PDA

View Full Version : Do you have a fast PC?



Duvel123
08-09-2010, 02:37 AM
If you have a reasonably fast PC, I could use your help. Im trying to gather some benchmarks for various systems.

If you want to help, just download my "Ceylon torture" test track here:
EDIT: Please do not link to modded content.
warning: seems to require UP!

Its a really stressful test, much more than "black death" with tons of objects, aaa, ships, lots of smoke and fire etc. Its pretty much a worst case scenario.

Here are some screenshots from the track:
<span class="flash-video"><embed type="application/x-shockwave-flash"
wmode="transparent"
pluginspage="http://www.macromedia.com/go/getflashplayer"
allowScriptAccess="never"
src="http://img29.imageshack.us/slideshow/smilplayer.swf"
width="426"
height="320"
name="smilplayer"
id="smilplayer"
bgcolor="FFFFFF"
menu="false"
flashvars="id=img29/grab0001s.jpg"
/> [</span>

Install fraps if you havent got it already, make sure you turn vsync OFF in your videodrivers. Disable all mods in JGSME (just default UP 2.01)

Preferred settings:
-No mods selected in JGSME
-resolution between 1280x1024 and 1600x1200
-AA and AF to ~2-4x
-Perfect settings, water = 2

You can stray from that as long as you are confident your videocard doesnt becomes a major bottleneck.

Then play the Ceylon torture track and bench it with fraps (hit F11 key to start).

Start benching when you are 5 seconds in to the track (look at the counter at the bottom). Ideally you would run the track twice and bench on the second run. I would need the "il2fb 2010 xx-xx-xx fps.CSV" file generated by fraps in c:\fraps\benchmarks as well as your system specs. You can mail it to duvel123 at gmail dot com. Ill make some nice charts if I get some useful results.

BTW dont worry too much about settings graphic options. On this track my PC (3.2 GHz C2D with radeon 4870) is completely CPU limited. The videocard is no bottleneck at all, despite perfect settings, water=2 and 8xAA/AF. Even if I clock the videocard down by 50% it doesnt impact performance one bit. Neither does running it without AA or perfect water or even in 800x600. CPU speed determines the performance completely, every over or underclock of the CPU changes my FPS pretty much linearly.

If your videocard isnt quite as fast or if you have a very high resolution monitor, run the track at a lower than usual resolution, or reduced AA settings to make sure the bottleneck isnt the videocard. Im looking for CPU performance here as that seems to be IL2 biggest bottleneck with todays graphics cards and Im particularly interested in results of Phenom IIs or Core iXs, although any CPU bound result would be interesting (also core2's with 4MB or more cache as mine is a 2MB model).

Thanks for the help, looking forward to see some results!

ytareh
08-09-2010, 05:56 AM
Very interesting project ,definitely appeals to the nerd in me .However it will be ,no offence intended ,pretty pointless if you dont get everyone to use standard graphics settings and ESPECIALLY screen resolution.To say that resolution doesnt affect fps is incorrect and similarly graphics settings (would need to specify at least AA/AF in Ati /Nvidia software and also IN GAME settings eg Perfect water is an fps killer)are very relevant.
There was an attempt to do something like this before using the Black Death track .
Will definitely post some figures here in next day or two...

Duvel123
08-09-2010, 06:13 AM
Originally posted by ytareh:
Very interesting project ,definitely appeals to the nerd in me .However it will be ,no offence intended ,pretty pointless if you dont get everyone to use standard graphics settings and ESPECIALLY screen resolution.

Thats what I also thought.
Until I tested it with decent videocard.

800x600, no AA, excellent setting; videocard at full speed

compared to

1440x900, 8xAA, perfect water=2 with my videocard running at *half* speed.

Youd expect enormous differences right? I would. I measured NONE. Check some charts here:
http://www.battle-fields.com/c...owthread.php?t=21901 (http://www.battle-fields.com/commscentre/showthread.php?t=21901)

I didnt chart everything I measured as the result was always identical; so much the graphs overlapped you couldnt see more than a single line. The only thing that mattered was CPU speed. Unless I set resolution to 1920x1080 (more than my monitor, running in a window) and 8xAA/AF then a small slowdown was measured as my videocard hit its limits.

