PDA

View Full Version : Deto 44.03 vs 44.67 vs 44.71 vs 44.90 comparison


BigFish7
07-11-03, 07:02 AM
some good stuff... (althought it's only 3dmark2k3)
the intersting thing is that the new 44.71 are not showing that high boost in performance in the P.S 2.0 test althought it's still quite big (82% boost from the 44.03v) as in the older 44.67 we see 142% boost from the 44.03v.
on the other hand we see different picture in the V.S 2.0 test
the 44.71 are geting 57% boost yet the 44.67 are geting 42% boost. (all compared the 44.03v)

P.S there are some Q3 scores as well.

here are some pictures:
DarkCrow's web site (http://www.darkcrow.co.kr)


http://www.darkcrow.co.kr/image/news/0307/4471_01s.jpg
http://www.darkcrow.co.kr/image/news/0307/4490s.jpg
http://www.darkcrow.co.kr/image/news/0307/4471_02.jpg
http://www.darkcrow.co.kr/image/news/0307/driver_result02.jpg
http://www.darkcrow.co.kr/image/news/0307/driver_result01.jpg

here is the table:

Deto 44.90 Deto 44.71 Deto 44.67 Deto 44.03
Single-Texture 1491.8 1491.7 1492 1491.5
Multi_Texture 3198.7 3198.4 3199 3198.7
Vertex Shader 2.0 16.8 16.8 15.2 10.7
Pixel Shader 2.0 29 29 38.5 15.9
Ragtroll 2.0 16.2 16.2 16.3 16.3

download links:

DetFX v44.90 (http://winbbs.intizen.com/board/content.asp?db=wxpdriver&number=2301)

DetFX v44.71 (http://download.guru3d.com/detonator/)

ragejg
07-11-03, 07:07 AM
Good read!

:)

StealthHawk
07-11-03, 07:23 AM
And I should add that before looking at the percentage gain, we all know that 44.03 scores dropped when using 3dmark03 330 "for a certain reason." ;)

Basically, by just looking at the percentage gains when using 3dmark03 330, you aren't seeing results of any practical use. As everyone should know 330 defeated driver detection for shader programs for all IHVs. All drivers released after 44.03 re-enable detection. To accurately compare scores you should use 3dmark03 220. Otherwise, you are not seeing how much better NVIDIA's optimizations are when comparing 44.03 to any later driver.

Grechie
07-11-03, 07:58 AM
are theyre any good improvments with any other nvidia card besides the FX5900U ?
are the Ti's improving much?
anyone tried

BigFish7
07-11-03, 08:44 AM
Originally posted by Grechie
are theyre any good improvments with any other nvidia card besides the FX5900U ?
are the Ti's improving much?
anyone tried

well someone should check it...

BigFish7
07-11-03, 08:49 AM
Originally posted by StealthHawk
And I should add that before looking at the percentage gain, we all know that 44.03 scores dropped when using 3dmark03 330 "for a certain reason." ;)

Basically, by just looking at the percentage gains when using 3dmark03 330, you aren't seeing results of any practical use. As everyone should know 330 defeated driver detection for shader programs for all IHVs. All drivers released after 44.03 re-enable detection. To accurately compare scores you should use 3dmark03 220. Otherwise, you are not seeing how much better NVIDIA's optimizations are when comparing 44.03 to any later driver.
the thing is that now the boost is much lower (P.S 2.0)
which make more sense and leades to the fact that maybe they aren't cheating this time (although we all know that they use their own shaders path, maybe now they are using "legal path").
i don't wanna see how much it got better then the 43.03v
i wanted to show that things are different now (they aren't socoring 38.5fps but 29 which is more like = "no cheats")
it's just a thought tho, it should be checked by experts first.

StealthHawk
07-11-03, 03:42 PM
Yes, but 44.03 is the baseline. 15 -> 29 certainly looks like a cheat to me, what does it look like to you?

This is a benchmark, changing the workload is cheating.

StealthHawk
07-11-03, 03:43 PM
Originally posted by Grechie
are theyre any good improvments with any other nvidia card besides the FX5900U ?
are the Ti's improving much?
anyone tried

Considering that only 3dmark03 was affected by these and other beta drivers, I'll rager that the Ti's aren't getting squat. Since they've been out so long I wouldn't expect any more improvements either. Unless a new popular game, such as Doom3 comes out, for which the drivers will be fine tuned.

BigFish7
07-11-03, 05:01 PM
Originally posted by StealthHawk
Considering that only 3dmark03 was affected by these and other beta drivers, I'll rager that the Ti's aren't getting squat. Since they've been out so long I wouldn't expect any more improvements either. Unless a new popular game, such as Doom3 comes out, for which the drivers will be fine tuned.

i don't know whether they cheat or not.
i do know (almost 100% sure) that the "work" they are doing is the same,
(might be the same might not, that's the experets work :) )
yet they are using their own path therefore they might get a faster process.
you can see it that way when they are using their own path they are going for the card's strongest points in P.S 2.0 (it's called optimization, same thing as intel's P4 HT tech)
intel working with something (althought it's not the same because it's still a tech and not simpale work through yet the point stays the same high speed in aplication due to optimizations).
if nVidia does use their own path (as i'm sure they are) you can see a better utilization of thier silicon moreover if you don't see any different in IQ from the point it was optimized with a specific path then i see no harm in thier optimiztions.
same thing goes for ATI's silicon we have never thought that due to a close work with FM they might built thier silicon in a better way to suit FM's P.S or something more reasonable such as different architecture which maches the P.S work of FM better.
therefore to examen this thing i would like to see more P.S 2.0 bechmarks such as aquamark 3 which should utilize P.S 2.0 effects
or even better, "games".
that way we would be able to see what's going on in a better way.

I hope you got my point...

batterbrain101
07-20-03, 11:14 AM
Point understood here, however the rest of the herd???? :cool:

AngelGraves13
07-20-03, 02:35 PM
I heard the 44.90 are clean of "optimizations"....that is they only have legit ones this time around.

reever2
07-20-03, 03:17 PM
Originally posted by 99 to Life
I heard the 44.90 are clean of "optimizations"....that is they only have legit ones this time around.

I doubt it, their definition of "legit" is messed up because before they were cheating, then futuremark said they werent, so the same cheats could be considered "legit"

StealthHawk
07-20-03, 08:42 PM
Originally posted by 99 to Life
I heard the 44.90 are clean of "optimizations"....that is they only have legit ones this time around.

They most certainly are not clean...read the earlier part of the thread.

According to Digit-Life's results the drivers are free of optimizations for the Code Creatures benchmark though.