For the gaming benchmarks all component and system settings remained at
default values. Specific settings for resolution and application of AA and AF
are identified in the tables below. The intent of the tests was to compare
performance between the single card and SLI configurations and to achieve
stability in all stress tests of the video cards and system components.
Recorded scores are in frames per second (FPS) and were the result of
averaging the scores of a minimum of three runs through the particular benchmark
at the indicated settings.
Doom3, HL2, and FarCry make up the benchmark suite for this review. The
latest patches/versions were used. Quality settings set to maximum detail in all
benchmarks. Resolutions were set in the benchmark and AA/AF settings applied by
the NIVIDA control panel. Forceware 77.77 drivers were used in all benchmarks.
Game Benchmark Results
The 7800GTX provided good gameplay in Doom3 up to 1600x1200, 4xAA/8xAF, but
struggled when pushed to 8XS/16AF. In actual gameplay I found 4xAA and 16xAF
provided good overall gameplay. The 6800 GTs in SLI best setting was 1600x1200,
no AA/AF, with playable frame rates found at 4xAA/8xAF. In actual gameplay
2xAA/8xAF, provided solid play.
HL2 provided a CPU bottleneck for the 7800GTX except when opened up to
1600x1200, 8XS/16xAF, where even though stressed put up good numbers averaging
72.9 FPS. A brief round of play in the game proved this setting would be
acceptable. Again, the 6800 GTs were handicapped by the CPU bottleneck. The SLI
proved also that the settings could be cranked to maximum and still get good
smooth gameplay. Either single card 7800GTX or an SLI-configured setup would be
a good choice for HL2 using this motherboard as the basis of a gaming system.
In game settings were set to 'maximum details' and maintained throughout all
benchmark runs in FarCry. The 7800GTX makes use of the extra 80 MHz clock
advantage and the 100 MHz in memory speed advantage does not hurt. Still, the
SLI setup with the GTs running 350/1.0 did a commendable job in this benchmark
staying even with the 7800GTX in all resolutions without AA or AF applied. Once
AA and AF applications were set the tide turned in favor of the 7800GTX
outperforming the SLI setup at around 10% at 1280x1024, 8xAA/16xAF, and about
19% at 1600x1200. 8XS/16AF put a strain on both setups and neither would be
acceptable except for maybe the 7800GTX at 1280x1024.
OVERCLOCKING
Overclocking starts with the motherboard and its adjustment capabilities
(tweaks). It takes quality components (both on-board and add-on), a good circuit
design, board layout, and a good Bios to set these adjustments. Probably the
best Bios for overclocking that I have seen to date is provided on the DFI
series of motherboards. Though adjustments on those can be intimidating to the
novice overclocker and can even make the most knowledgeable scratch their heads.
The EVGA 133-K8-NF41-AX did not disappoint in the overclocking area. The Bios
could be improved but provides some interesting possibilities for the
overclocker as evident with the VDIMM selection ranging from 2.55V to 3.25V.
High front side bus (FSB) settings are highly desirable but can be at a cost
to performance. Obtaining a high overclock does not necessarily translate into
higher performance. Indeed, the highest performance setting that you may acquire
with a given system will usually be at some clock setting less than the highest
actual overclock settings obtainable.
For the memory settings I am open to experimentation of trying both loose and
tight timings. I do avoid the DRAM 2T command rate and strictly go with 1T when
I am looking for the highest overall performance possible. I am talking about
the best overall performance achievable and not putting down the use of 2T
command rate as it may offer the only course to system stability in a particular
application. Of course, to get the highest possible it takes good high quality
fast ram.
Of the ram I have available for this review I choose Patriot PC3200 XBL
because of its low latency at 2-2-2-5, 1T, and the fact that I have used these
particular two sticks of 512MB memories on at least five different motherboards.
Performance of these memory modules have been consistent on DFI NF3, NF4, and
Intel 775, Gigabyte NF3 and 4, and MSI KT800.
The first overclock goal was 220MHz with the default multiplier (11x) and was
achieved with 2-2-2-5 timings, 225MHz was no-go as some instability became
apparent. Loosening timings to 2-3-3-7 produced a very stable and high
performing overclock at 225MHz up to 230MHz. The attempt at 250MHz was achieved
but with loosening the timings even further to 3-4-4-8. Performance was good but
still the 230MHz exhibited better overall results. Dropping the multiplier
allowed for passing 265 MHz but due to time restraints this concluded my
experimentation with overclocking this motherboard.
