Just how synthetic is 3dmark 2001 SE?

[squidman]

I dunno, i was just benchmarking my system to compare the results to those of other online benchmarkers. For some awkward reason, this is what i got:
14102 - no aa no af
11500 - 2aa
11100 - 2aa 4af
11500 (??!) - 2aa 8af
11130 - 4aa
11600 - no aa, 16af
11350 - 4aa 16af.
Whoa!!! This was done on a freshly installed drivers, default benchmark, cat 3.4 drivers... i thought something was wrong, so i reinstalled the 3d mark 2001 SE and repeated the process (i was busy doing other stuff while it was benching). Same stuff, in 11k's, with little detiriorations. "deafult" (no aa no af) is just stuck at 14102, doesnt change a point. FSB is at 175mhz, memory at 7-3-3-3.
I know that 3dmark is synthetic, but this??!! Anybody got a comment? Games performance (like it should) once i tinker with aa/af... but 3d mark? The ULTIMATE gamers benchmark? laughable.

 

[Jeff7181]

Well... for one thing, ATI cards are really good with AA and AF... they don't take as much of a performance hit as GeForce4's do.

If you wanna see how accurate that is... those scores are all about 80% of your score with no AA and no AF.

Now fire up a game and save the game at a certain point that has quite a few things on the screen... couple light sources, things off in the distance, things up close, even rain if the game has rain in it. Then use that to benchmark, use something like FRAPS to monitor the FPS and test it at each setting you tested 3DMark in. That should keep ya busy for about an hour or two... I'd be interested to see your results.

*EDIT* Also make sure VSync is OFF

 

[squidman]

vsync? what does vsync have to do with the framerates? Yeah, ill d/l the fraps to see how it goes. But so far - UTII is playable/enjoyable at 1024x768 32bpp, at aa and af...even at the beginning of ythe game, where theres a big-ass exposed world...The fps stays at 28, 4aa 16af. At 4aa its 40 (on that big ass level 1 + intro). At 2aa and 2af (the mode i use) its 50-60fps on that level. I use the same scene: when Marshalls ship flies over the entire island.

I've noticed, that theres no difference ( image quality-wise, not performance wise) between 2x aa, or 4 or 6x aa...thats why i stick with 2x. But difference between 2xaa and no aa - is immense!!! Jaggies are virtually unnoticeable. As of af...in games, when u add af to aa - the performance drops. If u just use af - almost no performance hit (?). framerates on that level didnt change (80 to 68 (no af and 16af respectively).
Sides, no game benefits from using 16, or even 4af as much as operation flashpoint. Pretty useless if u ask me. Also, i dont see difference in DX games when i use AF...weird.

 

[yhelothar]

first of all... to get an accurate score from 3dmark.. you need to be using 1600x1200 at least now.. because 1024x768, the cpu is clearly holding everything back..

oh yeah.. 3dmark03 is far more synthetic than 3dmark2001

 

[JonnyBlaze]

where did you change the settings at? in the driver or the program

 

[squidman]

driver, of course.

 

[squidman]

these are taken from ut intro, by fraps.
from fraps log:

2003-06-14 08:34:58 - Unreal2
Frames: 2843 - Time: 57406ms - Avg: 49.524 - Min: 33 - Max: 83 -4aa 16af

2003-06-14 08:37:20 - Unreal2
Frames: 3538 - Time: 57750ms - Avg: 61.264 - Min: 39 - Max: 99 -16af

2003-06-14 08:39:14 - Unreal2
Frames: 3720 - Time: 57484ms - Avg: 64.713 - Min: 44 - Max: 101 -16 quality af

2003-06-14 08:41:26 - Unreal2
Frames: 3705 - Time: 57937ms - Avg: 63.948 - Min: 48 - Max: 91 -2aa -2af, quality

2003-06-14 08:43:22 - Unreal2
Frames: 4095 - Time: 57703ms - Avg: 70.966 - Min: 53 - Max: 101 -no aa no af