[bitsandchips]: Pascal to not have improved Async Compute over Maxwell

[AtenRa]

they're AMD partners:

So every GameWorks game released the last two years are EXTREME biased.

Thank you.

 

[USER8000]

Sadly for some the fact that many major titles are developed with the consoles as the lead platforms and there are at least 55 million PS4 and XBox One consoles out there now, Async is not going anywhere.

Conservative Rasterization and Rasterizer Ordered Views are not supported on console hardware so again that limits their use in many games.

 

[Silverforce11]

Evidence? The huge AMD logo on their site, the two year promotion tour for AMD and their Mantle API, their hardware and their software.

Nobody really believes that they are neutral. Or?

So they are AMD sponsored, that's your example of most extreme bias?

UE4 is NV sponsored with PhysX and GameWorks.

https://developer.nvidia.com/unrealengine

https://developer.nvidia.com/nvidia-gameworks-and-ue4

Epic developed Unreal Engine 4 on NVIDIA hardware, and it looks and runs best on GeForce.

-Tim Sweeney, founder, CEO and technical director of Epic Games.

So you have Epic Game's CEO saying his engine was developed on NV hardware, looks and runs best on NV's GPU. That's more of an example of extreme bias wouldn't you say?+

The difference here, Nitrous is developed on all hardware, Oxide have said they work closely with all hardware vendors. So bias they even made Async Compute toggle-able. -_-

 

[3DVagabond]

Evidence? The huge AMD logo on their site, the two year promotion tour for AMD and their Mantle API, their hardware and their software.

Nobody really believes that they are neutral. Or?

So no evidence then, just guilt by association. I didn't think you had any, but I thought I'd ask. So, when all else fails play the shill card and walk away.

And I didn't ask if they were neutral, although evidence actually points to that. Your quote was, "And Ashes is the most extreme example for a biased engine". "The most extreme", which is a big implication that you should have at least some evidence to support. Or you should retract it.

 

[sontin]

The difference here, Nitrous is developed on all hardware, Oxide have said they work closely with all hardware vendors. So bias they even made Async Compute toggle-able. -_-

Unreal Engine 4 is open. You can bring all this up over and over again but unlike certain other companies Epic doesnt care about nVidia or AMD.

So either you can show me how i can get the source code of the Nitrious engine or you accept that Oxide doesnt care about nVidia users (which is fine btw).

So no evidence then, just guilt by association. I didn't think you had any, but I thought I'd ask. So, when all else fails play the shill card and walk away.

And I didn't ask if they were neutral, although evidence actually points to that. Your quote was, "And Ashes is the most extreme example for a biased engine". "The most extreme", which is a big implication that you should have at least some evidence to support. Or you should retract it.

An engine tailored around a specific hardware API is not "the most extreme"? Since the first day of their company Oxide has promoted AMD.

 

[3DVagabond]

Unreal Engine 4 is open. You can bring all this up over and over again but unlike certain other companies Epic doesnt care about nVidia or AMD.

So either you can show me how i can get the source code of the Nitrious engine or you accept that Oxide doesnt care about nVidia users (which is fine btw).

They've stated that they send builds with source code access to nVidia (and AMD). Why should you having it be relevant?

 

[nvgpu]

https://www.unrealengine.com/ue4-on-github

Biased Nitrous engine, open source and available? Guess not, your fallacies are sunk.

 

[krumme]

...but it won't provide the finishing blow to Nvidia that AMD fans are so eagerly hoping for.

Is this important for you?

Because for me its a gfx that performs best. I anticipated Pascal had it because its important for performance but now suddenly its waste of space and whatnot. I simply cant tell that from bm on the new dx12 games.

 

[Sweepr]

So every GameWorks game released the last two years are EXTREME biased.

Thank you.

Maybe, but 'Optimized for AMD Radeon Graphics' doesn't make it look like the definitive non-biased benchmark title for DX12 like some are painting it.

 

[Silverforce11]

Unreal Engine 4 is open

Tim Sweeney, founder, CEO and technical director of Epic Games.

Epic developed Unreal Engine 4 on NVIDIA hardware, and it looks and runs best on GeForce.

