Windows XP Professional sans Service Pack 1 / DirectX 9 NOTE: A fresh install of Windows was not used. With real world testing being
arguably more important than synthetic benchmarking I felt it unrealistic to use
a fresh install. I don't reinstall Windows before playing a game, do you? My
system was, however, thoroughly defragged with a third-party defragging utility
before and after all benchmark utilities, games, etc were installed.
NVIDIA Detonator XP Driver Version 43.00 was used because of improved scores
in 3DMark03 and UT2003 over the 41.09 driver.
The Gainward memory clock had to be reduced from 700MHz to 691MHz for UT2003
for only the Citadel Botmatch benchmarks at 1600x1200 with 4XAA and 4XAF. NOTE: This test DID run just fine with the 41.09 drivers, so the fact
that it did not under 43.00 drivers points to problems with the drivers and
NOT the Gainward card.
The Front Side Bus had to be reduced to 186MHz in order to complete 3DMark03
(specifically due to blue screening during the CPU Test 2).
Intellisample settings in the Detonator drivers were either "Aggressive" or
"Performance" (41.09 and 43.00 name changes respectively).
Texture Sharpening was unchecked in the Detonator Drivers in all cases.
I realized just before posting this story that I'd run all benchmarks with
my Windows monitor settings at a refresh rate at 100Hz (1024x768 resolution).
This was an oversight on my part and I should have run at are a more typical
75Hz or 85Hz refresh rate as the 100Hz likely slowed down many (if not all)
benchmark scores across the board. This review may be updated if dropping the
refresh rate of my monitor proves significant increases in the benchmarks.
190MHz Front Side Bus
Memory Timing 4-2-2
CAS Latency 2.5
MemTurbo II was used
before every benchmark to "scrub and defrag" system RAM. This saved time of
having to reboot before each benchmark. In every case, all benchmarks were begun
with 90~92% (469~472MB out of 512MB) of system RAM free. Before running MemTurbo
II only ~400MB was free.