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

Doco · 29 Dec 2013 at 13:45

Gregster said:
GFWL is finished and it will be Steam redeemable

I knew gfwl wasnt in operation any more (lost my arkham city save from.this

) and wasnt sure how the disk would work if it was gfwl,.thanks for clear up

tommybhoy · 29 Dec 2013 at 16:14

Doco said:
can get it for just under 15 quid

BAO is ~£[email protected], just got it myself to see what all the fuss is about. :eek:

humbug · 29 Dec 2013 at 16:21

Gregster said:
Sorry if I am not grasping it. Humbug is saying that the 7970 can handle tesselation better (and I think you are?) but the guy on the original article is claiming that Nvidia are adding this amount of tesselation to cripple the competitor.

I had a crap nights sleep and full of cold, so excuse me if I am missing something. Is he wrong in the original article then?

I said Tessellation was a problem for the HD 6K series, that the HD 7K series is much better at handling Tessolation, IE better than the HD 6K series, that Tessellation is not a problem for the HD 7K Series.

C2 was released before the 7K series, long before, i think the HD 6970 was still a relativity new GPU.

pgi947 · 29 Dec 2013 at 16:30

Relatively so, a little googling shows the majority of the HD6900 series releasing Dec 2010 with the 6990 following March 2011, Crysis 2 was released on the PC April 2011.

LtMatt · 29 Dec 2013 at 16:56

I'm not interested getting into debates, 'but its fine it has X amount of fps' that's not the point im trying to make. It never sits well to me that using GameWorks AMD are unable to provide any optimization to the game regarding drivers, Tessellation, Multi Gpu Performance etc. Instead they must rely on Nvidia to do the work for them. Could explain why crossfire has never worked on one of the Assasin Creed IV games and the fact that crossfire never worked on Cod Ghosts for a long time yet SLI was working from day 1. To this day im not even sure if crossfire works on Ghosts. I do know from friends with AMD cards playing it its not a nice experience

My opinion on this is not going to change as it changes the way things have always been done. Previously AMD were responsible for performance in games, now it appears that lies with Nvidia, assuming the game in question uses Gameworks. This is the proprietary form of Mantle and i can only hope it never gets off the ground if it means i have to rely on Nvidia providing optimizations for my AMD gpu.

I pulled out some quotes from the author that set the alarm bells ringing for me.

If you write a game 'the normal way", partnering with AMD or NV means that one company has better, more optimized drivers ready for launch. Nothing prevents the other company from optimizing drivers post launch. So in the long run, games get optimized on both platforms.

Optimized through GameWorks, games are never optimized for AMD at all. That's a fundamental change from how we used to do things. Instead of working with a developer to add support for specific NV functions, Gameworks actively works against the implementation of any AMD-specific functions.

Nvidia can optimize their drivers. AMD can't. That's not an "Nvidia advantage" like PhysX, or TXAA, or G-Sync.

The fundamental difference between Mantle and GW, to the best of my knowledge, is this: mantle does not hurt NVs ability to optimize games in DX11. Developers who agree to use Mantle can still optimize for NV. There are no new hurdles.

GW creates near-impossible hurdles for AMD. I seriously doubt a GW title can support Mantle without developers committing to enormous additional work.

Now, we have a situation where AMD's performance cannot be optimized for these DX11 functions. Writing new libraries may be technically possible, just as it was technically possible for AMD to write its own compiler, but the costs are prohibitive. Again, AMD's performance is resting in the hands of a company other than AMD.

The reason this situation isn't as bad as Intel's compilers is because AMD hasn't paid Nvidia for the right to use GameWorks. Nevertheless, I believe it creates a similar impact. People look at DX11 or the poor performance of Crossfire in Arkham Origins, and they blame AMD's drivers without realizing that AMD *cannot* optimize the drivers for those functions without access to libraries and support from the developer.

And having given WBM a month to reply, and multiple emails, plus talked to AMD about the situation, I think any statement will be CYA.

But since you want more info.

When AMD contacted WBM in October and offered to contribute code to improve tessellation and multi-GPU scaling, they were given three days to do so. AMD sent the code for both fixes over and was subsequently informed that the code would not be included.

That was early November. WBM has gone radio silent since.

You could call that hearsay, and you'd be right. That's why I don't lean on it. I present two statements I can personally verify and a third I have no reason to distrust:

1). WBM did not return my emails.
2). WBM couldn't optimize the GW libraries, even if it wanted to. (Meaning the greater issue exists and is problematic regardless of developer friendliness to AMD).
3). AMDs ability to improve Crossfire or tessellation without WBM's assistance is limited.

Edited to add: I spent a month on this story. It's easily one of the longer efforts I made this year as far as time invested. I investigated multiple titles and performed a great deal of performance testing to arrive at the conclusion that overt sabotage was not, in fact , occurring.

