ATI and App Specific Optimisations

[Diplo]

The line they previously used is that they prefered optimisations that were as generic as possible.

"Prefer" is a flexible word. I'm sure Nvidia could claim they'd prefer optimisations that are as general as possible, but it doesn't mean they won't do specific ones.

People need to make up their minds. Are optimisations OK? If so, what is the criteria for allowing them? Should we be told about them? Should we have the option to disable them? Then apply those conclusions to both Nvidia and ATI, instead of selectively finding instances to prove that 'their' IHV is perfect whilst the other is the devil.

 

[ninelven]

edited

 

[jvd]

What are they going to say when you can turn off shader replacement and app specific optimizations on ati cards but not nvidia cards ?

Well, it depends on what generation of cards your talking about. People really don't care that much about R3x0 vs NV3X anymore. Carmack stated that Nvidias optimizations were for the NV3X series (if I read the quote right).

which quote was that ?

Also i highly doubt nvidia stoped driver replacement and other optimizations just because they put a new card out

 

[ninelven]

edited

 

[jvd]

The quote where he said they were cheating.

this quote ? http://www.beyond3d.com/forum/viewtopic.php?p=331833#331833


In the hardocp review he makes no mention of hardware. Simply that nvidia falls off the fast track. That to me implys all of nvidia's products otherwise he wouldn't have made such a broad statment .

 

[FUDie]

The quote where he said they were cheating.

What Carmack said was that NVIDIA's driver "goes off the fast path" with very small changes (to the fragment shader) and that this was especially noticable on NV3x. No where did he exclude NV40 as he was talking about NVIDIA's drivers in general.

This has already been discussed in other threads.

-FUDie

 

[ninelven]

edited

 

[FUDie]

You mean other than the NV40 runs everything really fast comment?

Do you not understand what the word "exclude" means? Please show me where he excluded NV40 from his comments about "falling of the fast path"? Did it occur to you that it can still be fast even without optimizations?

Head to http://www.webster.com if you're unsure about the word "exclude".

-FUDie

 

[ninelven]

edited

 

[FUDie]

Oh great I feel like I'm back in the fifth grade; wow, its like I'm really there. The greatness of message board freedom, when all else fails use personal attacks.

Way to dodge the issue! Where's the quote from Carmack showing where he excluded the NV40 from his comments about "falling of the fast path"? See? my link to webster.com may come in handy after-all!

-FUDie

 

[LeStoffer]

 

[ninelven]

edited

 

[jvd]

Only problem with that quote is that we don't know where its from or when its from .

 

[FUDie]

FUDie,

Well, I've been around here long enough to know the people one can have a useful discussion with. Your reaction to my post that NV40 might not be using the optimizations like the NV3X is extreme to say the least, and it reveals more than enough to me about your character and bias.

I see. I am biased because you can't back up your claim? Odd logic. Yes I replied with sarcasm, I felt your post deserved it.

I must confess I am not really suprised though. I made no moral judgement about the use optimizations. (My first post in this thread was just a joking provocation though many didn't get it.) My original comment with regard to the topic at hand was made (based upon what I believe) in order point out where the focus in the gaming community actually is.

So you ignore Carmack's comment about "falling off the fast path" but embrace the other comment about "running everything fast"? What does "everything" mean? Does it mean everything Carmack ran on it? Does it mean everything DOOM 3 does? Does it mean NV40 runs Excel well? You can't pick and choose. Carmack made both quotes, I'm sure he had a reason for them

I have no further interest in discussing anything with you.

I'm heartbroken.

-FUDie

 

[Florin]

Only problem with that quote is that we don't know where its from or when its from .

Are you sure that's your only problem with that quote? :?

Anyway it'd be from this thread:
http://www.beyond3d.com/forum/viewtopic.php?t=11811&postdays=0&postorder=asc&start=160

 

[jvd]

Only problem with that quote is that we don't know where its from or when its from .

Are you sure that's your only problem with that quote? :?

Anyway it'd be from this thread:
http://www.beyond3d.com/forum/viewtopic.php?t=11811&postdays=0&postorder=asc&start=160

well no. Now that i know its from april and the hardocp one is much newer i tend to believe they are def cheating on the nv3x hardware and may well infact being doing it on the nv40 hrdware . I mean if your going to take the time to replace shaders if the increase performance on your other hardware why not use them.

 

[Florin]

well no. Now that i know its from april and the hardocp one is much newer i tend to believe they are def s/cheating/optimising on the nv3x hardware and may well infact being doing it on the nv40 hrdware .

Ah yes I'm sure you do but see that's the crux of the argument that has popped up here - how to interpret exactly what Carmack has said about this. Now if you'd ask me after looking at the quote above, as well as the quantifier 'especially on nv3x' on that other quote then I'd say he might well mean that NV40 hardware does not see a comparable 'falling off the fast track' in case of a modified shader.

 

[Geo]

[Judge bowler & cigar=ON]

Arguing over "exclude" in this context is roughly as nutritious as arguing over what the meaning of "is" is.

Unless Carmack is an idiot (not a chance) or a complete NVidiot (yes, I know, some think he is, but that would be a gross overstatement in my view) then the quote upsteam in this thread is clearly meant to communicate that in his opinion there are no significant shader replacement cheats in NV40 driver code re Doom3. This would not, btw, exclude --ah, that word!-- other more universal cheats. It's a narrow context he's pointing at.

Now, maybe he's wrong, but that's what he's trying to communicate.

[Judge bowler & cigar=OFF]

No gratuities please.

 

[Pete]

There is nothing inherently wrong with application detection.

It can be used for useful things such as fixing bugs, and applying optimizations. All of which are good things.

It sure makes app-specific optimizations easier to incorporate, though. Dark side of the Force(ware) and all that.

 

[Bjorn]

Are optimisations OK?

Yep.

If so, what is the criteria for allowing them?

Doesn't tamper with the IQ.

Should we be told about them?

Preferably. But i don't expect the IHV's to be completely honest about every optimization they did so i don't see that this would help us that much.

Should we have the option to disable them?

That would be great. But then it's a question of what happens when you click the "disable" checkbox. Are all optimizations really disabled ? Cause it looks better for the IHV if they don't loose that much performance.