Quad core testers are needed

[Skeptic]

Hi folks, I would like quad core users to run several tests for us in controlled manner to help understand how Arma2 performs.

UPDATE: It looks like Arma-Mark II is not ideal tool to measure CPU effectiveness of real Arma2 gameplay.

Test results (last updated on June 16th, 2009 ref# p5.n42):
Quad core improves performance over dual core - 2
Quad core does not improve performance over dual core - 1
Inconclusive/retesting - 2 (ref# p4.n35)

Background:

If you want to experiment with this, you can use -cpuCount=X. This will override our autodetection. Changing affinity on the fly is definitely not good, because worker threads are spawned on the game initialization, and by lowering affinity later you cause threads competing for cores. The best way to simulate dual core on quad core is to use -cpuCount=2 when you run the game and then change the affinity to 2 cores to make sure additional cores can never be used when some overshedulling happens. It might be also possible to set the affinity in the OS before you launch the process, that would work as well.

Requirements:

Taken from ArmaMarkII thread

This is an old mission from Toadlife from back in the OFP days. All scripting and idea's are his. Mr. Burns Converted this over to ArmA and now I converted to ArmA2 with a few small changes.

For those who did not use this in ArmA, this little mission will give us a good idea on which systems run ArmA2 the best.

I suggest a standard on settings
Everything on Normal and Fillrate should be 100%. View distance doesnt matter since the mission will automatically set this.

First You should Post specs You can just copy and paste what I have and place your own specs. I only think CPU, Ram, GPU, and the operating system is important we can leave the rest out.

If you want to try it on High as well go ahead and post the settings and scores.

I suggest posting a screenshot of your score to prove your scores.

Download

Set your settings:
Texture Detail - Normal
Anisotropic Filtering - Normal
Terrain Detail - Normal
Objects Detail - Normal
Shadow Detail - Normal
PostProcess Effects- Low

Report to us your specs:
Cpu -
Ram -
GPU -
OS -
Resolution -

Please run four tests and post armamark score for each. If possible please reboot your PC after each test.

Run and report these results
Test 1 - no changes to Arma shortcut (running with 4 cores)
Test 2 - add -cpuCount=3 switch to Arma shortcut (running with 3 cores) and when you run the game change the affinity to 3 cores
Test 3 - add -cpuCount=2 switch to Arma shortcut (running with 2 cores) and when you run the game change the affinity to 2 cores
Test 4 - add -cpuCount=1 switch to Arma shortcut (running with 1 cores) and when you run the game change the affinity to 1 cores

ArmaMark installation

Place the included .pbo file in your missions folder and select it from the SP missions screen.

Included .pbo files:
ArmAIIMark.Chernarus.pbo

ArmAMarkII is installed by simply copying the included file "ArmAIIMark.Chernarus.pbo" into the [...\ArmA 2\missions] folder. Then launch ArmA2, choose Singleplayer > Scenarios and run the testmission.

Shortcut change

"C:\Program Files\Bohemia Interactive\ArmA 2\arma2.exe" -cpuCount=3

Setting affinity
in task manager (firefox as example, you'll be changing arma2.exe)


You will see 4 boxes if you have 4 cores. Keep 3 checked for test 2, 2 checked for test 3 and etc.


Thanks for your help

[JimmyB]

Settings: All Normal (Post-Processing Low)

CPU: Intel Q6600 stock (2.4Ghz)
RAM: 3GB DDR2
GPU: ATI HD 4870 1GB

Listed in test order:

4 cores - 2622.55
1 core - 2554.92
2 cores - 2551.72
3 cores - 2508.04
4 cores - 2527.84
4 cores - 2557.00

Pretty...inconclusive. Results seemed to fluctuate regardless of how many cores were being used. No idea how those results stack up against similarly spec'd machines, but there was a fair bit of stuttering and texture/LOD pop-in throughout.

[W0lle]

Moved as I fail to see what this has to do with troubleshooting.

[Skeptic]

Settings: All Normal (Post-Processing Low)

CPU: Intel Q6600 stock (2.4Ghz)
RAM: 3GB DDR2
GPU: ATI HD 4870 1GB

Listed in test order:

4 cores - 2622.55
1 core - 2554.92
2 cores - 2551.72
3 cores - 2508.04
4 cores - 2527.84
4 cores - 2557.00

Pretty...inconclusive. Results seemed to fluctuate regardless of how many cores were being used. No idea how those results stack up against similarly spec'd machines, but there was a fair bit of stuttering and texture/LOD pop-in throughout.

