NVIDIA Pascal Async compute support: CONFIRMED

[i_build_nanosuits]

So, i was watching the reviews for the GTX 1060 that came out yesterday...and in their GTX 1060 video review, Digital Foundry (a well know and very reputable source for graphics cards and other PC hardware testing and reviews) clearly showed that the new GTX 1060 (and consequently all the other GPU's in the Pascal architecture familly) does INDEED support asynchronous compute under DirectX12.

 

They did testing with the new DX12 3d Mark TimeSpy benchmark, and clearly the GTX 1060 does gain a noticeable amount of frames per second output from enabling the Async compute feature...in fact the gains there are very close to those seen on the RX480.

 

Here, i put the link to the video...you should watch it in it's entirety but the interesting part about nVidia and Async compute start at 3:30

 

 

@3:30 ''one thing that is clear, is the nVidia's Pacal architecture does indeed support that all important Asynchronous compute feature''

[2FA]

*Concurrent Async Compute

[HKZeroFive]

Most people know this. Pascal supports async compute but doesn't have async shaders which is what gives AMD that extra advantage over NVIDIA (hardware vs software).

 

Maxwell, on the other hand, is a different story...

[Lord_Karango17]

Didnt Maxwell also support a software implementation of async compute as well? I though amd had the advantage with the rx480 due to having built in hadrware for async compute.

[2FA]

Just now, Lord_Karango17 said:

Didnt Maxwell also support a software implementation of async compute as well? I though amd had the advantage with the rx480 due to having built in hadrware for async compute.

Maxwell was never enabled in the driver though. The driver will block every attempt at async compute on the 900 series.

[Ertyrty]

Eh. And I was happy for a moment. 

[Morgan MLGman]

@i_build_nanosuits

 

Then why does it still lose FPS when transitioning to DX12?
 

From a GTX 1060 review and benchmarks:

ROTTR DX11:

 

 

ROTTR DX12:
 

 

Hitman DX11:

 

 

Hitman DX12:

 

[Dabombinable]

Just now, Morgan MLGman said:

@i_build_nanosuits

 

Then why does it still lose FPS when transitioning to DX12?
 

ROTTR DX11:

 

 

ROTTR DX12:
 

 

Hitman DX11:

 

 

Hitman DX12:

 

Why do people post the snips but not the link to the website? Unless they want people to think that they are cherry picking?

[Morgan MLGman]

Just now, Dabombinable said:

Why do people post the snips but not the link to the website? Unless they want people to think that they are cherry picking?

Because the article is in Polish and you're not from here probably? They benched only two games DX11->DX12.

Here: http://pclab.pl/art70607.html

[Dabombinable]

Just now, Morgan MLGman said:

Because the article is in Polish and you're not from here probably?

Here: http://pclab.pl/art70607.html

People on this forum know how to use Google Translate.

[intelCore]

Just now, Dabombinable said:

People on this forum know how to use Google Translate.

Some people*

[Morgan MLGman]

Just now, Dabombinable said:

People on this forum know how to use Google Translate.

Not even once have I seen it translate something properly

 

besides, what cherrypicking? A thing like losing FPS when going from DX11 to DX12 in a brand-new 2016 GPU should NOT happen AT ALL.

[Tomsen]

DX12 3d Mark TimeSpy benchmark have receiving lots of shit for not been a DX12 benchmark, not use async properly, and in whole been optimized towards nvidia hardware.

I would wait for more information, before stating anything.

 

What is still clear;

Neither maxwell nor pascal support async in hardware.

They can impose all their driver-trikery to simulate "support", but that require extensive support from Nvidia on the individual title.

[i_build_nanosuits]

6 minutes ago, Morgan MLGman said:

@i_build_nanosuits

 

Then why does it still lose FPS when transitioning to DX12?

 

i return the question: why does the other reputable outlets show completely different results then your obscure source?

[Morgan MLGman]

Just now, i_build_nanosuits said:

i return the question: why does the other outlets show completely different results then your obscure source?

Link? I was curious and searched for sites that benched the same games in both DX11 and 12 to see whether it really has proper Async after reading your topic, this was the first that I found, it's one of the three most well-known Polish tech-sites, but maybe they screwed the benchmarks up somehow

[patrickjp93]

7 minutes ago, Morgan MLGman said:

@i_build_nanosuits

 

Then why does it still lose FPS when transitioning to DX12?
 

From a GTX 1060 review and benchmarks:

-SNIP-

Because they all use far too many queues for Maxwell and Pascal to handle. There isn't a tuned code path for Nvidia in DX 12.

