Jump to content
Sign in to follow this  
Skeptic

Quad core testers are needed

Recommended Posts

dude my i7 is 8 core

whoever says im wrong should buy glasses

your i7 is NOT 8 cores. You have 4 cores with hyper-threading enabled. Windows will read it as 8 processor threads.

I had P4 3.4 HT, it's a one processor, but because of Hyper-Threading, windows lists an extra processor.

Share this post


Link to post
Share on other sites

so that means i need my target parameter on cpuCount=4or 8 ??

cuz device manager says 8 cores

and gleek excuse me for beeing offensive

Share this post


Link to post
Share on other sites

Does the Dedicated Server exe use all cores?

Cheers

GC

Share this post


Link to post
Share on other sites
Does the Dedicated Server exe use all cores?

Cheers

GC

Yep , I tried the dedicated server on my Q6600

And it does appear to run across all cores

Share this post


Link to post
Share on other sites
so that means i need my target parameter on cpuCount=4or 8 ??

cuz device manager says 8 cores

and gleek excuse me for beeing offensive

device manager will say 8

As the story goes your Core i7 has 4 physical cores that process but Intel added hyper threading again to the Core i7 series, this causes windows to detect double the amount of cores when there's really not, however there are performance advantages of having hyper threading when using multi-threaded code but having a 'true' core would net you better perfomance

http://en.wikipedia.org/wiki/Hyper-threading

Share this post


Link to post
Share on other sites
Ok after properly setting the affinities these are results:

Cpu - Core i7 (3.6Ghz)

Ram - 6Gb DDR3 1600

GPU - 4870 512mb

OS - Vista 64

Resolution - 1650X1050

4 cores: 4130.93

3 cores: 3717.23

2 cores: 3039.04

1 core: 2420.49

More along the lines of what I expected. In my case using 4 cores shows a 42% performance increase over 1 core, 27% increase over 2 cores, and 10% increase over 3 cores

doesnt i7 have 8 cores though? edit, the core 2 quads are not native, while i7's are and have 8 cores when HT is enabled

Share this post


Link to post
Share on other sites
doesnt i7 have 8 cores though? edit, the core 2 quads are not native, while i7's are and have 8 cores when HT is enabled

8 threads not 8 cores

Share this post


Link to post
Share on other sites
the command -cpuCount=x doesn't work ?

why using task manager ?

It only works together with affinity. Please read first post carefully.

@PhatVybz: Enabling HT on Core i7 to get 8 threads will actually slow you down a bit (since you don't get actual 8 physical CPU cores). 4 cores is plenty for Arma.

Share this post


Link to post
Share on other sites

Can someone with a2 and a quad or i7 do the following test, please?

Get procexp from sysinternals.com (it's microsoft, they bought sysinternals)

Run procexp (as admin under vista/w7).

Run arma windowed.

Start (editor?) a mission with 100+ AIs in a firefight.

Switch to procexp, double click on arma2.exe, select the threads tab, sort by CPU column.

Bring the small arma window to the foreground, while keepin the threads tab in view.

Now repeat with cpuCount set to 1, 2, 4 (and 8 for i7).

Report on how many threads there are and how much % of CPU each one is getting.

DO NOT SET AFFINITY, please.

thank you :)

Share this post


Link to post
Share on other sites

i did dl procexp

but i have no clue how to use it and where to find the details you need

can also someone confirm my question with a answer

if device shows 8 cpu does cpuCount needs to be 8 or 4 on that i7

and its weird but disbaling HT is better for arma?

its confusing to me to disable performance technologies

BTW the latest ati drives increase and stabilize from alot the fps tho it seems ArmA locks fps at 60

Edited by PhatVybz

Share this post


Link to post
Share on other sites
i did dl procexp

but i have no clue how to use it and where to find the details you need

can also someone confirm my question with a answer

if device shows 8 cpu does cpuCount needs to be 8 or 4 on that i7

and its weird but disbaling HT is better for arma?

its confusing to me to disable performance technologies

Depends, you can try with cpucount 8 and no affinity or cpucount 4 and affinity 1&2&3&4 or 1&3&5&7. I dont know how the virtual cores are assigned, maybe first real core is virtual cores 1&2 or 1&5 and last real core is virtual cores 7&8 or 4&8.

The idea is that you only have a thread per real core. If you have 2 threads per real core you are kind of reducing your per core cache in half. You can tell us what works best on i7.

