Intresting. I have the same configuration (though with 512 mb), and I did
also notice a slow down during some plays. However, is it the CPU that is
slowing down, or is it the GeForce? I have seen pictures of what the GeForce
MAP unit looks like, and that should be quite difficult to cool down. I
regularly run other processor intensive tasks, such as numerical simulations
and never noticed any slowdown of the processor speed (instead, the fan
speed goes up). Also, benchmarks such as Sandra run very stable if you
repeat them a number of times. So, why would the processor slow down during
a game?
However, the slowdown does occur and seems processor related. Below a number
of benchmarks made with UT2003 demo:
(I hoped this stays readable, because I directly copied from excel, all
results in fps)
performance nVidia GeForce4 440 Go,
Utdemo benchmark, settings: performance, no AF
800x600 800x600 800x600 1024x768 1024x768 1600x1200
no aa aa 2x aa 3x aa 2x aa 3x no aa
flyby 90.3 51.4 51.4 33.7 33.6 27
botmatch 45.4 27.7 25.9 24.8 26.3 22
640x480
flyby 117.8 118.85 118.95 118.87 118.86 118.8825 0.045735 high perf
botmatch 52.8 53.5 25.7 41.4 36.6 39.3 11.52244 high perf
flyby 118.73 118.82 118.78 118.65 118.84 118.764 0.076354 perf
botmatch 53.4 23.2 31.7 53.6 53.5 43.08 14.58139 perf
When you look at the flyby results at the 640x480 resolution you see that
these are very stable. When you look at the botmatch results (which should
be more processor intensive, espacially at this low resolution) you see that
there are large differences in the results. This is due to a clearly visible
slowdown, that occurs when two consecutive benchmarks are run without a
break. That does not occur at higher resolutions, thus when the GeForce is
likely the limiting factor. Thus, from this perspective you may be right
that the processor is slowing down, and not the GeForce.
Maybe system temperature as a whole is overheating because both chips are
producing a lot of heat? And it might be the motherboard which takes action?
(Though it could be the CPU as well)
What I will do (once I've got a little more time), is run the UT2003demo
benchmark again at low resolution twice, and then directly afterwards SiSoft
Sandra.
How to solve this problem? During system startup I notice that the maximum
fan speed is much higher then will ever occur during operation. If someone
knows a way to run the fans at this high speed, the problem may be solved.
Also running in a high resolution seems to prevent the slowdown....
Greetz,
Derob