On the public record admitting that UE4 is NV's favorite love-child. You won't find more bias than that, at the highest level.

Seriously, change the topic cos you're not going to convince many that UE4 isn't bias. We've all seen the benchmarks.

NV is going to need more of these biased games from UE4 in 2016 and 2017 if Pascal is gimped for Async Compute.

 

[USER8000]

Wait,so AMD sponsors games = unfair and Nvidia sponsors games= fair.

Yet,since Intel supports a higher tier of Conservative Rasterization and Rasterizer Ordered Views,it suddenly makes it more important now,but only for Nvidia and not for Intel but the 55 milllion consoles which are the lead platforms for many large budget games now,suddenly makes Async compute a pointless feature??

Not seeing the logic here,or have I missed it??

 

[Silverforce11]

Wait,so AMD sponsors games = unfair and Nvidia sponsors games= fair.

That's exactly what sontin/nvgpu is saying. They blame Oxide for being unfair due to AMD sponsorship and claim the game engine is so bias... yet we we look at the performance gap, it isn't that great. Not on the level of UE4 games like Ark or Hatred, where NV GPUs are 25-50% faster.

There's also nothing unfair about AMD sponsorship, because they hide nothing behind black boxes. NV was given Hitman access way before release, they even had "Game Ready" drivers for the beta. Same for AOTS, way in alpha, NV released "Game Ready" drivers for it. If they were not given access, you can accuse AMD sponsorship of being unfair, but it's not the case.

 

[NTMBK]

Unreal Engine 4 is open. You can bring all this up over and over again but unlike certain other companies Epic doesnt care about nVidia or AMD.

So either you can show me how i can get the source code of the Nitrious engine or you accept that Oxide doesnt care about nVidia users (which is fine btw).



An engine tailored around a specific hardware API is not "the most extreme"? Since the first day of their company Oxide has promoted AMD.

Don't make me trawl YouTube to find the number of Nvidia keynotes that Sweeney has appeared at. They are very close.

 

[krumme]

Maxwell is a damn fine arch. So therefore its just a shame and disappointment Pascal doesnt support one - if not the - most important features one could hope for as an improvement.

For my part there is not more to it. Lets hope the they can fix it or have some workaround and that the bm will be okey anyway. Looking forward to bf 5 as this starwars playing is not my style

 

[USER8000]

Don't make me trawl YouTube to find the number of Nvidia keynotes that Sweeney has appeared at. They are very close.

Yep,he even said UE4 would only run on Kepler when it was first released:

http://www.tomshardware.co.uk/Kepler-Nvidia-Tim-Sweeney-Unreal-Engine-UE4,news-38514.html

 

[Cookie Monster]

Im kind of confused. Since when did DX12 performance hinge on async compute?

Does Maxwell do poorly in DX12 games? I haven't seen anything like that unless I missed something? feel free to direct me these benchmarks..

As long as both IHVs can produce high playable fps in future DX12 games regardless how it gets it done (there is simply no one "right" way to do things in any aspects of technology), its fine by me. This stuff is just pure nonsense unless we see the benches and products themselves. If AMD has the upper hand and I need to upgrade, then AMD for my next upgrade vice versa.

Til then, the talk of async compute, DX12 performance and what not doesn't have any real substance. Both seem to relatively perform well but we know there aren't native DX12 games out there yet (or is there? haven't really followed games inawhile) to fully test the API's capabilities. By that time, these new products will be out with higher performance.

 

[3DVagabond]

An engine tailored around a specific hardware API is not "the most extreme"? Since the first day of their company Oxide has promoted AMD.

It's DX. Most game engines are tailored around it. So, no it's not extreme at all. And again, guilt be association is not evidence, it's just hearsay.

Your posting is getting ignore worthy. It hasn't happened yet, but it's really close.

 

[swilli89]

https://forum.beyond3d.com/threads/intel-gen9-skylake.57204/page-6#post-1869935

Lol did you just cite another forum post as a reference? No that doesn't cut it I'm afraid.

 

[96Firebird]

So every GameWorks game released the last two years are EXTREME biased.

