Nvidia’s GameWorks program usurps power from developers, end-users, and AMD

tommybhoy · 1 Jan 2014 at 20:38

eyetrip said:
But, but, but... if AMD could edit and make changes to Nvidias intellectual property, they could get a gajillion more % from it, by working their magicz! Innit!

No one needs to change ip, do you think Nvidia could alter TressFX code?

Nvidia Tress FX performance would have been abysmal if they were denied access to tweaking engine performance.

@Rusty,

To be clear on what I was pointing out in regards to AMD 290X performance, is-it's the easy way to dismiss the underlying point of this thread.

As mentioned earlier, iirc, 560/7850 performance levels is what the majority are playing with, that hardware is probably taking a hit where as the brute force of the 290X is pulling it through-BAO is a very undemanding title for the 290-it shows with the lack of advancement in better/higher IQ.

LtMatt · 1 Jan 2014 at 20:39

MjFrosty said:
What are you even talking about? It's an open library to developers, the only thing you're basing this on is one studio refusing an update for which we don't even know what it contains to WB Studios. AMD will be able to work with developers just as directly. You have no idea what the reasoning behind WB refusal is, and nor has it happened anywhere else yet.

You're just posting blanket statements based on one crummy article.
There is no facts that state that AMD optimisations will be directly effected. Stop going round in circles it's really quite boring. I could start up a thread about how Mantle could directly affected developers affordable time on Direct X optimisations, but I won't. Because I don't have a website low on hits that needs some attention.

Just because you don't understand nor agree does not make it a crummy article. Article > Your opinion, in my opinion.

If you find it boring don't reply or use the ignore feature.

Feel free to start a Mantle thread if you want though.

ToxicTBag · 1 Jan 2014 at 20:44

Gentlemen please keep it on topic and non personal - TBag

Silent_Scone · 1 Jan 2014 at 20:45

I understand fine mate, I'm wagering you've not even read the information on the NV site?

You'd rather read speculative spack from one bloke who couldn't even get a response to his question in the first place. So what else are you basing your concerns on? Other than this WB situation. Because that I don't understand, honestly.

LtMatt · 1 Jan 2014 at 20:46

@Greg The author of the article Joel, has replied to your comment.

Maybe this will clear up some confusions amongst us all... this is for the specific title in question.

I used FXAA because switching to MSAA is precisely what allows the 290X to leverage enough raw power to overcome the GTX 770. My benchmark results are not "massively wrong," simply because you don't like the tests I chose. My test results accurately summarize performance in the mode tested.

When you turn MSAA to full, you push fill rate hard enough that card performance comes down to brute force. Yes, the 290X wins that comparison. This does not explain the performance gap between FXAA performance of AO and AC under identical circumstances.

Finally: GameWorks locks out optimization of specific functions. It does lock out everything -- just the cutting-edge parts. In Arkham Origins, the following Gameworks libraries are used:

GFSDK_GSA
GFSDK_NVDOF_LIB (Depth of Field)
GFSDK_PSM
GFSDK_ShadowLib (Soft shadows)
GFSDK_SSAO (Ambient Occlusion)

Assassins' Creed IV, on the other hand, uses:
GFSDK_GSA
GFSDK_GodRaysLib
GFSDK_ShadowLib
GFSDK_SSAO

Clearly the GW library loadout is customized and tailored depending on the title. These are the libraries and functions AMD cannot optimize. The fact that AMD can optimize the game and improve performance 35% due to other changes does not change the fact that GW-specific changes are locked out. And I believe the original story makes this distinction quite clear.

tommybhoy · 1 Jan 2014 at 20:51

290X stock clocks [email protected]

Silent_Scone · 1 Jan 2014 at 20:51

tommybhoy said:
No one needs to change ip, do you think Nvidia could alter TressFX code?

Nvidia Tress FX performance would have been abysmal if they were denied access to tweaking engine performance.

@rusty,

To be clear on what I was pointing out in regards to AMD 290X performance, is-it's the easy way to dismiss the underlying point of this thread.

As mentioned earlier, iirc, 560/7850 performance levels is what the majority are playing with, that hardware is probably taking a hit where as the brute force of the 290X is pulling it through-BAO is a very undemanding title for the 290-it shows with the lack of advancement in better/higher IQ.

http://www.techspot.com/review/733-batman-arkham-origins-benchmarks/page2.html

Honestly don't see anything untoward there, bearing in mind that is before the patch.

Silent_Scone · 1 Jan 2014 at 20:55

LtMatt said:
@Greg The author of the article Joel, has replied to your comment.

Maybe this will clear up some confusions amongst us all... this is for the specific title in question.

Am I allowed to ask how he knows these functions are locked out? They're fundamental DX technologies.

LtMatt · 1 Jan 2014 at 20:56

MjFrosty said:
http://www.techspot.com/review/733-batman-arkham-origins-benchmarks/page2.html

Honestly don't see anything untoward there, bearing in mind that is before the patch.

LOL that backs up the claim perfectly!!

So 290X faster than a titan with AA but 19% slower with FXAA. So that is the GameWorks penalty of 20%. Maybe more than 20%, not sure how much faster the 290X was over the titan with AA applied. Without AA and AMD brute forcing their way to the top 20% is the performance penalty of GameWorks for a 290X vs a Titan in Batman.

See now if AMD could actually optimize a bit more for this, i wonder what the performance penalty would be. Would there be one at all?

MjFrosty said:
Am I allowed to ask how he knows these functions are locked out? They're fundamental DX technologies.

