Juhana Sadeharju wrote:
Hello. I found a reason why ATI nor NVIDIA provides us hardware
details:
http://www.futuremark.com/companyinfo/3dmark03_audit_report.pdf
Regarding ATI: "This performance drop is almost entirely due to 8.2%
difference in the game test 4 result, which means that the test was
also detected and somehow altered by the ATI drivers."
Nvidia is worse: they have 8 cheats in their driver.
Ugh, this is OLD news. You're 2 years late...
It is no wonder why they don't want release the hardware details.
They simply don't want a driver which does not contain the cheats.
Why would that make any difference for them? After all, the open-source
driver would be slower without the cheats, so they could provide an
additional reason why you should use the closed-source driver from them
(not that it likely wouldn't be faster anyway, even without cheating...).
Btw, both Nvidia and ATI use tricks which aren't really cheats, for
instance the "brilinear" filtering and aniso filtering optimizations
(and afaik aniso filtering isn't fully specified, so you can't even
really cheat there even if you wanted, though there is some general
expectation what it should do). You can control at least some of these
optimizations in the driver control panels (though every here and then,
usually when new cards are launched, disabling some of the optimizations
won't work mysteriously, a "bug" which is usually fixed when the initial
reviews of the cards are over...). "Brilinear" is supported for r200
even in the open-source driver, though it's manually controlled and
certainly not used by default. At one point I even experimented to
autocompress textures (what ati's driver does) though I gave that up as
its usefulness seemed limited (and it's not opengl conformant).
And, nowadays usually even app-detection "cheats" are not necessarily
considered evil, as long as the same output is guaranteed (and if it's
not just optimizing for a benchmark run, e.g. the static clip planes
nvidia did for 3dmark03). Though it's probably something you'd want to
stay far away from in a driver developed by the community, as it
certainly increases driver complexity - you want good general case
performance, not lots of app-specific optimized paths just to increase
performance in those particular apps by 3%.
Please continue developing reverse-engineered, open sourced drivers.
As time permits...
Roland
-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO September
19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
--
_______________________________________________
Dri-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/dri-devel
