What processor do those getting a slowdown between 5.13 and 5.21 have?

[Snake-Eyes]

I've noticed that there have been a variety of responses about the performance differences between 5.13 and 5.21 (haven't we all?).

I saw in the release notes that the new drivers "fixed" a problem with P3 processors, among other things. Now, I happen to be using a P3(450) for my system, and I got a slowdown going to the new drivers. My idea is that possibly the differences in results has something to do with the processor (AMD [K6/K6-2/K6-3], Intel [Celery/P2/P3]) and the optimizations in the code for whatever the specific chips supports (3D-Now, MMX, SSE). For instance, maybe they made some headway with 3D-Now optimzations, no change for MMX, and had to disable/remove some stuff for SSE support (due to bugs).

Please just post info about your processor and the difference between the two drivers. Maybe we'll see a trend emerge, and have an idea why there is such a reported difference between all of our results.

------------------
Ace

[rylan]

When I switched to the 5.21 from the 5.13 drivers I got a small speed incraese (on the order of 2~3fps). Now, I only have a o/c G200 with 8mb sdram so it may react differantly to the 5.21 than the g400 does. I also have an AMD K6-III 450 @496 so as you said, maybe the drivers made headway with 3DNOW and not with SSE.. guess we'll have to wait for more people to respond here to see.
In the meantime, I'll probably order a MAX

[paul]

With an amk6-2 350 @400 w98 and g200, I couldn't get proper shutdown of system using 5.13. The few games I played before fatal shutdown (rollcage, powerslide) did look enormously better than 4.51, though. 5.21 seems a bit slower but very stable for me. Switching icd betas and drivers will really confuse you in the end.

[Alfie]

amd k6-2, matrox g200,64 megs dram,efa viking mvp motherboard,quantum 9.1+ka hard drive.
Noticed slight slow down in quake II,but no more distortion or video breakup.After flashing bios for g200 (new flash program is cool,what an improvement over the old days !and installing latest drivers from via,had an increase in speed(Sorry no benchmarks,too busy playing!)So it seems so far amd owners may have benefitted the most,also g200 owners!

[Snake-Eyes]

I guess the consensus so far is that most G200 owners got an increase then. As far as the slowdown goes, I just installed the 5.25 drivers, ran the benchmarked with q3demo1 again, and got .1 FPS faster than the 5.21- still much slower than the 5.13's.

Something interesting- I then downloaded the beta ICD and replaced the G400ICD.DLL that was installed with 5.25 with the beta (after backing it up of course). Re-ran the benchmarks- FPS went back up to 25.5 FPS, almost exactly the same as with the 5.13 drivers. Wasn't too surprised, since the blurb on the Matrox download page says this ICD has been optimized somewhat for the Quake/Quake2/Quake3Test games. I went ahead and compared version numbers between the backed up 5.25 ICD and the beta. Guess what- the beta IS the 5.13 ICD! Oh well, not a problem for me- in fact, I'm very happy right now, cause at the moment, Quake based games are the only real OpenGL games I play.

It does prove (to me, at least) that something that changed between the 5.13 and the later drivers has hurt the OpenGL performance on some systems (still waiting on more replies to see if it is processor dependent or not though)

------------------
Ace

[Himself]

Paul,

I couldn't get a proper shutdown either, I found that a motherboard bios update fixed the problem.

[Himself]

ACE,

The only performance difference I've noticed on an AMD K6-3 400 was a 10% increase with the beta ICD. 5.13 and on have had consistant results, didn't vary a hair, vsync wasn't an issue, hard to imagine getting 44.2 fps at a 60Hz refresh rate with VSync on. The default could have been left with VSync on for all it mattered, just so long as the OpenGL bits to set it from within games worked properly.

[Rags]

Vsync was disabled in the 5.13 ICD. And it could not be enabled, that is why you didn't notice any difference.

Rags



------------------
P3 450@560, BH6, 128MB PC100, Delta CD ROM, Mill.G200/G400, 8.4G WD, 8.4G seagate, SB16 (temp.), Yamaha Burner, Zip, Sportster 56K int., Storm Scanner.

[Snake-Eyes]

Thanks for the Reply Himself. Sounds like the AMD users may not have had any real optimizations performed for 3D-Now (at least not any more) than before. Probably you benefit from the extra stability then. I found another P3 user that got a framerate hit with anything by 5.13 or the betas, and he's using a 500, not a 450. Starting to sound like most P3's are taking a hit. Still waiting for more info, especially about P2/Celery systems.

------------------
Ace

[Liquid Snake]

I have a P3-550. With the 5.21 drivers, my scores went UP compared to 5.13 (but only about 1fps). The beta ICD is even faster (up to 15fps faster) than 5.13 or 5.21/5.25. Hmm, so why am i the lucky one?

[Liquid Snake]

BTW, the GL performance in 5.25 was EXACTLY the same as in 5.21 (24.8fps in 1024x768x32 q3demo1). THe D3D performance went up a bit. About 10 or so more in 3dmark99 (wow i got 5305, never had a score above 5300 before.

[Scytale]

I experienced the slowdown on a P2, and I don't think it's a P3 in disguise

[Joel]

As one of the beta boyz I can assure you that the beta ICD that has just been released is not the same one as the one in the 5.13 drivers. There was a reason that that ICD was never released publicly and if you will take the time to run Half-Life in OpenGL you will see why.

Also as someone who has done programming for a living I know that sometimes fixing one bug can introduce or show up other bugs.

Joel

[Himself]

Rags,

With Q3Test using the reghack made no difference on vsync, the game sets it, the default doesn't matter. Quake 2, no idea, I don't play either game.