With this track (and in fact with most tracks) IL2 is *much* more CPU limited than I would ever have believed. At least with my core 2, even at 3.2 GHz, I am 100% CPU limited. Maybe faster CPUs will manage to shift the bottleneck back to the videocard, and certainly with my old videocard (7900GS) resolution, AA, water=2 etc made a big difference but with a modern videocard, I bet you are mostly or entirely CPU limited. Resolution, AA and detail settings wont change a thing, surprisingly enough.

ytareh
08-09-2010, 06:46 AM
Heres some old data for Black Death Track

I ALWAYS use 1024x768 as dots are invisible above that in combat I find.Similarly I dislike Perfect water as I find it hard to judge how far away it is !

E8500@4.066 ATI Cat 8.9 4870 Stock 750/900

AVG/BEST/WORST
99/248/30

E8500@4.066 ATI Cat 8.9 4870 OC 829/999

100/254/29

OVER AND OVER AGAIN WE SEE THAT THE GRAPHICS CARD IS DARN NEAR IRRELEVANT IN IL2-THIS INFO SHOULD BE INCLUDED IN A STICKY!

***

Heres a 4870 at stock speed with CPU speed changed

E8500 @3.166GHz 66/176/20
E8500 @4.066GHz 87/235/26

AND HERES THE SAME CPU AT 3.166 BUT WITH A 4870X2

E8500 @3.166GHz 66/169/19 (A BIT SLOWER THAN BASIC 4870!!!!)

4870X2 needs Cat AI enabled in CCC to run both GPUs

IL2 wont run (mission loading error ) in Cat AI Advanced Mode

in "Dual GPU Mode"

E8500@3.166GHz 61/164/18 (SLOWER AGAIN!-Game NOT Crossfire optimised )

E8500@4.066 4870X2 Cat AI STANDARD

80/220/24

***


Heres some figures comparing res and aa/af changes

BD Track @2560x1600 Perfect Landscape -Application Controlled Settings E8500@4.066 4870X2
56/148/18

As above with maxxed AA/AF
45/131/18

***

In a quick test 4x1gig ram sticks gave 2fps min speed more than 2x1gig -same cheap and nasty sticks!

***

In another quick test swapping a GTX95 for an 8800GTS 640mb gave the EXACT same min fps (38)

***

PROCESS AFFINITY CORE Testing

DISABLED 141/360/52 142/369/51
ENABLED =1 136/356/51 137/352/51
ENABLED =0 142/374/53 142/365/53

so set core =0 SEEMS to be the winner there for dual core C2D

***
GTX295 @STOCK 137/361/17 138/360/17
@695/1498/1196 OC 133/345/16 133/348/16
@498/1073/899 UNDERClock 110/277/16


Hope this is of SOME interest ...

Will download track and get to work on Project Ceylon asap!

Duvel123
08-09-2010, 07:12 AM
Your results on blackdeath match what i would have thought and said, cpu cpu and cpu speed is what matters as long as you keep resolution and AA settings below ~ HD resolution (depending on videocard).

Looking forward to your results on the ceylon track. Especially if you could run it with the 4870 (keeping resolution and AA "sane", so lets say no higher than the 1440x900 8xAA I used although a bit more shouldn hurt as at that resolution half or double the videocard speed didnt matter). Then we have the same videocard very similar cpu's but yours has 6 MB cache, mine has 2. Something tells me thats going to help a lot.

