More important is the relative performance with OpenGL and Direct3D. It is blatantly obvious that ATI does worse in OpenGL with respect to Direct3D, whereas there are simple tests that one can perform on games that use both API's to show that nVidia does pretty well at both.Sharkfood said:
Why Does ATI Get Beat Down on OGL
- Thread starter memberSince97
- Start date
Relative to what? Doom3 and IL2? Both with known methods that NVidia hardware handles admirably. If relative to "the grand scope of things"- you'll find R4xx and NV4x are quite neck and neck in SS, Quake3, GlExcess, VulpineGL, and various synthetic SDK/sample programs, etc.etc.Chalnoth said:
Just in currently popular OGL games. Once again, not "overall"..
ATI has had a few OGL "blunders" in past driver revs (see the thread here on fbench pure-API performance figures that were broken, fixed, broken, fixed, etc.etc.), but barring those, it's not as cut and dry as so many pointing fingers at Doom3 wish it to be.
The latest poop on Hexus also seems to indicate that Doom3 and other OGL performance in newer titles may specifically be isolated to AA + memory configuration. It almost begs comparisons with some water-cooled 4+ ghz system at 1600x1200 with no AA (in hopes of leaning off being CPU bound) to test this theory/hint towards AA related.
Mintmaster said:
I am asking a question of your testing methodology. Don't equate that to implying anything. I think you may need to stop being defensive about some posts (I can vouch for mine) as not everyone in this thread is out to ding on ATi. But thank you for the veiled insult nonetheless.
Perhaps, I don't know. Does Thief 3 use Carmack's reverse? Anyway, my original question (whether you need to find a game that "plays nice" with a IHV's hardware to properly measure that IHV's driver) doesn't have anything to do with D3 and stencils. I'm curious why X feature in a game is the "wrong" way to go about it while Y feature in a hardware part has been blessed by the heavens as THE "right" way to do things.
That question was answered quite well by that ATi's engineer I quoted. If you don't believe ATi's mea culpa then there's nothing I can say to change your mind.
P.S. Sorry for the late reply, I lost my net access. Anyway, I'm just now reading up on this AA fix for D3.
Subtlesnake
Regular
Because it's patented?Chalnoth said:
By Creative Labs. That is the stupidest f'in software patent in existence. And software patents are horrible things to begin with. Some guy posts the algorithm on their message board, and Creative patents it.Subtlesnake said:Because it's patented?
The only reason that Creative sued id software was because Carmack was nice enough to share with the world how he performed the shadow calculation. And he gets f'in sued over it.
I'm sure that if other development studios just don't tell everybody what sort of stencil shadow volumes they're using, they'll be safe.
Subtlesnake
Regular
Actually I did some research on this and it turns out that the technique was patented following a talk given by a Nvidia engineer to some Creative personal.
From Mr. Dietrich of Nvidia:
"Don't worry about it fellas. I described this technique publicly a few months before they filed the patent - hence Prior Art. Ironically, it was at a Creative Labs developer's forum.
During my stencil buffer talk, I described doing shadow volumes the 'reverse' way. At the time, I didn't realize the major reason why the z fail method is better than the z pass method, although I did realize they were logically equivalent, which is why it's now known as 'Carmack's Reverse' and not 'Dietrich's Reverse'!"
It could have been a coincidence, but the patent was filed before Carmack officially announced anything. Anyway, I thought it might be relevant because in the Beyond3D interview with Dean Sekulic he says:
"Oh, and don’t forget - while Creative Labs holds a patent for 1-pass shadow volume rendering, I'm not touching that with a 10-foot pole!"
From Mr. Dietrich of Nvidia:
"Don't worry about it fellas. I described this technique publicly a few months before they filed the patent - hence Prior Art. Ironically, it was at a Creative Labs developer's forum.
During my stencil buffer talk, I described doing shadow volumes the 'reverse' way. At the time, I didn't realize the major reason why the z fail method is better than the z pass method, although I did realize they were logically equivalent, which is why it's now known as 'Carmack's Reverse' and not 'Dietrich's Reverse'!"
It could have been a coincidence, but the patent was filed before Carmack officially announced anything. Anyway, I thought it might be relevant because in the Beyond3D interview with Dean Sekulic he says:
"Oh, and don’t forget - while Creative Labs holds a patent for 1-pass shadow volume rendering, I'm not touching that with a 10-foot pole!"
Bouncing Zabaglione Bros.
Legend
Yes, there is prior art, but who's going to go up against Creative in court to prove it, and maybe pay millions for the privilige of doing it? This is why the current patent system is so broken - the winner ends up being the person with the deepest pockets and most lawyers. So much for patents protecting the little guy and his innovations.
Subtlesnake said:Because it's patented?
I know you probably know this but it bears mentioning: just because something it's patented doesn't mean no one else can use it. Companies just need to license it and they could also license it from Creative without the need for any money to change hands, like id did.
In fact, UE3 uses the stencil buffer for some shadows and I just don't see Tim using anything but the most efficient method.
Bouncing Zabaglione Bros.
Legend
Mordenkainen said:
ID still had to work to support Creatives sound cards. That's still time and money. Sounds to me like ID was blackmailed into including Creative support for fear they would have to go to court, get sued, pay lawyers and end up delaying their games.
Mordenkainen said:
Why should you have to licence an invalid patent for fear of being sued?
Bouncing Zabaglione Bros. said:
Correct, but that effort can be absorbed into future games using that engine and even future engines. What's most important is actually license fees which hurt your bottom line everytime as it then becomes overhead.
That's pretty much what happened or at least what has seeped to the public. JC even talked he was considering making a stand and using a (slower) 2-pass algorithm but in the end he felt the end user would be most hurt by it (not having to pay fees to Creative was also a factor).
You shouldn't but until someone disputes Creative's patent in a court of law there's no way to tell if the patent is invalid or not. It's similar to the SCO vs IBM in the linux source code case. Even though many felt SCO's claim was bogus some companies still decided to go ahead and license UNIX from SCO just to keep using linux.
memberSince97 said:
Looks like they're still working on it. Catalyst 5.11 addresses some OpenGL difficiencies. See here http://www.hexus.net/content/item.php?item=3668
