AMD and NVIDIA get into a hairy argument

Subject: General Tech | May 29, 2014 - 07:43 PM |
Tagged: nvidia, gameworks, dirty pool, business as usual, amd

The topic on NVIDIA Gameworks was discussed at great length on last night's PCPer Podcast and from the live comments as well as the comments on Ryan's original story this is obviously a topic which draws strong opinions.  As it is always best to limit yourself to debating topics of which you are familiar with the facts The Tech Report's article on the aftereffects of the Forbes story is well worth a read.  Cyril had a chance to speak with a rep from NVIDIA's driver development team about Hallock's comments pertaining to NVIDIA's Gameworks and the legitimacy of AMD's complaints.  As you might expect there is a lot of denial and finger pointing from both sides; what long time enthusiasts might describe as 'business as usual'.  Both sides of this argument have vehemently denied ever attempting to undermine each others business but yet both sides can point to specific instances in which the competition has used questionable methods to get a leg (or hair) up on the competition.  

View Full Size

"Earlier today, I spoke with Cem Cebenoyan, Director of Engineering for Developer Technology at Nvidia, who offered a rebuttal to a Forbes story we covered yesterday. In that story, AMD's Robert Hallock alleged that Nvidia's GameWorks program prevents AMD from working with game developers on GPU optimizations."

Here is some more Tech News from around the web:

Tech Talk

May 30, 2014 | 02:30 AM - Posted by Areus

The AMD vs Nvidia argument is obviously a long lasting and very contentious one, but I feel the most important comment that can be made on this story is this:

That is an AWWWWDORABLE picture!

Who's a good semiconductor developer? You're a good semiconductor developer! Yes you are!
(。◕‿◕。)

May 30, 2014 | 01:47 AM - Posted by arbiter

Truth of fact that AMD has pulled same crap as they are claiming Nvidia is doing now, So pretty sad how AMD is whining about it now. AMD really needs to pick fights on fronts they don't have a history of being guilty on.

June 1, 2014 | 11:11 PM - Posted by Anonymous (not verified)

Hmmm... well then perhaps you can enlighten us by providing links that back up what you are saying?

May 30, 2014 | 02:27 AM - Posted by JohnGR

Nvidia has gone many steps further with PhysX by disabling it when an AMD card is primary.
To keep it as simple as possible and to use an example that even Nvidia fanboys there is a chance to understand

What would you say if tomorrow Intel was saying:
"You know something? If you want to use our ultra fast processors you'll have to buy ONLY Intel SSDs. If you buy SSDs from other manufacturers, those SSDs will only work in IDE mode, because we can not verify the compatibility of those SSDs on our platforms."

There are plenty more to write for Nvidia, but I really want anybody's opinion in this one.

May 30, 2014 | 08:30 AM - Posted by gamerk2 (not verified)

To be fair, NVIDIA never tested said configuration, so why should NVIDIA be on the hook for dealing with the long term support costs? The PhysX API is still open, and AMD is free to implement it via OpenCL any time they want to.

May 30, 2014 | 10:24 AM - Posted by JohnGR

So Intel have any right, based on your thinking, tomorrow to say that they support only Intel SSDs, because they can't test the compatibility of so many other SSDs on their platforms. In other words, if you want a 4770K, Intel have any right to choose the SSD for you.

As for PhysX, patches that came out and bypass the locks show that PhysX does work with an AMD card as primary. And of course Nvidia could give the PhysX as a beta when detecting an AMD card and just throw a big red box when installing PhysX informing the user that it can't guaranteed that PhysX would work with an AMD as primary and an Nvidia as secondary card as good as when using only Nvidia hardware.

June 1, 2014 | 11:18 PM - Posted by Anonymous (not verified)

In actual fact the Justice Department has odereded Intel in the AMD vs Intel settlement agreement to refrain just from this type of action. Intel shall not in anyway identifiy and disable performance in AMD or nVidia silicon.

Page 13....

"IT IS FURTHER ORDERED that Respondent shall not make any engineering or design change to a Relevant Product if that change (1) degrades the performance of a Relevant Product sold by a competitor of Respondent and (2) does not provide an actual benefit to the Relevant Product sold by Respondent, including without limitation any improvement in performance, operation, cost, manufacturability, reliability, compatibility, or ability to operate or enhance the operation of another product; provided, however, that any degradation of the performance of a competing product shall not itself be deemed to be a benefit to the Relevant Product sold by Respondent. Respondent shall have the burden of demonstrating that any engineering or design change at issue complies with Section V. of this Order."