Also rather than providing min/max/avg, I like to compare the FPS csv file and chart it as it says a lot more and you can easily spot anomalies (pressing F11 too soon or too late and you can get a very different min/max result) and having 200FPS where another system scores "only" 150 (or even 60 with vsync) changes max and average by a lot but changes nothing to the experience.You can see that on charts but not by looking at those 3 figures.

edit: one more thing about core affinity; at least in windows 7 it doesnt matter. In XP there was a benefit setting it to either 0 or 1 as XP would hop the thread between cores; but thats no longer the case in 7 (and I think vista). Wether you set it to 0 or 1 really makes no difference. Anything you'd measure is measurement error, as one core is not faster than the other (assuming you dont have another thread locked to a core).

ytareh
08-09-2010, 07:18 AM
I have a GTX295 EVGA Coop Ed (single PCB) (Had an earlier Dual PCB -this one runs much cooler).Only kept my 4870X2 for a few weeks was a bit of a loud oven !4870 was sweet card but didnt bring THAT much extra to the game over 8800 series (had all flavours...) X1950 was last gpu that mattered for IL2!!!

ytareh
08-09-2010, 07:48 AM
Just sent you my first set of results ...no wonder youre looking for "FAST PC's "! That track would melt ,if not set on fire older pcs!!!Its awesome but a bit long at 5 min for benching.Anyway your 30fps challenge has been reached by my pc with cpu underclocked by over 50%.Forgot to mention in email all testing at 1024x768 as usual...Just about to run on 4.2 OC now.I dont think its going to be very practical to do it on a SECOND five minute run ,would suggest you remove that step...

Duvel123
08-09-2010, 08:02 AM
The challenge was *maintaining* 30 FPS, not averaging it http://forums.ubi.com/groupee_common/emoticons/icon_smile.gif. And I dont think it will be done this decade http://forums.ubi.com/groupee_common/emoticons/icon_smile.gif I got your results; thx, ill process them in a bit.

As for the 5 minute length.. the longer, the more accurate http://forums.ubi.com/groupee_common/emoticons/icon_smile.gif. Besides I recorded that track as part of a movie I planned on making, just thought it would make a good torture track as well as I had a hard time frapsing it at good FPS http://forums.ubi.com/groupee_common/emoticons/icon_razz.gif

ytareh
08-09-2010, 08:16 AM
No NOBODY is gonnna keep MINIMUM 30fps in this track!(unless MAYBE at 600x480 windowed etc !)Have sent my 60fps avg 4.2 run and now doing it at 1680x1050 which I expect will be a fair bit slower ...interesting to see...Only 30 secs to go so may as well post them ...

RESULTS SO FAR

Ceylon Track -Latest Nvidia Driver all default settings apart from x32Q AA supersampling and VSync off


E8500/GTX295 (overclocked at 672/1449/1166)


@2GHZ MIN/MAX/AVG 7/89/32.5
@4.2 GHz 8/165/59.8 1024x768
@4.2GHz 12/163/43 1680x1050

Note big drop in fps at higher res and strange INCREASE in minimum fps !I started at bang on 5 seconds in both cases.

Duvel123
08-09-2010, 08:22 AM
Dont read too much in to the min FPS without looking at the charts. That minimum FPS happens when you see the zero pull up and the bombs fall on the base. It only lasts a very brief moment, you get 0 FPS for like half a second aand can give some funny readings.

Im very surprised at your drop in FPS at higher res though. And shocked how much faster your system is than mine overall. I gotta find out if that is due to the L2 cache or nvidia drivers. Ill see if I can borrow a fast enough nvidia card to bench with.

edit: oh.. 32xAA. No wonder your framerates drop. There you are pushing your videocard a bit too much and its impacting your framerates. Set it to 8x AA and I think you will get the same result as at 1024x768 with whatever AA setting.

mortoma
08-09-2010, 11:00 AM
I played at 1920X1080
Water=4
Effects=2
Forest=2
Sky=2

Nvidia Control Panel settings:
Aniso Filtering = 16X
AA gamma correction = on
AA mode = override any application setting
Antialiasing setting = 16xQ
AA transparency = Off
Conformant texture clamp = Use Hardware
Error reporting = Off
Extension Limit = Off
Force mipmaps = None
Maximum pre-rendered frames = 3
Multi-display/mixed-GPU acceleration = Multiple display performance mode
Texture filtering - Anisotropic sample optimization mode = Off
Texture filtering - Negative LOD bias = Clamp
Texture filtering - Quality = Quality
Texture filtering - Trilinear optimization = On
Triple Buffering - Off
Vertical sync = Force off

System:
Intel i7 965 ( overclocked to 4.020Ghz )
EVGA Nvidia 285GTX ( overclocked; core clock=675, shader clock=1537, memory clock=1290 )
6gb Corsair PC3-8500F

Frames:
Min. = 23
Max = 224
Avg. = 75.223

Duvel123
08-09-2010, 11:27 AM
mortoma, could you mail me the CSV files please?

voyager_663rd
08-09-2010, 11:33 AM
I don't have UP.