I wonder if ArmaMark loads CPU significantly to be used as reliable benchmark. Let's get several more scores before we start drawing conclusions...

No problem W0lle.

[BlueSteel]

JimmyB, I think you made the same mistake I did on the first try: You forgot to set the affinity in taskmanager to the appropriate number of cores. Suma said: "...and then change the affinity..." If you don't set the affinity in taskmanager, ArmA 2 will still be able to use all the cores; it will just think it has fewer cores available and spawn less worker threads.


What is also really important: After running the test once, hit esc on the resultscreen and choose "restart". The first run is worthless, since the performance is heavily influenced by loading data from the HDD. On the second run this data is already preloaded, so the result is closer to real-world results. (during gameplay it would stutter only for a few seconds while loading and then run at full speed)

Here is my data:

CPU - Intel Core2Quad Q6600 @ 2.4GHz
RAM - 4GB DDR2
GPU - NVIDIA GeForce GTX 260 (896MB VRAM)
OS - Windows 7 x64 RC
Resolution - 1680x1050

Test 1 - 1 CPU Core - 1536 OFPMarks
Test 2 - 2 CPU Cores - 1671 OFPMarks
Test 3 - 3 CPU Cores - 2396 OFPMakrs
Test 4 - 4 CPU Cores - 2816 OFPMarks



These tests were run at a fairly low CPU speed of only 2.4GHz; I failed to notice that my mainboard did not apply the overclock at the reboot. This happens sometimes after rebooting, I have not been able to find the cause.

So there will be a second set of benchmarks from me, that time with 3.4GHz. But what we can say already is that the game performance does scale pretty nicely with the number of available cores. I will try to include CPU utilization charts on the next run to see how much the cores are actually working during the tests.

[trini scourge]

Cpu - Core i7 (3.6Ghz)
Ram - 6Gb DDR3 1600
GPU - 4870 512mb
OS - Vista 64
Resolution - 1650X1050

4 cores: 4091.98, 4089.21
3 cores: 4052.68, 4061.83
2 cores: 4052.39, 4060.62
1 core: 4129.73, 4123.81

Quite strange that utilizing 1 core was better than all 4.

[Skeptic]

Here is my data:

CPU - Intel Core2Quad Q6600 @ 2.4GHz
RAM - 4GB DDR2
GPU - NVIDIA GeForce GTX 260 (896MB VRAM)
OS - Windows 7 x64 RC
Resolution - 1680x1050

Test 1 - 1 CPU Core - 1536 OFPMarks
Test 2 - 2 CPU Cores - 1671 OFPMarks
Test 3 - 3 CPU Cores - 2396 OFPMakrs
Test 4 - 4 CPU Cores - 2816 OFPMarks

BlueSteel, I think Suma meant to leave affinity automanaged by windows and control number of cores by shortcut switch. EDIT: You're right - updated first post.

Your results seem to show some significant improvements with 4 cores

Cpu - Core i7 (3.6Ghz)
Ram - 6Gb DDR3 1600
GPU - 4870 512mb
OS - Vista 64
Resolution - 1650X1050

4 cores: 4091.98, 4089.21
3 cores: 4052.68, 4061.83
2 cores: 4052.39, 4060.62
1 core: 4129.73, 4123.81

Quite strange that utilizing 1 core was better than all 4.

Did you mess by any chance with affinity settings? Your results also show no improvements with more cores as well as JimmyB's, hmmm. May be once you're running super fast CPU - you shift bottleneck to GPU?

More, more tests we need!

Thank you guys!

[ZeeDD]

I will test that tomorrow on my Q6600@3.6ghz and HD4850 512Mo system.

DD.

[Skeptic]

Be careful using unofficial English patch - it might be activating FADE copyright protection and messing up your performance. Proved to be BS. Updated first post.

[Redkid Joker]

Skeptic, I think Suma means to manually set the affinity in Windows, to make sure the game does not "overflow" onto more cores than you want it too. In other words, I believe BlueSteel got it right.

EDIT: Also, I thought going from one core to two cores would have more of a performance change.