What about IHV optimization a from Vendor 1 that lockout IHV optimization a from Vendor 2? The only way AMD can match this is if the developer agrees to work with them from Day 1 to include AMD optimizations. By the time the game launches, it's too late.

I can't speculate on whether or not NV has baked restrictions into the GW contract because I haven't seen one and don't have visibility on that issue. But I think the central topic -- that GW locks in optimizations for NV but leaves AMD out in the cold -- is a valid one. I'd feel the same way if this was Intel holding the keys to control NV's DX11 performance, or if AMD had created a system that gave them control over Nvidia. Specific optimization for one side or the other is not the same as *preventing* optimization for one side or the other.

Remember, it's publishers making this call more than developers. And that matters for the devs that aren't big enough to call their own shots.

Supporting Mantle does not hurt DX11 performance on NV or AMD hardware. It does not prevent Nv from optimizing DX11.

Gameworks does prevent AMD from optimizing its own performance.

I apologise if people find these opinions offensive, or me posting this article in bad taste. I suggest you make use of the ignore feature rather than attacking me though, thanks.

This post was sponsored by drunkemasters ramblings. Filling your screen like no other.

humbug · 29 Dec 2013 at 17:08

So should we should hold Nvidia responsible For AMD bug fixes and xFire support in those games?

Martini1991 · 29 Dec 2013 at 17:09

But Gregster posted where AMD's driver improved performance?

LtMatt · 29 Dec 2013 at 17:17

humbug said:
So should we should hold Nvidia responsible For AMD bug fixes and xFire support in those games?

As it stands with GameWorks, yes. There is very little AMD can do if what im hearing and seeing is true.

Martini1991 said:
But Gregster posted where AMD's driver improved performance?

Did you actually read the comments in the link Greg provided? AMD can only do so much with driver tweaks. Most of them won't work with GameWorks it seems.

Regardless its not so much about performance in this particular title, more the fact that using GameWorks AMD are no longer responsible for Optimization or performance. This is what worries me.

humbug · 29 Dec 2013 at 17:25

LtMatt said:
As it stands with GameWorks, yes. There is very little AMD can do if what im hearing and seeing is true.

Did you actually read the comments in the link Greg provided? AMD can only do so much with driver tweaks. Most of them won't work with GameWorks it seems.

Regardless its not so much about performance in this particular title, more the fact that using GameWorks AMD are no longer responsible for Optimization or performance. This is what worries me.

Me to, if there is a problem we can Tweet Roy or Thracks and they get right onto it, now what? Tweet Nvidia?

Have we just moved into a two tier PC Gaming model? where you have to have an Nvidia GPU to get serving for some games.

Martini1991 · 29 Dec 2013 at 17:28

LtMatt said:
Did you actually read the comments in the link Greg provided? AMD can only do so much with driver tweaks. Most of them won't work with GameWorks it seems.

Regardless its not so much about performance in this particular title, more the fact that using GameWorks AMD are no longer responsible for Optimization or performance. This is what worries me.

People are making it sound that they flat out can't do jack, which obviously isn't true.

LtMatt · 29 Dec 2013 at 17:28

humbug said:
Me to, if there is a problem we can Tweet Roy or Thracks and they get right onto it, now what? Tweet Nvidia?

Have we just moved into a two tier PC Gaming model? where you have to have an Nvidia GPU to get serving for some games.

Hopefully it doesn't gain too much traction. As long as it sticks with games like Batman, COD and Ubisoft titles im not too worried. If it starts becoming common place with other AAA games then its certainly causes for concern for AMD users.

Martini1991 said:
People are making it sound that they flat out can't do jack, which obviously isn't true.

You can't really do much with a closed library. So we have something where Nvidia can optimize for but AMD can't. That's never a good thing, not now but certainly not in the future with other games.

Martini1991 · 29 Dec 2013 at 17:34

People were flat out with blanket statements of "AMD can't optimise drivers".
Which wasn't true.

LtMatt · 29 Dec 2013 at 17:37

Martini1991 said:
People people were flat out with blanket statements of "AMD can't optimise drivers".

If you didn't already, have a read of the article, including page 2 and comments section. The author deals with a number of questions and even replies to a pessimistic Greg. It also explains that AMD submitted code to improve tessellation performance and multi gpu scaling which was rejected. Even if the Dev was willing to help, it would not have been able to because of GameWorks.

http://www.extremetech.com/extreme/...surps-power-from-developers-end-users-and-amd

Kaapstad · 29 Dec 2013 at 17:39

If you are a games developer and you find that your games start running poorly on half the graphics cards available, you are not going to just accept it, you are going to do something about it.

LtMatt · 29 Dec 2013 at 17:45

Kaapstad said:
If you are a games developer and you find that your games start running poorly on half the graphics cards available, you are not going to just accept it, you are going to do something about it.

You'd think that wouldn't you? Depends how much the developer got paid though, right?

WB Montreal refused to comment on the situation. If I knew that NV had ordered WBM to refuse to work with AMD, than the article would say so. If I knew that developers everywhere were refusing to work with AMD, I would say so.