Is that why I get the error message "no 634 sqd" (from memory, not written down) when I try to run it?

Duvel123
08-09-2010, 11:38 AM
Yes. I made the map with ultrapack and used some planes and squadrons not available in the stock version. I would have avoided that if I had set out to make a benchmarking track; but it was raw material for a movie project..

Here is the prelude for that movie, complete with horrible AF shimmering on my old nvidia card (nvidia users, set AF to default for IL2 and not to something high as it looks horrendous:
<span class="flash-video"><object classid="clsid:d27cdb6e-ae6d-11cf-96b8-444553540000"
codebase="http://fpdownload.macromedia.com/pub/shockwave/cabs/flash/swflash.cab#version=8,0,0,0"
height="385"
width="640"
><param name="wmode"
value="transparent"
></param><param name="allowScriptAccess"
value="never"
></param><param value="http://www.youtube.com/v/9H2WlD0oTCk&hl=en_US&fs=1"
name="movie"
/><param value="true"
/><param value="always"
/><embed wmode="transparent"
pluginspage="http://www.macromedia.com/go/getflashplayer"
allowScriptAccess="never"
type="application/x-shockwave-flash"
height="385"
width="640"
src="http://www.youtube.com/v/9H2WlD0oTCk&hl=en_US&fs=1"
/></object></span>

And this is well.. the movie that I didnt make:
<span class="flash-video"><object classid="clsid:d27cdb6e-ae6d-11cf-96b8-444553540000"
codebase="http://fpdownload.macromedia.com/pub/shockwave/cabs/flash/swflash.cab#version=8,0,0,0"
height="385"
width="640"
><param name="wmode"
value="transparent"
></param><param name="allowScriptAccess"
value="never"
></param><param value="http://www.youtube.com/v/FzpMCtgGQ60&hl=en_US&fs=1"
name="movie"
/><param value="true"
/><param value="always"
/><embed wmode="transparent"
pluginspage="http://www.macromedia.com/go/getflashplayer"
allowScriptAccess="never"
type="application/x-shockwave-flash"
height="385"
width="640"
src="http://www.youtube.com/v/FzpMCtgGQ60&hl=en_US&fs=1"
/></object></span>

mortoma
08-09-2010, 03:53 PM
Originally posted by Duvel123:
mortoma, could you mail me the CSV files please? I might but I have no idea what a CSV file is!

Duvel123
08-09-2010, 03:58 PM
csv= comma seperated values. its a file format, like a simple excel file.
Fraps generates 3 such files in the benchmark folder (one you probably opened to see min/max/avg). Id need the one called "il2fb 2010 xx-xx-xx fps.csv" (the extension .csv might be hidden on your system)

BillSwagger
08-09-2010, 04:06 PM
ever hear of PassMark?

I used it a few months ago to see how my computer compared to similar product lines.
Its a great way to know if your computer is running as it should.
In doing some reading, i also discovered that a 64bit OS can actually return better results than a 32bit OS system, possibly because it is more efficient in some ways.
have you ever heard of upgrading the OS from 32 to 64 bit, improving the performance?
It probably depends on the application.

bill

Treetop64
08-09-2010, 04:25 PM
...make sure you turn vsync OFF in your videodrivers.