Ask him. He'll reply to your quote. Check the comments section.

http://www.extremetech.com/extreme/...om-developers-end-users-and-amd#disqus_thread

Silent_Scone · 1 Jan 2014 at 21:02

I was looking more at 7950 performance, seems to scale well enough with competing hardware, in this case the 580 GTX. Of course the 290X is a newer card and thus the drivers are in their infancy. Something you choose to ignore in this case. In fact yes, let's look at the performance in the mid section shall we.

Any thing you care to add on that?

On why a 580 GTX and 7950, which are both very similarly performing cards - are performing very similarly?

Maybe we should ask your friend what it is in AMDs current drivers that they're unable to optimise, and what optimisations exactly were rejected? Would he be able to comment on that at all?

triss · 1 Jan 2014 at 21:06

Since amd drivers have helped performance as listed by AMD basically this locked idea gimping amd is wrong,, they have some access or they simply couldn't do it.
and dont go other they overpowered the AA engine,, as it runs at X speed u cant suddenly double it to compensate for one game,, What ur actually doing is optimising it
which is exactly what you have said cant be done.
Tell me when WB say its locked and they personally cant see into it and that its at the request of nvidia not themselves , Rather than one reporter assuming its so because he didn't get a reply from WB

LtMatt · 1 Jan 2014 at 21:12

MjFrosty said:
I was looking more at 7950 performance, seems to scale well enough with competing hardware, in this case the 580 GTX. Of course the 290X is a newer card and thus the drivers are in their infancy. Something you choose to ignore in this case. In fact yes, let's look at the performance in the mid section shall we.

Any thing you care to add on that?

On why a 580 GTX and 7950, which are both very similarly performing cards - are performing very similarly?

Maybe we should ask your friend what it is in AMDs current drivers that they're unable to optimise, and what optimisations exactly were rejected? Would he be able to comment on that at all?

Take a look at these.

GTX580 gets near to a 7950 with FXAA.(somehow)

Yet once we pile on AA x8 what happens to the 580? It vanished into oblivion. I wonder how it managed to get so close to a 7950 with FXAA?

The GameWorks effect. Only nullified via GCN's superior AA brute force performance.

Maybe you should ask Joel, as you don't quite seem to grasp it. I think you should because we're just going round in circles and he has a superior knowledge to me or you.

Silent_Scone · 1 Jan 2014 at 21:18

You'll also notice there are no other 1gb or 1.5gb cards in that list. MSAA 8x, right well. That's proven me wrong then. Crickey.

I grasp exactly what Joel is saying, but Joel doesn't know any more than you or I other than what he's being told. Subsequently by AMD. News Flash. FXAA works better on NV cards.

No whay?
.

triss · 1 Jan 2014 at 21:18

The GameWorks effect Nvidia sending spending there $$ by sending out there employees out to work with game devs at the start of projects to ensure games run well on purchasers of their hardware.** fixed

LtMatt · 1 Jan 2014 at 21:25

MjFrosty said:
You'll also notice there are no other 1gb or 1.5gb cards in that list. MSAA 8x, right well. That's proven me wrong then. Crickey.

I grasp exactly what Joel is saying, but Joel doesn't know any more than you or I other than what he's being told. Subsequently by AMD.
.

Ok lets try this again. Lets take your vram argument out of things. So a basic 660 2gb card 3.5% faster than a 7950 boost at 1080P with FXAA and the GameWorks effect. How this is possible, ive no idea. Its a god damn 660, low spec.

Slap on some AA and what do we have? 7950 boost now 36% faster. 36%, not far off the 35% AA boost, eh?

triss said:
The GameWorks effect Nvidia sending spending there $$ by sending out there employees out to work with game devs at the start of projects to ensure games run well on purchasers of their hardware.** fixed

Whilst also ensuring AMD cannot provide 'complete' optimization for their users. Great for the gaming industry hey.

MjFrosty said:
I grasp exactly what Joel is saying, but Joel doesn't know any more than you or I other than what he's being told. Subsequently by AMD. News Flash. FXAA works better on NV cards.

Lol you're getting desperate aren't you. So now you're telling me that Nvidia are better than AMD at providing FXAA which normally costs 1-2fps MAX in more demanding games? I rest my case Frosty.

triss · 1 Jan 2014 at 21:31

Complete optimization now,, it was any optimization not long ago...
Maybe if amd spend some $$ on helping during that same cycle they wouldn't have a so called "issue"
Yes all companies should work together does it happen? no
Should we condemn one manufacturer for trying to get the best experience for there buyers? No , And thats applies to either/any side eg mantle or w/e each company wants to develop
Was there a problem with the reject code from AMD to WB? only WB can answer that and until they do its just speculation

Silent_Scone · 1 Jan 2014 at 21:32

It's 8x multi sampling Matt. There is no VRAM aargument. Try running that much MS on a 1.5gb card in any recent game.

Gregster · 1 Jan 2014 at 21:32

So let me try again...

AMD owners are fed up, as they are locked out of FlameWorks/Flex/TXAA/PhysX/FaceWorks/And what ever else nVidia come up with?

Rroff · 1 Jan 2014 at 21:33

Be more interested in 4x results, 8x will hammer older Fermi, etc. cards even the 6970 dissapears off the list and nVidia have put more effort into FXAA and TXAA than 8x in AO.

LtMatt · 1 Jan 2014 at 21:38

Gregster said:
AMD owners are fed up, as they are locked out of FlameWorks/Flex/TXAA/PhysX/FaceWorks/And what ever else nVidia come up with?

No problem with those at all as they don't affect AMD performance.

Answer Joel Greg.