I dont have arma2 yet and only a e21xx dual core, cant test myself.

What im curious about is the load on the different threads

  • Run procexp (as admin under vista/w7). Open menu view->"update speed"->10 seconds
  • Run arma windowed, create a shortcut adding -window to the command line parameters. Let arma run in a smallish window 800x600 or even less.
  • Start (editor?) a mission with 100+ AIs in a firefight, should tax your cpus. If you have a uber-system, copy-paste a few more hundred AIs until you see your frame rate go down.
  • Switch to procexp while arma is running the heavy firefight.
  • In the list of processes locate arma2.exe and double click on it
  • A new window appears with information about the selected process (arma2.exe). The window has multiple tabs on the top, select the threads tab.
  • The threads tab shows all the threads the process is running, you can sort by CPU usage by clicking on the top of CPU column (2nd column from the left).
  • Bring the small arma window to the foreground, while keepin the threads tab in view. The foreground arma window should not overlap the procexp
    window with the CPU per thread usage data. This way the game runs in the foreground but you can still see the reported data in the other window.
  • Report on how many threads there are. Each line in the threads tab is a single thread. Each thread has a "start", thats the address where the thread started running, i believe. In any case, let's name the thread by it's start address.
  • Now report how much % of CPU each one is getting. Usually not all threads are running all the time. See which ones are running (using the "name" from the last point) and report what CPU% column for each thread reports. The cpu usage % usually is not constant, just try to give us an average for it.

Now repeat with cpuCount set to 1, 2, 4 (and 8 for i7).

Doint this test will tell us: how many threads is arma2.exe starting for each configuration selected (1,2,4,8). And how much are those threads doing.

For example, if there is a thread calculating sounds, that thread should be doing around 5% or less. If you are in the menu, possibly only 2 threads are running, one doing sound and another drawing. If you start a heavy firefight with 1000 AIs, 8 threads should show CPU usage as arma2 lets each thread handle some AI each.

DO NOT SET AFFINITY, please.

Also, running fraps while testing could be usefull to get a quick and unscientific feel for performance.

thank you

Share this post


Link to post
Share on other sites

it runs 27 thread count and 18 working hard no matter what cpucount

when on the editor 10X10 infantry group units fighting each other

and sometimes some threads go disabled and end up with 24 or 25 thread

otherwise my fps and cpu usage is same

is this as it should be or am i doing something wrong

i cant check the procxp while doing the stuff on arma but i have it in front of me while doing arma

arma was windowed mode taking half of my screen and procxp taking the other half of my screen

is there a way to create a log file with procxp??

also i can confirm that the i7 show 8 core but cpu 0 and 1 use core 0

cpu 2 and 3 use core 4

cpu 4 and 5 use core 8

cpu 6 and 7 use core 13

Edited by PhatVybz

Share this post


Link to post
Share on other sites

Hey guys I'm going to test cpucount=4 and cpucount=8 with my i7's hyperthreading disabled and enabled respectively. I'll post back with results.

My rig:

i7 920 @ 4.00ghz

6GB DDR3 PC-12800

ATI HD 4870 512mb

Asus P6T Deluxe

Win7 64-bit

Test Settings:

Video Memory - High

Post Processing - Low

Anti-Aliasing - Off

All other settings - Normal

Resolution - 1680x 1050

RESULTS:

Test 1: Hyperthreading Enabled in BIOS -cpucount=8

1nd Run - 4440

2nd Run - 5129

Test 2: Hyperthreading Disabled in BIOS (max 4 cores)

1st Run - 4626

2nd Run - 5353

Test 3: Hyperthreading Enabled in BIOS -cpucount=4

1st Run - 4410

2nd Run - 5201

Core Utilisation

4VS8.jpg

Conclusion & Findings:

Firstly a caveat to my results, since I'm using an ATI card I have VSync forced on and cannot disable it so my frame rate was capped at 60fps throughout significant portions of the benchmark.

Now to the results, there doesn't seem to by a big difference between HT enabled and disabled, however I must say there was a significant increase in texture pop-in pauses and frame rate dips with hyperthreading enabled. You should be able to discern from the graphs that the hyperthreaded cores are acting quite erratically with sudden drops in thread utilisation. Even when I added the -cpucount=4 command to my hyperthreaded processor, erratic behaviour was observed.