[Morgan MLGman]

Just now, patrickjp93 said:

Because they all use far too many queues for Maxwell and Pascal to handle. There isn't a tuned code path for Nvidia in DX 12.

But why is that? I mean, I want to understand why Nvidia doesn't fix that issue and what is the cause of the issue, care to explain? Thanks in advance

[i_build_nanosuits]

i should have known the AMD fanbabies would come at me with pitchforks.

That was not the goal of this post, i don't feel like arguying, the proofs are there.

Pascal DOES get improved performance from Async compute support under DX12, now deal with it.
i'm out. (this forum really sucks, it's a good thinhg for nvidia users...that DOES NOT mean it's a BAD thing for AMD fanboys...GROW UP!!)

[Dabombinable]

Just now, i_build_nanosuits said:

i should have known the AMD fanbabies would come at me with pitchforks.

That was not the goal of this post, i don't feel like arguying, the proofs are there.

Pascal DOES get improved performance from Async compute support under DX12, now deal with it.
i'm out. (this forum really sucks, it's a good thinhg for nvidia users...that DOES NOT mean it's a BAD thing for AMD fanboys...GROW UP!!)

Did you see the videos comments section?

[patrickjp93]

Just now, Morgan MLGman said:

But why is that? I mean, I want to understand why Nvidia doesn't fix that issue and what is the cause of the issue, care to explain? Thanks in advance

The issue is simple: AMD is the only one helping developers with DX 12 right now.

 

The cause is simple: Nvidia knows you don't need 32+ queues to get the main benefits of Asynchronous Compute and Shading. We've been using 1 up until now, so it logically follows we don't suddenly need 32 or 64 despite what AMD keeps pushing to make up for its own lousy drivers. Nvidia's hardware handles less, and AMD is taking advantage of it right now. Does that mean Nvidia's hardware is any less fit for DX 12? IMHO, no, because AMD is being wholly disingenuous with this push for maximum asynchronous compute and shading as if it's magic fairy dust and more is better when clearly there's an upper limit being surpassed here.

[2FA]

Just now, i_build_nanosuits said:

i should have known the AMD fanbabies would come at me with pitchforks.

That was not the goal of this post, i don't feel like arguying, the proofs are there.

Pascal DOES get improved performance from Async compute support under DX12, now deal with it.
i'm out. (this forum really sucks, it's a good thinhg for nvidia users...that DOES NOT mean it's a BAD thing for AMD fanboys...GROW UP!!)

Only reason I put my original comment is because concurrent async has a higher latency compared to native hardware based async. 

[i_build_nanosuits]

Just now, Dabombinable said:

Did you see the videos comments section?

no not really...the level of stupidity in the youtube comments section is usually beyond bearable..the amount of un-educated clueless morons that post there is usually simply too much for me to handle...i don't waste time there anymore.

 

I will say though, i will download this timespy benchmark tonight at home and i will test with my 980ti and see if i get improvements from Async compute as well (i hope so :p)

[M4st4M1nd]

3 minutes ago, i_build_nanosuits said:

i should have known the AMD fanbabies would come at me with pitchforks.

That was not the goal of this post, i don't feel like arguying, the proofs are there.

Pascal DOES get improved performance from Async compute support under DX12, now deal with it.
i'm out. (this forum really sucks, it's a good thinhg for nvidia users...that DOES NOT mean it's a BAD thing for AMD fanboys...GROW UP!!)

Well yes but we all knew it had Async support and it does get performance gains. Nobody has really disputed that. Does it benefit as much as AMD when used properly? Nope. So the same conclusion everyone has had for weeks now...

[Morgan MLGman]

Just now, patrickjp93 said:

The issue is simple: AMD is the only one helping developers with DX 12 right now.

 

The cause is simple: Nvidia knows you don't need 32+ queues to get the main benefits of Asynchronous Compute and Shading. We've been using 1 up until now, so it logically follows we don't suddenly need 32 or 64 despite what AMD keeps pushing to make up for its own lousy drivers. Nvidia's hardware handles less, and AMD is taking advantage of it right now. Does that mean Nvidia's hardware is any less fit for DX 12? IMHO, no, because AMD is being wholly disingenuous with this push for maximum asynchronous compute and shading as if it's magic fairy dust and more is better when clearly there's an upper limit being surpassed here.

Hmmmm, but considering what transitioning from DX11 to DX12 in Ashes of the Singularity, which is a DX12 showcase game benchmark, does to a 390X (that it surpassed 980Ti's DX12 score), is it really that bad that they used that many queues? If they do improve performance THAT much, what's wrong with that? Not like it's the 64x tesselation that decreases FPS by half when used ;-;

[Kobz360]

Didn't we all know Pascal had software Async support already?