games into async, and nvidia will try to keep devs on dx11 as much as possible.
Click to expand...

Preemption is a hardware implementation. Otherwise can you explain how to do pixel-level preemption in software without using magic.
 
TheRealDeal said:
They were using the high preset. Here is the Extreme preset.

http://www.tomshardware.com/reviews/nvidia-geforce-gtx-1070-8gb-pascal-performance,4585-3.html

Fury X is now only 10% behind the 1080 and Fury Pro is now faster than the 1070. It does look like Nvidia have moved performance on around 10% in DX12. I wonder if there was any difference in drivers used for NV cards.
Click to expand...


The benchmark is pointless because the developerschool haven't implemented an async path for pascal yet. You need to wait until oxide have played with the 1080 to see what the real results are like.
 
D.P. said:
The benchmark is pointless because the developerschool haven't implemented an async path for pascal yet. You need to wait until oxide have played with the 1080 to see what the real results are like.
Click to expand...

Nvidia said that about Maxwell, the game has been around for an age, how much longer are we going to have to wait?
 
humbug said:
It was 10x Maxwell at one point in one of their slides :p
Click to expand...

Which was made perfectly clear to be the case specifically in deep learning, and the number was substantiated with real world benchmarks proving it was accurate.

So where exactly are you going with this?


All the performance numbers Nvidia announced have been fairly accurate, if anything they have been under reporting slightly. They said things like 1080 is 80% faster than 980 when it is commonly 75-82% faster in a majority of games.
 
humbug said:
Nvidia said that about Maxwell, the game has been around for an age, how much longer are we going to have to wait?
Click to expand...

Ask the developers. The 1080 was only just released so I doubt oxide has had time to do much yet, especially when getting good paid by AMD.
 
You must log in or register to reply here.
Back
Top Bottom