How to get reproducible results for comparison:
Both tests use FRAPS with 'save detailed benchmark statistics' all checked.
Additionally, you must set FRAPS to stop benchmark automatically after 120 seconds.
Test1: Argon Prime (Highly CPU intensive)
Start FRAPS with settings mentioned above.
Start game with desired resolution and settings.
Start a 'Normal' game.
When materializing in the Argon Prime system facing towards the south gate, use the left or right arrow key to turn and face your retical roughly at the north gate bounding box.
Optionally - you can disable HUD now (SHIFT h).
Press TAB for boost extension and at the same time press F11 to start the benchmark.
Benchmark will complete in exactly 120 seconds, when you see the red or yellow fps numbers appear again you can exit out of the game.
Test2: Cloudbase Southwest (more GPU dependent and serves as an average for unpopulated systems)
Start FRAPS with settings mentioned above.
Start game with desired resolution and settings.
Start an 'Aspiring Explorer' game.
When materializing in Cloudbase Southwest you are facing the North Gate.
Press T to target the north gate.
Optionally - you can disable HUD now (SHIFT h).
Press "u" to start autopilot simultaneously with F11 to begin the benchmark.
When entering the gate press F11 to stop the benchmark at the point the gate animation begins (actual runtime is about 40 seconds).
Exit game.
Caveat: Each time you start a game there is some randomness to other ship positions in the sector so there will be some variation among tests, but in my findings they are minimal to about 1 fps difference.
In the Fraps folder under the benchmark folder you will see your results, most interesting being minmaxavg.csv
Observations and performance results:
My system specs:
- Gigabyte GA-8IHXP MOBO
P4 3.06 GHZ
NVIDIA 6800 GT (Forceware 81.87 for these tests)
1GIG RAM (850mb free)
Creative Audigy 2
XP SP2
As reference 3DMark05 score ver 1.20 - 5272 (High Quality) 5418(High Perf)
Test Settings:
All tests run with high quality textures and shaders.
Image settings on driver set to high quality, and AA and AF set to application controlled unless specified below.
Vsynch disabled.
Test 1 Results (Argon prime a populated system):
1280x960 with ingame AA and AF checked and AQC=no
Max 14
Avg 9.667
1280x960 with driver 8xAAS and 16xAF and AQC = no (aggressive GPU settings)
Max 14
Avg 8.442
NO HUD - 1280x960 with ingame AA and AF checked and AQC=no
Max 23
Avg 14.017
Conclusion of Test1, Argon Prime system: Showed no real difference at 1024x768 from any of the above. Also windowed mode, which would effect a graphics bottleneck and not CPU as much, showed no impact. So Argon Prime is completely a CPU crippler, or a NVIDIA driver compatibility or engine optimization bug. Only more comparable tests by others will show. No driver settings made any difference with these numbers because the 6800GT was never a bottleneck until finally at 8xAAS and 16xAF it started to slow down a bit but no settings could ever speed the fps numbers up. Removing the hud shows a huge benefit in busy systems, 30% improvement here.
Test 2 Results (Cloudbase Southwest an average system):
1280x960 with ingame AA and AF checked and AQC=no
Max 30
Avg 22.26
1280x960 with driver 8xAAS and 16xAF and AQC = no (aggressive GPU settings)
Max 22
Avg 17.022
NO HUD - 1280x960 with ingame AA and AF checked and AQC=no
Max 62
Avg 42.018
Conclusion of Test2, Cloudbase Southwest: This test shows definite impact by the video card settings and especially showing great improvement when removing the HUD meaning it's not entirely CPU limited for a P4 3GHZ here. Removing the HUD improved average by almost double. Windowed mode showed no improvement here either for my system. In Test1 Argon Prime there is about 550MB of system Ram useage while in Test 2 there is about 400MB of useage. Users with only 512MB of RAM in Argon Prime and other busy systems will see disk thrashing for the paging file causing additional stuttering, which only increases as the game goes on, reaching up to about 1.1 GIG of actual X3 only RAM usage over time.
Overall summary - populated systems don't benefit from a high end video card above a 6800 GT hardly at all since it's very much a cpu bottleneck, while in an average system like Cloudebase Southwest the video card gets to stretch it's legs a little bit.
We really needed a way to compare apples to apples and I think this is an easy solution to compare a busy system and an average system, each showing bottlenecks in either a CPU or graphics card respectively. It's not the end all-be all, but it gives us a point of comparison that utilizes a standard methodology to produce comparable and quantifiable numbers. I expect to see ATI cards paired with fast CPU's do better than NVIDIA in Test 2-cloudbase southwest, just guessing from info I have seen in other threads. Also, AMD processors will probably shine in Argon Prime although it is a beast in that solar system on any gear and some serious optimization needs to be done to the game yet IMO. Many onboard sound cards commonly utilize 30% of CPU useage which is crippling, so generally speaking, you could subtract 1/3 from your cpu speed if you use onboard sound.
Post your results and system specs if you desire and we will see some real patterns develop and see if NVIDIA cards really have some issues with the game comparet to ATI as seems to be the case. Then when patch 1.3 comes out we will have a point of reference to compare against.
