Date: 02/03/07 (Computer Geeks)    Keywords: no keywords

Many of you know this already, but I'm going to share my recent experiment.

Take the following 3 computer configurations (I say configurations, because I only have one Geforce 6800 card and I have swapped for the tests involving it) for example. The general results are as follows:

3D Mark scores versus relative "real life" gaming performance (Unreal Tournament 2004 Onslaught Gametype)

Network machine name and motherboard type	Specs	Approx. 3DMark05 score	Relative game performance (frame rate, responsiveness, etc.)
"ARIA", server	Dual PIII overclocked 1.41GHz, 2GB PC2100 ECC DDR-SDRAM, Geforce 6800@AGP 2x	2500	Sluggish in general
"MAI", desktop	P4 1.8GHz, 1GB PC100 SDRAM, Geforce 6800@AGP 4x	2500	Fair, very occasional hiccups
"SAKURA", notebook	P4E 3.0GHz, 512GB PC3200 DDR-SDRAM, ATI Mobility Radeon 9700@AGP 8x	1700	Good all around

I was expecting the server to do better than the desktop, since it has two Tualatin processors installed (the very last model of Pentium III). Even though the amount of RAM kept it at a very consistent framerate, it was consistently SLOW. Although the PIII server pretty much matched the P4 desktop in 3DMark05 scores, their gaming performances were very different. The notebook got a low 3DMark score but did much better at gaming than the other two computers. This shows that 3DMark does not give a very good indication of gaming performance at all, especially for machines with specs that are fairly comparable. Of course, if the scores are WAY different, then it would be a more trustworthy indicator. Then again, if the difference is that big then there is no need to even benchmark ("Revy", my gaming machine, posts close to 10000 3DMarks- but of course it's going to score higher because the spec is not close enough to these other three machines to be a fair comparison)

I think something like PCMark might work as a better indicator, but I have not tried it yet. I think I'll try PCMark, Aquamark, Sysoft Sandra and maybe a few others to see which of them actually match up to the differences in gaming performance.

Source: http://community.livejournal.com/computergeeks/1033514.html