For ATI users: Leaving VSync ON at full quality, with "Triple buffering" clicked ON, will yield smoother performance in OpenGL with no hit on frames.

mortoma
08-09-2010, 04:34 PM
BTW, I never remember putting my conf.ini water setting to = 4!! I didn't even know that there was a setting above 3? Anyone an expert on the water setting in conf.ini and knows what the settings mean and how high it should be set? My water does look nice and has that nice shiny sheen to it in OpenGL. But I always thought that water settings were restricted to 1, 2 and 3 only.

mortoma
08-09-2010, 04:43 PM
Originally posted by BillSwagger:
ever hear of PassMark?

I used it a few months ago to see how my computer compared to similar product lines.
Its a great way to know if your computer is running as it should.
In doing some reading, i also discovered that a 64bit OS can actually return better results than a 32bit OS system, possibly because it is more efficient in some ways.
have you ever heard of upgrading the OS from 32 to 64 bit, improving the performance?
It probably depends on the application.

bill 64 vs. 32 bit does not seems to make any difference in IL2-1946 at all. I have Win XP-32 and Win 7-64 and when I play the BDT, I get slightly better frames with Win XP-32. It probably makes a huge difference with other software and games but IL2 is a 32 bit game and has only one thread. I can't play the game in Win 7-64 ( no USB joystick ) but I can watch tracks and catch frames from them and therefore can evaluate it's performance against Win XP-32.

Treetop64
08-09-2010, 04:44 PM
Max water settings used are specific to which video card brand you have installed. I think it's:

2 for ATI
3 for NVdia

Is there a Water=4 setting?

mortoma
08-09-2010, 04:47 PM
Originally posted by Treetop64:
Max water settings used are specific to which video card brand you have installed. I think it's:

2 for ATI
3 for NVdia I wonder why mine's at 4 then? I didn't put it on 4 but it's working like 3 because my water looks shiny and gorgeous.

Urufu_Shinjiro
08-09-2010, 05:00 PM
Originally posted by Treetop64:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">...make sure you turn vsync OFF in your videodrivers.

For ATI users: Leaving VSync ON at full quality, with "Triple buffering" clicked ON, will yield smoother performance in OpenGL with no hit on frames. </div></BLOCKQUOTE>

Originally posted by Treetop64:
Max water settings used are specific to which video card brand you have installed. I think it's:

2 for ATI
3 for NVdia

Is there a Water=4 setting?

For benchmarking purposes always leave vsync OFF. For playing certainly, vsynch and triple buffer, but it will mess with benchmark numbers.

As for water=4, it was introduced in one of the patches along with improved ATI water=2, so now it is ATI has 0,1,2 (the new two being more like the old nvidia 3) and nVidia 0,1,2,3,4.

mortoma
08-09-2010, 05:05 PM
Weird, I just went down to Aniso 8X and AA 8XQ and my min. frames went down a bit from what I posted earlier. Wow, I guess I'll stick to my insanely high 16X and 16XQ then! My max. and avg. went up some but not much.

Frames:
Min. = 17 ( before was 23 )
Max = 234 ( before was 224 )
Avg. = 78.04 ( before was 75.223 )

<span class="ev_code_RED">POST EDIT: Never mind, I should have run the track at least twice this last time. From my experience you should run a track at least once to warm up your video card before benchmarking with FRAPS. The second or third time a track will run smoother from my experience. The first test I did I had run it twice before I benchmarked because I wanted to see how long the track lasted.</span>

Treetop64
08-09-2010, 05:10 PM
Originally posted by Urufu_Shinjiro:

For benchmarking purposes always leave vsync OFF. For playing certainly, vsynch and triple buffer, but it will mess with benchmark numbers.

As for water=4, it was introduced in one of the patches along with improved ATI water=2, so now it is ATI has 0,1,2 (the new two being more like the old nvidia 3) and nVidia 0,1,2,3,4.

Good to know. Thanks.

ytareh
08-09-2010, 07:08 PM
So how do ya convert these files into graphs ?

WTE_Galway
08-09-2010, 07:52 PM
Originally posted by mortoma:
64 vs. 32 bit does not seems to make any difference in IL2-1946 at all. I have Win XP-32 and Win 7-64 and when I play the BDT, I get slightly better frames with Win XP-32. It probably makes a huge difference with other software and games but IL2 is a 32 bit game and has only one thread. I can't play the game in Win 7-64 ( no USB joystick ) but I can watch tracks and catch frames from them and therefore can evaluate it's performance against Win XP-32.