May 30, 2014 | 01:45 PM - Posted by renz (not verified)

i can understand that why nvidia do that. but prior to that nvidia did offer AMD a chance to license from them. IMO nvidia just don't want to deal with or fix problems that might come with such combination. if there are problems people will blame nvidia first since PhysX is nvidia product. they will be forced to make sure the combination between amd gpu as primary and nvidia gpu as dedicated physx card will work in any condition. that is more work for them and for AMD they have no reason to help nvidia so such combination will work because above all else physx is nvidia product and not theirs. also back then (2009) they also hope to push their own solution that should rival PhysX in gpu acceleration by working with Bullet Physics. so before things get complicated for nvidia they made decision to disabled physx when such combination is used.

anyway this gpu accelerated physics is a long war between amd and nvidia. but to me at the very least nvidia deliver their promise and we did see game use physx. for amd part i think they were the first that talk about gpu physics stuff and yet they failed with their promise about superior gpu acceleration on radeon cards. TressFX? that is only one solution. that thing is not even complete physics engine like havok, bullet or physx.

May 30, 2014 | 02:56 PM - Posted by JohnGR

The problem is that whatever excuses you can find for Nvidia's decisions can also be applied in the Intel example. But in the case of Intel many who now try to find a good, logical and kind reason behind Nvidia's decision will rebel if Intel was doing what I am describing in my example.

The rest of your post about AMD and Nvidia doesn't really have to do anything with us the consumers. It is a mistake to put our selves in the middle of their fight and let them step all over us.
As a consumer when I buy an Nvidia graphics card also for it's PhysX support I expect it to use it for that matter. I expect Nvidia to deliver the full product that I just bought, not a product that disables features if if doesn't like my graphics card, my USB secondary monitor, my cpu, my hdd or my ram modules.

May 30, 2014 | 07:22 PM - Posted by renz (not verified)

"It is a mistake to put our selves in the middle of their fight and let them step all over us."

but that's how the reality is. whatever you stand was in regard to such situation when amd and nvidia did this the customer will always get the negative impact from it.

May 31, 2014 | 03:01 AM - Posted by JohnGR

So? Lets be the victims? No. And in many occasions companies had to step back or apologize to their customers for their tactics or at least change/abandon those tactics.

June 1, 2014 | 11:34 AM - Posted by renz (not verified)

nope. not really. you were asking for opinion and i just gave mine. there is all to it

June 1, 2014 | 06:52 PM - Posted by JohnGR

OK

May 30, 2014 | 08:53 AM - Posted by Anonymous (not verified)

There really needs to be a third player in the discrete GPU market, I am sure one of the Mobile GPU microarchitectures could be scaled up into a discrete/desktop SKU, it's not a very hard thing to do, it just takes enough funding. The mobile GPUs already utilize OpenGL, OpenCL, drivers, so it's more a matter of scaling whatever amount of execution units up from the lower amount of execution units on the mobile SKU, and wider/hider bandwidth data paths, and some driver tweaking.

One of the major makers of mobile GPU IP was, at one time, in the discrete GPU market, and they could re-enter the market with some low to mid level value based discrete offerings, for the upcoming Steam Box market, and then gradually build out into the higher end market at a later time. Starting out in the Steam Box market and offering the drivers for OpenGL, etc. should not be too costly where driver development funds are concerned, and the Linux drivers are already made, for the Mobile market SOCs. These mobile GPU microarchitectures are already optimized for low power usage, and just may scale up into some very power efficient/low cost home Theater based low to mid range, discrete GPUs for this and the mobile Laptop/ChromeBook/MintBox market.

This constant proprietary graphics API/Middleware/Driver infighting between AMD and Nvidia, and lack of OpenGL, OpenCL, etc. Standards, Linux driver support, among the 2 big discrete GPU market players is getting old, and there is a niche to fill in the Low/Mid Open graphics API driven discrete GPU market, that the Big Boys intentionally do not want to fill.

May 30, 2014 | 08:56 AM - Posted by Anonymous (not verified)

Edit: wider/hider,
to: wider/higher

May 30, 2014 | 11:50 AM - Posted by 7stars

it's true...
completely unnecessary war, to the detriment of pc gamers...
i don't buy Nvidia for its closed view

but the main reason and the main cause is the devs behavior, they are unfair 'cause that's a bad port from console

"Making an open world run on [next-generation] and [current-generation] consoles plus supporting PC is an incredibly complex task."

too bad, i don't care about consoles, i don't care if you have a certain workload, you have to release a finished product, a good optimized PC product...otherwise I DON'T BUY your product, at day one not for sure... so you force customers to wait and then see if you can patch and let the product be almost good as it should...
good PC programming = I BUY
bad PC programming, bad console port without the necessary attention = I DON'T BUY

May 30, 2014 | 12:01 PM - Posted by ZoA (not verified)

Frankly I think pcper's coverage of this issue is dreadfully bad. You persistently repeated several demonstrable lies:

1) you falsely claim this issue is new and generated by AMD after publication of Watch Dogs. This is outright brazen lie, problem of optimization of GameWorks titles on Intel and AMD hardware is known for a long time, and journalists more competent then you have written on the issue for quite some time. Here is example from end of last year:
http://www.extremetech.com/extreme/173511-nvidias-gameworks-program-usur...