What you call conspiracy is, in fact, a carefully considered approach to a potentially incendiary topic. This article points out that no overt smoking gun has been detected, but that the inability to optimize is itself a fundamental difference from game development in the past. It does not speculate on why the WBM team has refused to work with AMD, but notes the team's noncommunication. It's more important to realize that WBM couldn't help AMD optimize the GW functions even if it was falling-down willing to do so.

And having given WBM a month to reply, and multiple emails, plus talked to AMD about the situation, I think any statement will be CYA.

But since you want more info.

When AMD contacted WBM in October and offered to contribute code to improve tessellation and multi-GPU scaling, they were given three days to do so. AMD sent the code for both fixes over and was subsequently informed that the code would not be included.

That was early November. WBM has gone radio silent since.

You could call that hearsay, and you'd be right. That's why I don't lean on it. I present two statements I can personally verify and a third I have no reason to distrust:

1). WBM did not return my emails.
2). WBM couldn't optimize the GW libraries, even if it wanted to. (Meaning the greater issue exists and is problematic regardless of developer friendliness to AMD).
3). AMDs ability to improve Crossfire or tessellation without WBM's assistance is limited.

Edited to add: I spent a month on this story. It's easily one of the longer efforts I made this year as far as time invested. I investigated multiple titles and performed a great deal of performance testing to arrive at the conclusion that overt sabotage was not, in fact , occurring.

humbug · 29 Dec 2013 at 18:12

WB also made a statement to the fact that "Batman: Arkham Origins isn't planned for release on the new generation of consoles (e.g. XBox One and PS4)."

CoD Ghosts (apparently another Gameworks title) is avalable on the XB1 and PS4.

The XB1 and PS4 sold 4.1m Units in the first two weeks alone, its a very strange decision for WB to apparently not want a piece of that.

What Dev support level does AMD have for ACIV and CoD Ghosts?

Rusty0611 · 29 Dec 2013 at 19:09

Nothing like sensationalism. I thought it was limited to British newspapers. Apparently not

.

Edit: Sorry this comment was meant for a page or so back. Didn't realise it had moved on that much! I'll get my coat

humbug said:
What Dev support level does AMD have for ACIV and CoD Ghosts?

ACIV runs amazingly well on the PS4. Not far behind PC to be honest.

humbug said:
The XB1 and PS4 sold 4.1m Units in the first two weeks alone, its a very strange decision for WB to apparently not want a piece of that.

The thing is the game didn't sell that much (comparatively) on PS3/360 so the decision to not bring it onto next-gen was probably financial as opposed to any lurid conspiracy theory. I like it, it's a good game but I'm just one guy. The market didn't show much appetite for it for whatever reason. Probably more timing than anything to be fair.

Silent_Scone · 29 Dec 2013 at 19:58

LtMatt said:
You'd think that wouldn't you? Depends how much the developer got paid though, right?

Have you been on godlike productions a lot over Christmas. AMD released a driver that fixed Origins performance the-following day, Matt.

There is no TWIMTBP titles that have adverse performance currently on AMD hardware.

I'm not even sure what AMD would have actually sent WB. They had clearly already been working on a fix prior to get it out so soon after launch. Which in my eyes means they probably had a fairly gold version of the game to work with anyway. If AMD were so uninvolved in the development of any one title, the performance would be far worse in GamesWork titles than it is. If you care to read the information on the NV website developers are free to work with the toolbox. But naturally the most logical answer is that NV pays everyone off I guess. Firms do that all the time. Few million here few million there. On a franchise worth barely that.

The poor chaps being ignored by the studio, that's not great practice granted. I know people in the media and it isn't uncommon to be ignored as companies in the public eye get bombarded by them every single day. There could be any number of reasons as to why the code was rejected, it doesn't automatically mean foul play. Getting a working beta driver out 24 hours after release could indicate the code was rejected to be amended. This is the single and most vague case at the moment that there is anything deliberate happening. I would just be happy you'll have your own API which will hopefully be impressive shortly, and stop worrying about the words of one man.

tommybhoy · 29 Dec 2013 at 20:39

290X stock clocks [email protected]

If anyone can compare results on those settings from the Nvidia crew, maybe we can work out if any performance is gimped or not at the high end gpu side of things.

MjFrosty said:
The poor chaps being ignored by the studio, that's not great practice granted. I know people in the media and it isn't uncommon to be ignored as companies in the public eye get bombarded by them every single day. There could be any number of reasons as to why the code was rejected, it doesn't automatically mean foul play. Getting a working beta driver out 24 hours after release could indicate the code was rejected to be amended. This is the single and most vague case at the moment that there is anything deliberate happening. I would just be happy you'll have your own API which will hopefully be impressive shortly, and stop worrying about the words of one man.

LtMatt said:

WB's response:

Gregster · 29 Dec 2013 at 20:41

How do I get the bench to run?