AMD's new offensive in the Dev Bribe Wars!

Rroff · 12 Mar 2010 at 16:44

AA on Batman was not blocked on ATI cards... its a subtle but key difference... AA was only enabled on nVidia cards...

...and that was only the last part of the resolve target, AA was fully functional on non-nVidia cards upto the last stage - which for best results required testing and optimizing for a specifc vendor path, that the nVidia tested path happens to work on ATI cards is coincidence no one can be sure it does in actual fact work 100% properly.

Why the game had no generic AA path - well you should be blaming the developer/publisher for that... why ATI didn't provide their own path is a more interesting one... everything points to ATI being too lazy to do it and would rather the nVidia one was enabled untested on their cards because it apparently appeared to work.

Rroff · 12 Mar 2010 at 16:46

james.miller said:
there's no proof though, it's all very circumstantial. That, and the dx10.1 mess with assassins creed. That wasnt nvidia's doing either, it seems. not because there's any proof they didnt have a hand in it, but because nobody can prove they did.

...i mean, maybe they didnt? maybe ubisoft just chose to...remove it, because it didnt work on nvidia card. who knows?

Actually you have a point with assassin creed... nVidia did effectively bribe the developer to remove DX10.1 features, tho it was not removed so as to implement nVidia features instead - but rather to neut any potential ATI benefits.

Orangey · 12 Mar 2010 at 16:47

Blame Epic for not supporting AA in UE3?

RavenXXX2 · 12 Mar 2010 at 16:48

Blah BLah..I don't believe out of the unsubstantiated crap you type, you're so far up Nvidias crack, you're well corrupted.

james.miller · 12 Mar 2010 at 16:48

Rroff said:
Why the game had no generic AA path - well you should be blaming the developer/publisher for that... why ATI didn't provide their own path is a more interesting one... everything points to ATI being too lazy to do it and would rather the nVidia one was enabled untested on their cards because it apparently appeared to work.

i could have almost typed that out for you, Rroff. Word for word.

Answer this, whats so special about this last stage you mention, so special that it took special development from nvidia to make it work when it also worked (specially) on ati hardware?

consider the idea that it's not actually special at all. or, more specifically, it actually is the generic path?

Rroff said:
Actually you have a point with assassin creed... nVidia did effectively bribe the developer to remove DX10.1 features, tho it was not removed so as to implement nVidia features instead - but rather to neut any potential ATI benefits.

As i mentioned: at the expense of the competition's 'features'. Remove dx10.1 is removing that feature, doing so removed the advantage ATi had. But like i also said - i (we) cant prove anything.

Rroff · 12 Mar 2010 at 16:49

RavenXXX2 said:
Blah BLah..I don't believe out of the unsubstantiated crap you type, you're so far up Nvidias crack, your well corrupted.

Are you talking to me? what I said does make sense and much of it can be backed up by anyone with experience with game development.

Rroff · 12 Mar 2010 at 16:51

james.miller said:
i could have almost typed that out for you, Rroff. Word for word.

Answer this, whats so special about this last stage you mention, so special that it took special development from nvidia to make it work when it also worked (specially) on ati hardware?

consider the idea that it's not actually special at all. or, more specifically, it actually is the generic path?

As it happens it is very similiar to the generic path which is why it apparently works on ATI, theres nothing that special about it other than the fact its only tested against how deferred shaders work on nVidia cards. Implementing AA with deferred shaders/HDR is quite tricky even slight differences in how its handled can be the difference between playable fps and <10fps, forcing it on in something like stalker clear sky can demonstrate that.

rhys216 · 12 Mar 2010 at 16:55

Rroff said:
AA on Batman was not blocked on ATI cards... its a subtle but key difference... AA was only enabled on nVidia cards...

It had code to to prevent Ati cards from using AA, therefore they blocked out the competition. There isn't any room for interpretation or spin Rroff, it is what it is...

I really don't know how you can defend this Rroff? Would you argue that if AMD added code that blocked/prevented DX11 on Nvidia GPU's, that they were simply adding DX11 support for for ATI GPU's and not blocking anything?

I mean seriously if your going to provide spin on Nvidia's behalf at least choose your battles...

james.miller · 12 Mar 2010 at 16:56

Rroff said:
As it happens it is very similiar to the generic path which is why it apparently works on ATI, theres nothing that special about it other than the fact its only tested against how deferred shaders work on nVidia cards. Implementing AA with deferred shaders/HDR is quite tricky even slight differences in how its handled can be the difference between playable fps and <10fps, forcing it on in something like stalker clear sky can demonstrate that.

Right! now we're getting somewhere. So the question now then, is how much testing did they actually do if it's that close to a generic path? 'does it work? (switch on) yes it does?'

The one thing that remained consistent throughout that debate was the absolute lack of details concerning what nvidia actually did to get this working. i'm betting they did next to bugger all.

as for the 'they enabled it on x, not disabled it on y' debate - that's just semantics which i myself said in a previous debate, to you Rroff. when you add detection code, of any kind, to enable something on one group of hardware but not the other, the end result is identical - it works on one, and not the other; you are still locking hardware out.