I have not tested it but suspect the main difference from the IL2 point of view will be addressing more ram, if you had win 7-64 with 6Gb of ram you might see a difference.

mortoma
08-09-2010, 09:33 PM
Originally posted by WTE_Galway:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by mortoma:
64 vs. 32 bit does not seems to make any difference in IL2-1946 at all. I have Win XP-32 and Win 7-64 and when I play the BDT, I get slightly better frames with Win XP-32. It probably makes a huge difference with other software and games but IL2 is a 32 bit game and has only one thread. I can't play the game in Win 7-64 ( no USB joystick ) but I can watch tracks and catch frames from them and therefore can evaluate it's performance against Win XP-32.

I have not tested it but suspect the main difference from the IL2 point of view will be addressing more ram, if you had win 7-64 with 6Gb of ram you might see a difference. </div></BLOCKQUOTE>I do have 6Gb of ram! Exactly that much. But really I don't think IL2 uses much more than 1Gb of ram. Remember, this game was coded back in the late nineties and early 2000s. If a PC had a half gig back then it was a big deal. IL2 had to be developed for that period of time and it's memory usage has not changed much, got bumped up a bit with the intro of Forgotten Battles about '03 but that's about it. Some mods exist to change memory usage but that's something different, I'm talking about the default un-modded game.

Duvel123
08-10-2010, 12:19 AM
Originally posted by ytareh:
So how do ya convert these files into graphs ?

Yes I will and I will post the findings here; just give me a day or so to do it properly and make sure the results are actually meaningful. Im also waiting for a few AMD results.

As for the RAM amount; more doesnt help, once you have enough to keep windows from swapping. For XP thats probably 2 or 3 GB RAM depending how much bloat you have installed, for 7 its 3 GB. I tested 3 vs 4 GB ram on that torture test and found no difference at all:
http://www.battle-fields.com/c...owthread.php?t=21901 (http://www.battle-fields.com/commscentre/showthread.php?t=21901)

(post 10). More wont hurt, especially if you like to keep lots of apps open in the background, but dont expect it to increase your FPS.

As for Passmark.. its a decent general benchmark, but it doesnt show how good a PC is for IL2.

Thanks to all who sent me results so far.

PF_Coastie
08-10-2010, 04:34 AM
I would be glad to give you an AMD965 result, But do not have UP....only HSFX.

This test should be done with stock game. Too many "other" variables with mods.

rfxcasey
08-10-2010, 08:52 AM
Originally posted by mortoma:
Weird, I just went down to Aniso 8X and AA 8XQ and my min. frames went down a bit from what I posted earlier. Wow, I guess I'll stick to my insanely high 16X and 16XQ then! My max. and avg. went up some but not much.

Frames:
Min. = 17 ( before was 23 )
Max = 234 ( before was 224 )
Avg. = 78.04 ( before was 75.223 )

<span class="ev_code_RED">POST EDIT: Never mind, I should have run the track at least twice this last time. From my experience you should run a track at least once to warm up your video card before benchmarking with FRAPS. The second or third time a track will run smoother from my experience. The first test I did I had run it twice before I benchmarked because I wanted to see how long the track lasted.</span>

Duh... http://forums.ubi.com/images/smilies/16x16_smiley-wink.gif

Oh and I think IL-2 adds +1 to the water setting if your system is liquid cooled...... http://forums.ubi.com/images/smilies/halo.gif

But I regress, you can set effects to 0 from 1 in your OpenGL settings in the conf.ini. You will not see a visual difference but the fire and explosions will create less of a performance hit slightly. I also found that the game actually runs better when I leave the processor affinity mask commented out with a ;. Utter then that make sure your HD is defragged as any hickup in disk access with skew your benchmark results me thinks.

mortoma
08-10-2010, 10:53 AM
Question for Duvel123. Do you really have to have us send you our Fraps files? Couldn't you just open up one of your own and type in our data as a substitute? Seems to me it would be the same thing. There's really not much to those Xcel files except the frames data written in them. Just asking, but I could still send the data if you need it.