So what does this mean to i7 users? The game definitely has issues with hyperthreading, and there is definitely room for improvement for i7 processor performance with this game. Bohemia Interactive just has to get around to optimising the game for HT.

Edited by choC

Share this post


Link to post
Share on other sites

Guys I gotta tell you that I just over clocked my CPU and I noticed a little bit of improvement, myself, I'm definitely, determined that ArmA II prefers Quad Core CPUs plus high GHZ CPUs since I was doing 3.0GHZ and pushed up (Factory Over clocking in BIOS) to 3.3GHZ and noticed little but stable increased in FPS without those suddenly drops in the city.

My rig:

AMD x2 Dual Core 6000+ @ 3.3GHZ OC

3.5GB DDR2 RAM

ATI HD 4890 1GB OC (940/1095)

Asus M3N72-D

Win XP Pro

Edited by Gringo85

Share this post


Link to post
Share on other sites

I don't know about the benchmark, but with hyperthreading on I tried loading up chernarus with a large amount of aircraft and it stuttered like crazy. Looking at task manager it seemed like only 1 thread (out of 8 threads) was running heavily, which probably shows ALL the AI use a single thread. Turning hyperthreading off obviously helped to reduce the stuttering by a significant amount as now that thread was running on its own core rather than 1/2 a core. I wouldn't be surprised if a no-AI scenario didn't show a difference with HT on vs off.

I don't know how many heavy threads the game runs in other situations, though. It would be nice to have some info from BIS regarding this topic, after all those who write the software should know how many threads they placed in it and how many of them can actually take noticeable CPU power to the point of slowing down your game.

In the end the first step towards multi-core optimization is spreading the load around with different threads. For example if the game only has 3 (heavy) threads that load up the CPU, then you're not going to see any real benefit going above 3 cores, and definitely won't see any benefit from HT.

Edited by galzohar

Share this post


Link to post
Share on other sites
i did dl procexp

but i have no clue how to use it and where to find the details you need

can also someone confirm my question with a answer

if device shows 8 cpu does cpuCount needs to be 8 or 4 on that i7

and its weird but disbaling HT is better for arma?

its confusing to me to disable performance technologies

BTW the latest ati drives increase and stabilize from alot the fps tho it seems ArmA locks fps at 60

Make sure you dont have vSync on in your video card settings in the software (ATI Catalyst or Nvidia's) since ArmA II has no setting for this unless in Arma2.cfg file 60 sounds like refresh rate issue to me so that's why I suggest you take a look into vSync, for me it seems like I'm capped at 30 FPS but after my CPU overclock I' had got to 35 when raining lol, wierd yeah.

Share this post


Link to post
Share on other sites

My system is using core 2 duo but for comparison purpose,

3246 at 2nd test.

Core2 Duo 6750@3.2 GHz

3.5 GB ram

4890 1GB at 1.0 + GHz

XP SP3

Arma2 on Raptor 74GB drive

Resolution: 1680x1050

I wonder it is really using Quad core..or just fast HDD.

---------- Post added at 02:44 AM ---------- Previous post was at 02:38 AM ----------

my settings might have been higher then demanded

but my results are

test 1 -35.1197

test 2 -39.854

test 3 -31.9574

test 4 -44.2478

test 5 -16.61

OFPMARK =3355.78

i run 8 core i7 cpu 940 @2.93Ghz

6MB triple channel ddr3

vista sp1 64bit

ati 4890 crossfire enabled X2

anyway when they say multicore suport it means it support as many cores you want

this post is ridiculous ..just cpuCount=X if the game cant read the cores

and other target parameters

mine looks like this -cpuCount=8 -maxmem=2047 -winxp -noCB -nosplash

Your score is not much better than mine....But you have i7 core and mine is Core 2 duo (OCed to 3.2GHz)

Looks like what BI said about quad or dual core before Arma2 is released is right. They said higher clock is better and quad is better in the same clock.

Edited by ArmedAnarkie

Share this post


Link to post
Share on other sites

Based on observed CPU load during different situations, it seems like the game is running 2 heavy threads for the actual game and then a 3rd thread for the AI, and that 3rd thread would go anywhere from very little use all the way up to using up an entire core 100% and slow your game down since your 3rd core can't keep up and the effort can't be split to the 4th due to it only being 1 thread.

Of course it would be very nice to get an official or at least more definitive information from someone who actually knows how to check how many threads a process is running and how much processing power each is taking.

Share this post


Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
Sign in to follow this  

×