Thank you.

According to you and plenty others here, yes.

Is this news to you?

 

[dark zero]

If this ends true, it will be a massive dissapointment from nVIDIA and a critical hit to gamers.

 

[Slaughterem]

NV is going to need more of these biased games from UE4 in 2016 and 2017 if Pascal is gimped for Async Compute.

Polaris has a lot of changes to the GCN arch, I think Mahigan pointed out that tessellation ROV and other NV advantages will not be much of an advantage going forward. There is no proof of this yet but I tend to believe Mahigan since he has a very good understanding of GPU arch's. IMO If Pascal does not have support in Hardware for Asynch shaders and Polaris improves DX11 single thread results along with DX12 levels features then NV will have to resort to brute force to compete which will negatively effect performance per watt going forward. Even though they have massive market appeal and lots of money to throw at some Devs consumers might decide AMD is the right choice.

 

[.vodka]

If this ends true, it will be a massive dissapointment from nVIDIA and a critical hit to gamers.

As if Gameworks wasn't already a critical hit to gamers, both sides of the fence, for the past few years. If anything this forces nV to stop doing their BS and start to play more fairly. They have to stop sabotaging the competition and themselves for once.

Of course this all depends on AMD playing their cards well, let's hope they don't unexpectedly shoot themselves in the foot again. So far their execution seems better than usual...

 

[swilli89]

Basically this is looking like AMD already paid the price of adding this computational ability into their uArch early on with Tahiti. nVidia skimped out (pretty much a good move for 2014-2015) which is how it achieved better FPS/watt with Kepler onward. This is starting to hurt those gamers who purchased Kepler/Maxwell.

However with DX12 NV will be forced to add this functionality back in and AMD may well be ahead at that point.

 

[casiofx]

Polaris has a lot of changes to the GCN arch, I think Mahigan pointed out that tessellation ROV and other NV advantages will not be much of an advantage going forward. There is no proof of this yet but I tend to believe Mahigan since he has a very good understanding of GPU arch's. IMO If Pascal does not have support in Hardware for Asynch shaders and Polaris improves DX11 single thread results along with DX12 levels features then NV will have to resort to brute force to compete which will negatively effect performance per watt going forward. Even though they have massive market appeal and lots of money to throw at some Devs consumers might decide AMD is the right choice.

Don't forget that by adding compute performance back to the gpu it'll consume more power. Remember last time the original Titan had a toggle for the fp64 compute speed? It had to be disabled to so the titan can boost.

 

[omek]

NV is going to need more of these biased games from UE4 in 2016 and 2017 if Pascal is gimped for Async Compute.

UE isn't the game changer it used to be. Unreal itself is in a bit of a niche.

There was a point in time when almost everything (AAA, small studio and indie, console and PC) ran on or was a derivative of Unreal 2 / Unreal 2.5 and Epic had real persuasion but if you take a look at how many engines are available and "free" today UE4 is an outlier engine that a portion of the indie devs will endeavor.

Most studios have developed their own proprietary engine at this point.

The Division - Snow Drop engine
Battlefield - Frostbyte engine
The Witcher - REDengine
Doom 4 - ID Tech 6
Assassin's Creed - AnvilNext
Dirt - Ego engine
Project Cars - Madness engine
Arma / DayZ - Real Virtuality
Call of Duty - IW engine
Far Cry 4 - Dunia Engine
ect. ect....


It goes on and on. Even a lot of the indies are starting to move away from Unreal, my bet is that there's a three way split between UE, CryENGINE and Unity - also, if these newer engines were derived from any UE version they are so heavily modified and re-written at this point that you'd never be able to tell it's heritage.

It's put into perspective if you look at the engine as a means for a small group of starter developers to get going on one specific platform compared to the behemoth Unreal used to be with the better half of triple A developers using it consuming both platforms, PC and console.

Oh, and about Pascal being without any asyncronus engine, it was obvious. The way NVIDIA kept quiet about the whole thing, pushing developers to detour away from ASC and now downplaying ASC's advantage. NVIDIA's marketing department is extremely transparent which doesn't say much for the people who believe their drivel.