Duvel123
08-10-2010, 11:19 AM
Originally posted by mortoma:
Question for Duvel123. Do you really have to have us send you our Fraps files? Couldn't you just open up one of your own and type in our data as a substitute? Seems to me it would be the same thing. There's really not much to those Xcel files except the frames data written in them. Just asking, but I could still send the data if you need it.

Those framerates are what matter http://forums.ubi.com/groupee_common/emoticons/icon_smile.gif

So yes, I would really like the entire file. min/max/avg is useful but it doesnt tell the whole picture, especially not when there are already so many other variables. Its also an assurance no mistake was made, as I plot FPS over time I can "shift" the values so they match perfectly in time and i would have to see correlation with the other results. some people sent me results that didnt correlate and it turned out they sent me the wrong benchmark result and one didnt have the required mods, resulting in a ton of objects not being rendered. Installing UP cut his framerate in half on that track.

Anyway if you dont want to send me an email, upload it to mediafire.com and post (or PM) a link.

Also.. im getting a few results but hardly anyone follows the instructions I posted. When you enable "effects_low" in UP that does more to your framerates than any overclock could achieve rendering any comparison void (or forcing me to rebench all my results with those same settings) Likewise (but opposite) for 32xAA/AF at uber high res.

Duvel123
08-10-2010, 01:13 PM
Okay, some stuff I did so far.

First lets look at the impact of resolution and AA settings as most benchmarks I received have different settings. Tested with my rig

Core2duo 3.2 GHz
Radeon 4870 512 Mb stock speed (750/900)
UP 2.01 default settings
perfect, water=2
effects=1


http://img412.imageshack.us/img412/842/11942106.png

Those results look funny to me. On the right side, there is exactly what youd expect. At ultra high res and especially in combination with high AA/AF settings; my videocard collapses and performance plummets. I suspect it would do so in any game.

From 1600x1200 down to 1280x1024 there is virtually no performance difference whatsoever, even regardless of AA/AF settings used. That would hint at the videocard being idle most of the time (which is the case according to catalyst, showing anywhere between 0 and 25% "load") but at even lower resolutions there is a weird performance boost. I cant really explain this.

I looked a bit further and charted FPS over time for a few of the above averages:

http://img534.imageshack.us/img534/3910/17165994.png

As you can see, the slowdown is across the board. Its not caused by FPS spikes up or down in particular area's.

So i figured Id test if my videocard really was no bottleneck. I clocked my videocard down from 750/900 to 500/450 at a resolution and AA setting where there is a clear slowdown (compared to 800x600):

http://img713.imageshack.us/img713/7371/55057597.png

No difference whatsoever. My videocard really isnt the bottleneck so something must cause an increased load on the CPU (or unlikely some other component?) as you increase resolution above 800x600. Must be an IL2 engine peculiarity ? Anyway the difference really isnt very big at all but its consistent.

So I had another thought.. those resolutions have different aspect ratio's (width/height). That impacts the FoV (field of view) in IL2. With a square resolution you get a much wider FoV than with a wide resolution. As a result, different things get rendered on any given track. To show this I took it the extreme and benched 900x900 and compared it to 1800x450. Its the exact same amount of pixels, so the same load on your videocard but the track looks entirely different obviously; and so does the performance:
http://img411.imageshack.us/img411/1611/18812148.png

The (ultra) wide aspect is a shocking 33% faster on average. I suspect that depends entirely on the track (eg flying straight or 90 degree bank might well reverse those results), but found it interesting nonetheless. Of course the resolutions we actually use arent nearly as far apart, ranging from 4:5 (or 16:20) to 16:9. Still, a 16:9 setup is probably a bit faster than a 16:10 or 4:3 display for any given amount of pixels. You never thought of that before did you ? http://forums.ubi.com/groupee_common/emoticons/icon_smile.gif