Rroff · 12 Mar 2010 at 16:59

There is absolutely NO code in there that active seeks for ATI cards and disables them... I know the difference is a little hard to comprehend but this is pseudo of what the code does:

Code:

switch(vendorID) {
  case 'nVidia':
    donVidiaAA();
  default:
    //dogenericAA();
}

theres a subtle but key difference between that and:

Code:

switch(vendorID) {
  case 'nVidia':
    donVidiaAA();
  case 'ATI':
    doNOTHING();
  default:
    //dogenericAA();
}

Rroff · 12 Mar 2010 at 17:02

james.miller said:
Right! now we're getting somewhere. So the question now then, is how much testing did they actually do if it's that close to a generic path? 'does it work? (switch on) yes it does?'

The one thing that remained consistent throughout that debate was the absolute lack of details concerning what nvidia actually did to get this working. i'm betting they did next to bugger all.

as for the 'they enabled it on x, not disabled it on y' debate - that's just semantics which i myself said in a previous debate, to you Rroff. when you add detection code, of any kind, to enable something on one group of hardware but not the other, the end result is identical - it works on one, and not the other; you are still looking hardware out.

Yes but theres a difference between actively seeking out and blocking a third party and only enabling code on a tested platform.

rhys216 · 12 Mar 2010 at 17:02

^^^
No Rroff there is no difference both code = the same outcome and accomplish the same goals, and the fact you are defending this is borderline shameful

james.miller · 12 Mar 2010 at 17:04

Rroff said:
There is absolutely NO code in there that active seeks for ATI cards and disables them... I know the difference is a little hard to comprehend but this is pseudo of what the code does:

come on now, dont try to patronise people.

Rroff said:
Yes but theres a difference between actively seeking out and blocking a third party and only enabling code on a tested platform.

no, there really isnt Rroff. you are still achieving the desired result - locking the competition out.

let's change the scenario:

Wanted:
black people with a background in electronics for a new position opening soon

Not wanted:
white people with a background in electronics for a new position opening soon

Whats the difference? you tell me. silly example, but it fits.

Whether this is right, should be allowed, tolerated, is up for debate thats for sure. But what they did is not up for debate.

Oh, i have a question: has anybody looked at the code and confirmed how nvidia actually did it?

rhys216 · 12 Mar 2010 at 17:05

Rroff said:
Yes but theres a difference between actively seeking out and blocking a third party and only enabling code on a tested platform.

The right thing to do would be to leave out the code and should Ati cards have issues (which they didn't) then let Ati fix it in their drivers!!!!!

Rroff · 12 Mar 2010 at 17:05

There is absolutely every difference... if the claim your making is that nVidia purposely went out to deny ATI access.

Which leads to the more important question of why ATI didn't include their own path... and its equally possible that deals were done behind closed doors to prevent them or that they were too lazy to bother... as I pointed out in the old thread from the discrepancies email/twitter posts it would seem that ATI couldn't be bothered.

kylew · 12 Mar 2010 at 17:08

Rroff said:
There is absolutely every difference... if the claim your making is that nVidia purposely went out to deny ATI access.

You're arguing semantics that don't matter.

"are you locked in, or locked out depending on which way the door swings?"

Is akin to what you're arguing.

It doesn't matter which way the door swings, you're behind a door that you're prevented from getting through.

rhys216 · 12 Mar 2010 at 17:08

@Rroff
No there is no difference Rroff unless you were to claim that Nvidia were locking out ALL competition and not just Ati, in either case the actions are anti-competitive and Nvidia could probably be successfully sued!

AMG · 12 Mar 2010 at 17:09

ATI did have the advantage of DX10.1 only reason why it didn't work on Nvidia cards is cos they chose not to have it till much later and now Nv have 10.1 cards okay that is odd

you could blame DX11 for ATIs doing but again no prove

also who destroyed phyX? we know it used to be able to run on CPUs now it can't do that to save its life...hmmm

Rroff · 12 Mar 2010 at 17:10

rhys216 said:
The right thing to do would be to leave out the code and should Ati cards have issues (which they didn't) then let Ati fix it in their drivers!!!!!

How could they leave out the code? AA + deferred shaders is hard enough to deal with effectively in the drivers, implementing work around in the game source code is far more effective - and you don't want your code working on untested platforms - everyone would have been up in arms if nVidia had included their own code path with no sanity checks and performance had been <10fps on ATI cards.

Rroff · 12 Mar 2010 at 17:12

kylew said:
You're arguing semantics that don't matter.

"are you locked in, or locked out depending on which way the door swings?"

Is akin to what you're arguing.

It doesn't matter which way the door swings, you're behind a door that you're prevented from getting through.

rhys216 said:
@Rroff
No there is no difference Rroff unless you were to claim that Nvidia were locking out ALL competition and not just Ati, in either case the actions are anti-competitive and Nvidia could probably be successfully sued!

If your going to claim that nVidia purposefully locked out ATI cards to disadvantage them then the subtle difference in that code means everything.