Anybody else with a GTX 580 having performance issues?

[[CW]Aresye]

I'll post my graphics settings later on for in-game and the nVidia control panel, but the quick and dirty run down is:

Built brand new computer in May:
Sandy Bridge 2600K OC'd to 4.5GHz
Ripjaw Series RAM - 8GB
nVidia GeForce GTX 580
7200RPM HDD - 1TB

Current Settings (Short Version):
Running CCC Lvl 4
Phys CPU Set to 3
1920x1080 Resolution
Force DX9
4X AA

My frame rates never seem to average above 40FPS, with some maps running consistently in the low 20s (Death Valley).  I've tried running w/ and w/o CCC, w/ and w/o AA, and various other settings to try and get this game working at playable frame rates without sacrificing all the visual quality, but right now it seems the only way to get decent frame rates is by lowering all the graphics and/or resolution so far it looks like complete crap, and then it's only maybe a 20fps increase.

Just wondering if anybody's been in a similar boat, or perhaps have some settings that work well with this kind of setup?

[sleepysheep]

I'll post my graphics settings later on for in-game and the nVidia control panel, but the quick and dirty run down is:

Built brand new computer in May:
Sandy Bridge 2600K OC'd to 4.5GHz
Ripjaw Series RAM - 8GB
nVidia GeForce GTX 580
7200RPM HDD - 1TB

Current Settings (Short Version):
Running CCC Lvl 4
Phys CPU Set to 3
1920x1080 Resolution
Force DX9
4X AA

My frame rates never seem to average above 40FPS, with some maps running consistently in the low 20s (Death Valley).  I've tried running w/ and w/o CCC, w/ and w/o AA, and various other settings to try and get this game working at playable frame rates without sacrificing all the visual quality, but right now it seems the only way to get decent frame rates is by lowering all the graphics and/or resolution so far it looks like complete crap, and then it's only maybe a 20fps increase.

Just wondering if anybody's been in a similar boat, or perhaps have some settings that work well with this kind of setup?

Hmm tried running dx10?

I doubt its your gc, turn aa off in crysis and turn 32xAa on in nvidia graphic panel, if you notice a huge decrease in performance your gc is a problem, if there is no difference in fps which there should be little or no fps then drivers or cpu or problem in mwll.

Sadly I never tried running mwll with only one gc when my pc was working recently so I am not sure on the exact fps you should be expecting. I do know most people who have high fps configure their settings ALOT, yet you really shouldn't be going down as low as 20.

[Steelbarrage]

It definitely sound like an issue with your 580.
I have my 2500k at 4.3 and a 768 mb gtx 460 and I can run 1920x1080 at high with an average of high 20 to low 30 in heavy battle.
You should see dramatically greater improvement with your 580.

[=KoS= Eldragon]

Got a GTX580 here. Run it on CCC level 4+ with vsync and triple buffering. runs like a dream. However I don't have AA turned on.

[sleepysheep]

Got a GTX580 here. Run it on CCC level 4+ with vsync and triple buffering. runs like a dream. However I don't have AA turned on.

Aa makes almost no difference, gtx580 is awsome at aa. You can get away with 8*aa and 16xaf with almost zero fps difference in most games, and certainly not enough to worry the card. Not that I find all games need aa cranked right up.

Triple buffering if I remember correctly doesn't work in dx10, you guys should be ashamed buying a gtx 580 and running dx9 in a dx10 game lol.


@steel, while the graphic does have more work to do in heavy firefights its generally the cpu that has more work to do and craps out on most peoples machine in heavy firefights. It depends on what's happeneing in the fight.

@Aresye are you getting the same fps in an empty server? And on all maps as well? Have you run stress tests on your oc so you know its 100% stable? Just goto rule out the cpu. Also what are your fps in other games?

[[CW]Aresye]

OC was tested with 24 hours straight of Prime95 with no issues, and yes, my FPS increase is marginally better with no people.  Death Valley might give me an average of 30-35 with no players.

As for the whole DX9 vs DX10.  MWLL doesn't really have much detail difference between the two.  I went with DX9 because it's somehow more stable for me.

[sleepysheep]

OC was tested with 24 hours straight of Prime95 with no issues, and yes, my FPS increase is marginally better with no people.  Death Valley might give me an average of 30-35 with no players.

As for the whole DX9 vs DX10.  MWLL doesn't really have much detail difference between the two.  I went with DX9 because it's somehow more stable for me.

Only 30-35 wow I get better with a broken cpu lol. If drivers are all upto date and temps are fine im stumped.

[Freeborn_Toad]

Something is wrong.

I was running the game with an OCed GTX 480 before I blew it up; that card was only slightly faster than a stock 580, and a 580 is faster than the unlocked/OCed 6950 I'm using now. I typically get 50-60 fps cruising around Death Valley with a mix of high and very high details, 1920x1200 resolution, and forced 4x AA, 16x AF.

I don't use CCC as I find their presets to be less than ideal, sometimes even harmful to visual quality/performance. I did reference them when making my own configs though.

Only issue I can think of that could be slowing your setup down, assuming you have everything installed and configured as it should be, and aren't seeing any throttling, is the amount of video memory MWLL can use on some maps. While my current card is generally slower than my last card, it does have more VRAM, and this does make a perceptible difference in some areas. I've seen MWLL use 1.7GiB of video memory on occasion.

If you could post the driver versions you are using for all major hardware and what processes you typically have running, that might help.

Also, if you are willing to try my custom config files, that would give me a base line to compare performance with your system.


For my config I set the in game quality settings to all high, except for textures (very high), sound (very high), particles (medium), and motion blur (completely disabled).

