Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Porksmuggler]

I did some more searching/testing and found some interesting answers.  I can't take credit, but CrossFire will "work" without a crash to desktop if you set AA to 16X in CCC.  CrossFire still crashes to desktop at 100% loading screen for any other setting, 2X-8X.  This thread tipped me off, it seems to be a driver bug:
http://forums.ubi.com/eve/forums/a/tpc/f/49310655/m/3681049478#3681049478

Testing confirms what PA_Willy posted, there isn't really any CrossFire scaling, but I couldn't find any real SLI scaling either (Please post if you have links to SLI scaling for IL-2).

TheBlackDeath.trk using in-game fps SHOW START, conf.ini settings from previous post, AVG/MAX/MIN:

71/142/22 CrossFire enabled 16X AA
71/142/22 CrossFire disabled 8X AA (single GPU doesn't allow 16X AA)
81/184/22 CrossFire disabled no AA
54/70/22 All settings with v-sync enabled

I run with v-sync enabled, so now I don't have to remember to disable CrossFire to avoid a crash to desktop.  I just have to remember to turn off 16X AA before playing other games, oh joy!

[cheruskerarmin]

@messer

THX for the advice but altering the adaptive AA in CCC didn't change anything. Still got the problem with contrails/disappearing props/glass parts. I like to play US-fourmot interceptions in realistic altitudes very much, therefore i have to see this issue quite often. However 10.8 works fine with all other games and programs on my rig. Maybe any other
advices?

Vertex arrays are working fine meanwhile, no need to disable them since 10.7.
Water = 2 ist the best (!) ATI setting, no need to set it to 3 or 4, which are NVIDIA related settings only! 3 or 4
doesn't change anything with ATI cards.
The game is NVIDIA optimized, therefore it is important to disable all 'nv...' settings in confini.

Are there any windows 7/64/5870 users who can check if they got the contrails/prop disc/glass issue? I can email you the mission files for testing. Would be very helpfull to know if i am the only one with this issue or if maybe just using another ones confini is getting this solved...

Thank you very much and all the best

[PA_Willy]

Testing confirms what PA_Willy posted, there isn't really any CrossFire scaling, but I couldn't find any real SLI scaling either (Please post if you have links to SLI scaling for IL-2).

The IL2 engine is scaling with Nvidia-SLI from day one:





http://www.xbitlabs.com/articles/video/display/gf6800u-sli_22.html#sect1

It's not a secret: if you play IL2 the most, you should buy a Nvidia card. Less troubles and better compatibility in OpenGL.

[Porksmuggler]

Are there any windows 7/64/5870 users who can check if they got the contrails/prop disc/glass issue? I can email you the mission files for testing. Would be very helpfull to know if i am the only one with this issue or if maybe just using another ones confini is getting this solved...

I've seen other threads mentioning issues with the 5 series I'm not seeing on the 4 series, I personally only had a texture corruption issue because of Steam overlay on my digital version.  I linked to an ATI thread here that might help you (first link):
http://forums.steampowered.com/forums/showthread.php?t=1358851

The IL2 engine is scaling with Nvidia-SLI from day one:

It's not a secret: if you play IL2 the most, you should buy a Nvidia card. Less troubles and better compatibility in OpenGL.

Thanks for the charts...I'm not disputing SLI works, it's more the question is there any real scaling with modern hardware.  It's good to know a pair of 6800 Ultras scale at 1600x1200 with 4XAA, but the results are obviously much lower than a single 4850 at 1920x1080 with 8XAA.

The only charts I found for modern hardware showed no real SLI scaling (the only comparison is one GTX295 vs two in quad SLI, 129fps vs 140fps at 1920x1200 and 109 vs 111fps at 2560x1600)
http://www.overclockersclub.com/reviews/crossfire_vs_sli/11.htm

I did find this IL-2 testing which is nice:
http://www.battle-fields.com/commscentre/showthread.php?t=21804

Does anyone have links for SLI scaling on somewhat modern hardware, say at least 8800GT or above?

[PA_Willy]

Quad SLI is rubbish. Everybody know that games don't scale correctly with 4 GPU,s. I have got a GTX295 in my rig. I know what I'm saying. But, with a 295 you can disable SLI. One only GPU is working, so you can compare results (GTX295 are two GPU,s in internal SLI).

[Porksmuggler]

Quad SLI is rubbish. Everybody know that games don't scale correctly with 4 GPU,s. I have got a GTX295 in my rig. I know what I'm saying. But, with a 295 you can disable SLI. One only GPU is working, so you can compare results (GTX295 are two GPU,s in internal SLI).

