The where and why you dont get the same scores as Ant, Thresh, etc.

[franksch3]

The reason why most of you don't get benchmark results as high as Ant or Thresh, is the simple fact that they benchmark without sound.

Now, from their point of view this is correct benchmarking, because the impact on framerates is different depending on your soundcard. This is a valid assumption.

On the other hand, it makes a card look better than it will be in real life.

I personally are more interested in benchmarks with sound enabled, since that gives me more natural scores. However, for comparing cards without sound is the best way, or using the same soundcard.

I hope that Ant and others will say next time when they put benchmarks on their site wheter they have sound disabled or not. That way, people might not feel fooled when they have lower scores.

Frank

[Marmita]

for ur information Ant DID mention he tested with sound off and still I can't get his scores...

[Ant]

Sound does have an impact on performance but there is no way it would half it. I always state that I benchmark with sound disabled, no one is trying to fool anyone. Sound introduces far to many variables into the equation to produce results that can fairly be considered to be indicative of the performance on a similar system. Here's a snippet from my original <a href="http://www.murc.ws/reviews/TurboGL/TurboGL.html">TurboGL article</a>:

"OK lets cut to the chase I've tested the G400 TurboGL driver against PowerDesk 5.25 with the beta ICD. Test system: PIII 500, 128MB SDRAM, Millennium G400 MAX (150/200), Windows 98. All benchmarks were run 3 times with sound disabled, all were run in 32-bit colour I don't see any point in running in 16-bit, why have a card like the G400 MAX and cripple it with 16-bit graphics?"

[Marmita]

I have a suggestion... from many tests I made with the new drivers I discovered that my results with HQ settings and Bilinear filtering I get the same scores (~0.5fps) so mybe the Beta TGL didn't support Trilinear or mybe it didn't support any other 3d feature ??? I really don't know but it's odd that even when o/c my G400 to MAX's speed and o/c my p3-500 to 575 I don't get better results or at least the same results as Ant.

[Ant]

I'm in the middle of one or two things with my system at the moment which is preventing me from getting the MAX back in and testing 5.30. As soon as I'm finished I'll do a comparison as I've not even laid eyes on 5.30 yet so can't really comment on the performance of anything but the beta TurboGL.

[Marmita]

Ok great, thanks for the afford... I'll check back in a couple of hours to see what you came up with.

[Maniac]

How do you turn sound off in Q3??

------------------
Celeron 300A@464 2.0V, Abit BH6 G400 16meg DH@150/200, Western Digital Expert 18gig, Ricoh mp7040A, Pioneer 6X DVD slot load, Motorola Cable Modem w/DEC ethernet card, Soundblaster 32PNP

[Marmita]

goto to the console and type:
s_initsound 0 <enter>
snd_restart <enter>

the console will close and u will have no sound (don't expect more the %2 increase in fps). to turn the sound back on:
s_initsound 1 <enter>
snd_restart <enter>

[Grimace]

Using HQ settings except Bilinear filtering, 16Bit Z buffer, all Graphics Options except VSYNC on, and with or without sound, I now get closer to Ant's numbers from my similiarly configured machine:

25.4 @ 1280x960
41.5 @ 1024x768
63.3 @ 800x600
73.1 @ 640x480

There definately still seem to be some differences though. It will be interesting to see if his 5.30 numbers are as good has his Beta numbers.

[Snake-Eyes]

Here's something to think about. I realized it after I tried out Q3Test1.08 with the turboGL, but someone else beat me to it in this and another post.

Anyway- it appears that the turboGL is the first to (correctly, maybe?) implement trilinear filtering. I know that all the Matrox ICD's that I've used (including the beta) have shown no difference between the bilinear and trilinear settings in Q3Test. But my eyes just about popped out when I enabled it with the turboGL. I don't know how to state it but to say that things are somehow much more clear and vibrant with trilinear enabled than without it. On the other hand- without it my framerates go up by a large margin (37 FPS on my rig w/o, vs. 27.6 w). This is a timedemo using q3demo1 with EVERYTHING else set to high, vsync disabled, and sound on.

Anyhow the framerates appear more than playable even with trilinear on now, so unless I run into a situation online where it starts to noticably slow down, I'll probably leave it on.

Two thumbs up from this reviewer, Matrox!

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


[This message has been edited by Ace (edited 12 October 1999).]