Then I use the following autoexec.cfg and system.cfg files:

e_cbuffer = 2
e_detail_materials_view_dist_xy = 4096
e_detail_materials_view_dist_z = 1024
e_precache_level = 1
e_water_ocean_fft = 1
q_Renderer = 3
r_colorgrading = 1
r_DepthOfField = 0
r_GeomInstancing = 1
r_sunshafts = 1
r_UseEdgeAA = 0
r_UsePOM = 0
s_FormatType = 3
s_MaxChannels = 128
sys_physics = 1
sys_physics_CPU = 5
r_TexturesStreaming = 2

con_restricted = 0
e_hw_occlusion_culling_objects = 1
e_hw_occlusion_culling_water = 1
e_recursion_occlusion_culling = 1
e_particles_thread = 1
e_terrain_occlusion_culling_version = 0
e_vegetation_static_instancing = 1
es_ondemandphysics = 1
gpu_particle_physics = 1

Note, I also use x64, DX10.

[=KoS= Eldragon]

Got a GTX580 here. Run it on CCC level 4+ with vsync and triple buffering. runs like a dream. However I don't have AA turned on.

Aa makes almost no difference, gtx580 is awsome at aa. You can get away with 8*aa and 16xaf with almost zero fps difference in most games, and certainly not enough to worry the card. Not that I find all games need aa cranked right up.

Triple buffering if I remember correctly doesn't work in dx10, you guys should be ashamed buying a gtx 580 and running dx9 in a dx10 game lol.

I'm strictly talking MWLL here.  Other games get the full AA and dx11 treatment. (Witcher 2 is just jaw dropping when everything cranked to the max).

AA and dx10 may look awesome and do fine on an empty map, but even my OCed GTX580 drops to <20 fps when there is a lot of mechs on screen. Especially on maps like Kagoshima.  And really with MWLL I don't even notice that I have AA turned off except for inside TC bases (Edges of buildings). I want my frames to be smooth as can be when in a dense firefight.

I did a lot of testing with my rig, and I average 5 fps more with dx9 over dx10. With dx10 I get the occasional crash to desktop. dx9? No crashes. Also with dx10 I get the occasional "Freeze" where the game pauses for ~1 second then goes back to drawing frames. This does not happen on dx9.

I do all of my gfx setting tests by going into spectator mode on servers with a full map, and measuring the fps when there were at least four mechs on the screen shooting each other.  The GPU/CPU load created by mechs firing their weapon is MUCH higher than an empty map.

[Freeborn_Toad]

Biggest issue I've seen with AA was in combination with higher particle quality where thick smoke and such was being anti-aliased. In my experience, turning down particle quality results in large FPS gains for almost no noticeable different in appearance.

[sleepysheep]

Got a GTX580 here. Run it on CCC level 4+ with vsync and triple buffering. runs like a dream. However I don't have AA turned on.

Aa makes almost no difference, gtx580 is awsome at aa. You can get away with 8*aa and 16xaf with almost zero fps difference in most games, and certainly not enough to worry the card. Not that I find all games need aa cranked right up.

Triple buffering if I remember correctly doesn't work in dx10, you guys should be ashamed buying a gtx 580 and running dx9 in a dx10 game lol.

I'm strictly talking MWLL here.  Other games get the full AA and dx11 treatment. (Witcher 2 is just jaw dropping when everything cranked to the max).

AA and dx10 may look awesome and do fine on an empty map, but even my OCed GTX580 drops to <20 fps when there is a lot of mechs on screen. Especially on maps like Kagoshima.  And really with MWLL I don't even notice that I have AA turned off except for inside TC bases (Edges of buildings). I want my frames to be smooth as can be when in a dense firefight.

I did a lot of testing with my rig, and I average 5 fps more with dx9 over dx10. With dx10 I get the occasional crash to desktop. dx9? No crashes. Also with dx10 I get the occasional "Freeze" where the game pauses for ~1 second then goes back to drawing frames. This does not happen on dx9.

I do all of my gfx setting tests by going into spectator mode on servers with a full map, and measuring the fps when there were at least four mechs on the screen shooting each other.  The GPU/CPU load created by mechs firing their weapon is MUCH higher than an empty map.

That's strange since dx9 for me runs alot worse than dx10 and crashes all the time compared to dx10. Ofc if you are experiqncing this with dx10 then its understandable why you wouldn't run it. However I really wish we could work out why dx9/10 work so differently on everyone pc for mwll.

@free born, mwll doesn't use 1700mb of vram even in triple screen on any map I have looked on...so no idea how you are getting it to go that high lol

[Freeborn_Toad]

@free born, mwll doesn't use 1700mb of vram even in triple screen on any map I have looked on...so no idea how you are getting it to go that high lol

Frame buffer doesn't use any where near as much as texture buffer, and texture buffer doesn't change with screen resolution.

Here is 1.6GB of usage in a TC_Marshes game I was just in a minute ago. I've seen 200MB more in some other maps, like Death Valley.

[sleepysheep]

Hmm the most I've seen on triple screen is about 1500mb total vram used and on 1900x1200 about 1200mb give or take. I wonder why ou use more vram than pretty much anyone I've heard of for 1900x1200. Ofc it doesn't matter if you got the vram to handle it anyway.

[[CW]Aresye]

Thanks for the config settings Freeborn.  I'll wipe all traces of CCC and try yours out and let you know all the details soon.

[Ingrater]

Try not using CCC and start Wars without mwll loaded. Then use the graphics config menu of wars and go to high / very high there.