I have an integrated GeForce 4 with 64 mb shared memory DDR333, an AMD Athlon-XP 1700+(1400~ Mhz) and over 300 mb of RAM.
Right after installing I configured my NIC and Internet connection and started emerging a few games, just to see how well they performed under my new source-optimized Linux system.
I first emerged Tenebrae, which I had already ran under my Win2K where it had performed pretty poorly and thus I considered to be a nice benchmarking tool. I was very impressed when I first ran it as I got twice as many FPS as I got under Windows.
I thought I was in heaven, and couldn't wait to see how well Quake III would perform, expecting this twice-as-fast thing to be a constant in all games. I immediately installed the Quake III binaries and used a symbolic link to my NTFS partition for the data files. I ran it, and it worked just fine, except for the fact that the FPS count was about 25% lower than what I get under Windows. What puzzled me the most was the fact that I had done this before on this very same machine running an unmodified Mandrake Linux (default kernel, default everything) and the FPS count was very much the same as in Windows, not lower not higher.
At this point I should've done a few more comparisons between Linux and Windows performances on same games, but being a knowledge-thirsty noob as I am I decided to just ask you people for some technical explanation. I have come up with a few theories myself, but I really doubt they may count as I have no real proofs of what I think:
a) 'Quake III is slow': I've read somewhere that Quake III has no 3dnow! support whatsoever, that it's optimized for Intel chips extensions instead. This might be total bullshit, but could by any chance exist some sort of binary match (or mismatch) in terms of brand chip optimizations between the kernel and the game itself? Like it would run faster on a kernel compiled in favour of Intel chips (eg: Windows, Mandrake) against those optimized for AMD?
b) 'Tenebrae is fast': Perhaps Tenebrae just runs faster cuz I've compiled it from source, thus squeezing every possible optimization out of it. I haven't been able to compile it under Windows yet, and I probably won't be in the future, so I guess I'll never know. Anyway, if this is the case then that could mean my whole system is running under-average, which leads me to the next question:
c) 'GRP is slow': How optimized is the GRP for Athlon-XP processors? Would it just be the same if I recompile everything from source using my own optimized GCC options?
d) 'My system is misconfigured': Perhaps my current system configuration is causing this strange behaviour, but I really doubt it as I haven't messed with configuration files more than it was strictly necesarely to.
Sorry about the long post, hope anyone takes the time to read it and answer it.
Many thanks