Lets move on to some CPU scores. I was dead curious to see how a core ix would fair. And it does brilliantly:

http://img192.imageshack.us/img192/3720/87325236.png

Our videocards are rather similar. That is, a 4890 should be around 10% or so faster than my 4870 if the videocard is the bottleneck but at these resolutions and AA settings Ive already shown you the GPU barely matters. Its all CPU driving the performance.

http://img34.imageshack.us/img34/5530/74516171.png

At the same clock, the Core i7 is almost 50% faster than my overclocked Conroe. Stunning. Even more so if you consider other tests I did that showed that IL2 doesnt care one bit about memory bandwidth (core i7 has spades of that compared to core 2s). I wonder if its the cache? The lower memory latency?

Overclocked to 4 GHz the i7 goes on to humiliate my scores by nearly doubling them. I have yet to see any AMD phenom scores, but you can already guess whats on my shopping list http://forums.ubi.com/groupee_common/emoticons/icon_smile.gif

The next comparison I wanted to see if there was a difference between a C2D with 2 Mb cache and one with 6. Unfortunately the test is skewed somewhat due to very different videocard, and UP settings were used that I think make no sense (uranium effects mod enabled but combined with effects_low - which resets all effects again to default). Anyway, Im taking this one with a grain of salt:

http://img205.imageshack.us/img205/7047/75936027.png

http://img375.imageshack.us/img375/3420/84063297.png

The combination of 6MB cache and a nVidia card does make mincemeat of my setup. I just wish id know which of both causes that. Could it be that even though at these settings our videocards arent the real bottleneck that nVidia drivers put less burden on the CPU or something? Lets see;

http://img814.imageshack.us/img814/4033/55428521.png

Oh yes. That could well be it. Two comparable 4 GHz Core i7 systems with rather comparable settings but the nvidia machine is indeed considerably faster, even though neither card would break a sweat at those settings. Some more datapoints would be good here, but so far it looks as if an ATI videocard puts an extra burden on your CPU, probably less optinmized OpenGL drivers. Well, thats a guess.

Ill be adding charts and data here later as they come in (AMD guys where are you ?), but my ad interim conclusions are clear:

1) Core iX 's rock. At least i7s, and the i5's with IMC will likely not be very different. Core i5's without IMC and i3 is anyone's guess at this point.
2) nVidia is what you want for IL2. We all knew that already, because of better drivers but I hadnt expected this kind of performance lead for cards that generally are no faster.
3) You can never have too much CPU speed for this game. Ever.

Thanks to those who sent results, I will add due credits later.

ytareh
08-10-2010, 01:34 PM
Interesting findings ...looks like an upgrade from C2D to "i" may be in my future !

Duvel123
08-10-2010, 01:57 PM
Id wait for a few AMD results. Either K8 or phenom would be interesting as both have integrated memory controllers (therefore low latency memory access) like the Core i7 and unlike my C2D.

I dont know if that makes the bulk of the difference but it wouldnt surprise me. L2/3 Cache could be a factor too.

mortoma
08-11-2010, 07:11 PM
Duvel, you say you want people to try different resolutions but what about when you have a LCD monitor with a native resolution? They say you aren't supposed to change. I play only at 1920 X 1080 for that reason.

WTE_Galway
08-11-2010, 07:37 PM
Originally posted by Duvel123:

1) Core iX 's rock. At least i7s, and the i5's with IMC will likely not be very different. Core i5's without IMC and i3 is anyone's guess at this point.



My impression is that if the ONLY game you played was IL2 the i3 530 would be right up there with the more expensive CPU's as its limitations (only dual core, no turbo boost, less memory bandwidth etc) are not that significant to IL2 and it has reputation for a phenomenal 4Ghz overclock.

.. and no I have not done any testing http://forums.ubi.com/groupee_common/emoticons/icon_biggrin.gif

willyvic
08-12-2010, 09:23 PM
Hate to pull the plug guys but linking to and benchmarking a modded install won't do. If you want to benchmark and discuss a vanilla install then go for it. In the interim this thread is closed.

WV