If you mean quad SLI is rubbish for IL-2, I agree   I know how the GTX295 functions, quad SLI works well depending on the game:
http://www.guru3d.com/article/geforce-gtx-295-quad-sli-gaming-test/1

Anyway this is what I'm looking for, will you run the TheBlackDeath.trk using in-game fps SHOW START with and without SLI enabled?

[PA_Willy]

Anyway this is what I'm looking for, will you run the TheBlackDeath.trk using in-game fps SHOW START with and without SLI enabled?

This track is not useful to compare SLI preformance because is hardly CPU bottlenecked. Where you can see the performance boost is in blind skies with AA and AF maxed. I have a quick mission on Slovakia that I use (with FRAPS) to measure the CPU, GPU and drivers variations. Believe me, this game is CPU-bottlenecked. In The Black Death track, CPU is the brake.

I have got a G15 keyboard, and I check inlive temps and percentage of use of GPU,s, CPU, RAM and VRAM. I see when CPU is at 100% limiting the GPU,s work, which are near 30-40% each one. The interesting benchmark would be one in with a very high VGA demand. For example, 16AA, 16AF, HQ textures, no ST3C compression, blind clouds, etc.

EDIT: quad SLI is the worst option to consider. Temps, power consumption, price, and later, a lot of games are not getting benefits in performance with it. In your link, of 5-6 games tested, two don't get benefits. Very bad idea.

[Porksmuggler]

TheBlackDeath.trk is the standard track for testing, will you max AA and the conf.ini as I have done and test with and without SLI enabled?  Please?

None of the testing above is concluding whether CrossFire or SLI scaling or CPU bottlenecking is the issue for modern hardware.  You're claiming CrossFire doesn't scale at all with modern hardware, but not mentioning the CPU bottleneck is an issue when it's CrossFire.  Then claiming SLI works perfectly for modern hardware, but the CPU bottleneck is an issue...

[PA_Willy]

Here you have my test (using FRAPS).

Firstly, my rig:

- C2D E8400@4.15Ghz (overclocked)
- Gigabyte X48DS4
- 4Gb RAM DDR2 1100 (Gskill)
- GTX295@700-2100 (overclocked)
- LG 30" playing at 2560*1600 (native resolution).

About the test:

- SLI mode is in AFR1 way (alternative frame rendering).
- UP. 2.01 enabled with other mods installed (HG & P Plutonium, Flushy clouds...for example).
- Effects=1
- Water=2
- AAx4 (because the high resolution employed) and AFx16. High Quality textures without any optimization. No ST3C compresion in game.
- Vsync off (of course).
- Track: TheBlackDeath.ntrk. Completed.

Result in SLI: Time: 155793ms - Avg: 83.958 - Min: 27 - Max: 185

Result with one only GPU: Time: 155914ms - Avg: 63.381 - Min: 29 - Max: 145

Minimum frames are lower in SLi (and CF) because SLI spends CPU performance to work.

Later I will test in SFR (Split frame rendering) and I will repeat the test with AAx16 (excessive at 2560*1600... only for testing purpose)

Regards

[PA_Willy]

Update in SFR way.

Result in SLI: Time: 155838ms - Avg: 81.065 - Min: 28 - Max: 206

Next --> Antialiasing x16 (multisampling)

[Porksmuggler]

Thanks for those test, I'm really just interested in the 16XAA with and without SLI.  The 6800Ultra SLI you linked increased the Avg. almost 100% over the single card, obviously not a CPU bottleneck for the older hardware, so I'm curious how much SLI is scaling with modern hardware.

[PA_Willy]

Same parameters with AAx16 at 2560*1600:

SLI in SFR way: Time: 155817ms - Avg: 79.670 - Min: 28 - Max: 197

Single GPU: Time: 155853ms - Avg: 59.575 - Min: 29 - Max: 130

Several considerations/recommendations:

- AFR way is the standard (Nvidia). BUT, I recommend SFR (Splt frame rendering). Why? Because in AFR I feel stutters (microstuttering). If you rise AA, more stutters you will feel in AFR. In SFR all is fine. Only in this game.
- With water=4 I feel stutters too IF I USE AF defined by me in nHancer (the program I use to enable/disable the different parameters and SLI modes). But I can use water=4 perfectly if I leave AF= defined by application (maybe any kind of incompatibiliy).

Regards.