2) claim that source code of GameWorks is available for optimization is also outright lie. Source code was closed to all up until less then two months ago, meaning all current GameWorks tiles were developed without any insight in source code by game developers, or Intel or AMD driver coders. Now it is still completely unavailable for optimization of Intel and AMD drivers, and only selected few developers that are able top pay Nvidia's license fees are able to gain insight in to source code, with caveat that even then developers cannot collaborate with Intel and AMD to optimize licensed game code or graphic drivers to run that code, because that would breach the license.

In addition to those lies one of the speakers in podcast maintained it is completely legitimate for nvidia to engage in such source code obfuscation. While this is probably indeed their legal right it is also extremely anti competitive an thus also anti consumerist position. If, on top of what nevidia is already doing, AMD retaliates by adopting same nvidia's practice and starts dumping black box code in their titles that would make them rune like shit on nvidia's and Intel hardware that would be very bad for consumers, and having such practice legitimized and defended by members of your writer staff is a disgrace.

May 30, 2014 | 08:02 PM - Posted by Jeremy Hellstrom

1) err, huh?  Specifically mentioning a website in that post which launched 14 years ago to demonstrate the length of time this has been going on is claiming the issue is new?

2) uh, it is available right now, that statement would have been true 2 months ago but it ain't April anymore.  We did mention that.

3) I was there and I'm pretty sure that sentence was a paraphrase of what our thoughts were

Argue with our points, disagree with our conclusions, disparage our opinions, belittle our mothers ... but we aren't going to be able to have a sensical conversation if you are hallucinating; maybe a fun conversation, but not one that makes any sense.

thanks for reading and watching.

May 31, 2014 | 02:55 AM - Posted by ZoA (not verified)

1) Ryan Shrout maintained in his article this is new and generated by AMD after Watch Dog publication. I quote: “NVIDIA also seemed confused as to why AMD would be on the war path with this story now, with Watch_Dogs” Furthermore Shrout repeated on podcast this is issue AMD created after Watch Dogs, as can be verified by anyone that listens the podcast.

I see no link or reference, in either yours or Shrout's article, that would demonstrate that GameWorks problems is long one, all I see is Shrout repeatedly maintaining it is new, created by AMD after Watch Dogs

“2) uh, it is available right now, that statement would have been true 2 months ago but it ain't April anymore.  We did mention that.”

In fact statement is still true as license is issued to specific developer, Intel and AMD still have no access to source code, access they need if they are to optimize their drivers to run that code.

Furthermore, as code is issued under payed license, developer cannot collaborate with Intel or AMD to modify that code because that would mean Intel or AMD having access to the code. How much optimization developer can make under such conditions is very much questionable.

This is on top of nvidia charging fees for source code access, effectively discouraging optimization for Intel and AMD graphic hardware because that would incur additional, nvida imposed, development cost.

So to say there is access to GameWorks code after April is a lie because Intel and AMD drivers developer have no legal access, and game game developer have only access if they can afford increase in development cost, and under conditionalities that obstruct code modification to ran on other hardware.

May 30, 2014 | 02:01 PM - Posted by Lord Binky (not verified)

Writing specialty code for a manufacturers graphics card is inherently limiting because of the limited coding resources a company can throw at a game. If they really want to progress the industry as a whole, they'd do work on the API standards and refine those, but that doesn't give an edge that can easily take away the edge from your competitors.

May 30, 2014 | 04:08 PM - Posted by Anonymous (not verified)

All the more reason to fund a third player's re-entrance into the desktop, with a condition that the funded product be based on a standard graphics API/Driver codebase, the Khronos group's standard appears to have been around for quite some time, and is heavily used in the Mobile graphics market, as well as the PC/Laptop market, and has many Contributor Members among the leaders of the industry. Google, IBM, Sony, Oculus VR(Facebook), they could fund an discrete GPU, derived from a mobile SKU, and make it so the software APIs/development SDKs/Middleware are open source based. Let AMD and Nvidia fight with each other, and do an end run right around the problem!

June 3, 2014 | 04:32 AM - Posted by Anonymous (not verified)

nVidia never tried to make the industry of videogaming progress.

They just try to vendor lock people to milk them (and it works, given last generation we had the mid range card at 500€ named 680 and the high end at 1000€ named geforce titan, and people were actually happy, because it showed how their beloved video card builder was sooo great to be able to do that).

To do that, they'll screw all gamers, as long as AMD ones are screwed more. It's not like the over tesselated walls and hidden seas were not hurting geforce players ... just less than AMD ones.

Keep throwng money at nVidia ppl, with the same happines you do it now :) After all, the kepler generation proved you want to be raped and like it.

June 2, 2014 | 01:57 PM - Posted by darknas-36

Those two company have been fighting back and forth for the longest time and its no wonder people get pissed at hearing the same crap time after time. It is just a matter of time when the game developers are going to say we are not going to work with you if you keep acting like children.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.