Tips for better performance?

[Mjolnir]

My system is as follows:
i5 2500K
MSI z68
Gskill ripjaws X 1600Mhz DDR3 (4gb)
EVGA gtx 560Ti SC

For some reason I still get fps drops where GPU usage goes below 50% and my FPS drop to the mid 20s. I am running at 1920 x 1200 and the CCC config at level 4 (or 4+ sometimes, the results are about the same). I have the config file set to run physics on a separate core and all of those other enhancements, but I still don't get much above 40% gpu usage at the very most. Overclocking my CPU to 4.2 Ghz doesn't really help the slowdowns either; they are mostly on 30 player servers or smaller servers when a lot is going on. I am running the 64bit .exe in DX9 mode on 64bit windows 7.
Is there any way I can avoid these FPS drops? All of my components are brand new, so I don't see why they shouldn't be able to handle this just as well as any other PC, and from talking to people in servers it seems like some people get better performance than me for some reason with other i5 or i7 cpus.

[sleepysheep]

noone even with a super computer can run the gane without slowdowns on a 30+ server. its physically impossible, and if you are able to do it somehow, your just plain lucky. I sure as hell never could.

if your getting this on server with 24 or less people in the game then you might have problems. just make sure dx and graphical drivers are fully updated.

also try running in dx10 some people find it runs alot better than dx9 always worth a try and your rig should easily be able to run it.

oh and leave your cpu oced at 4.2ghz it does make a difference for general gameplay, maybe not huge but for zero voltage increase, and barley any extra heat why not?

[whatever]

run it in dx10 mode also turn your aa to x2 at 1680 by 1050 res

and set it to good settings and see if that helps.. i have a 560 gtx ti card also

also sometimes the game servers get wonky also and need to be reset every so often..

[Mitchpate]

Lowering shaders increases performance in almost every game.

[Ressk [CSF]]

Quick question, how much am i missing out on with shader's on medium? with it on high im in the low 20's for FPS, with them on medium im in the dreamzone of consistent 40's 

and yes the shadows on sandblasted look funny, but thats all ive seen so far.

[Freeborn_Toad]

Mjolnir, what are you using to measure GPU utilization?

I ask this because I've come across several situations where GPU-Z and MSI Afterburner have failed to register correct utilization. For example, when stress testing with OCCT any shader complexity setting other than 0 results in GPU-Z telling me that utilization is 0%, while MSI AB reports 33%. This is obviously incorrect because all settings above 0 result in greater power consumption and higher temperatures for the GPU.

I'd recommend using Sysinternals Process Explorer (http://technet.microsoft.com/en-us/sysinternals/bb896653) to monitor GPU utilization, if you are not doing so already.

[sleepysheep]

Mjolnir, what are you using to measure GPU utilization?

I ask this because I've come across several situations where GPU-Z and MSI Afterburner have failed to register correct utilization. For example, when stress testing with OCCT any shader complexity setting other than 0 results in GPU-Z telling me that utilization is 0%, while MSI AB reports 33%. This is obviously incorrect because all settings above 0 result in greater power consumption and higher temperatures for the GPU.

I'd recommend using Sysinternals Process Explorer (http://technet.microsoft.com/en-us/sysinternals/bb896653) to monitor GPU utilization, if you are not doing so already.

Occt is a pile of shite. That about sums up everything about it. Gpu-z for me works flawlessly even when i thought it wasnt, it was me not the program!

There are far better graphic card stress tests just I have no pc to find the one I use. I think its furmark is what i used after asking my friend.

Shaders greatly effects the quality of the picture, but i'm sure you can live with it turned down, just try. Oh make sure you delete your shader folder each time, just in case.

I think you may be just trying to run the game on too high a setting for your gc...what settings do you actually use, aa and af etc as well as wars graphic settings.

Edit forgot you use ccc, try just lowering it to level 3? You have a mid range card and if you want awsome fps your going to have to lose out on the eye candy and stop playing in severs with more than 24 people.

[GB-72AT Talon]

run it in dx10 mode also turn your aa to x2 at 1680 by 1050 res

and set it to good settings and see if that helps.. i have a 560 gtx ti card also

also sometimes the game servers get wonky also and need to be reset every so often..

No absolutely not.  Mjolner do not do this except the part to turn up AA use.

1. Change from DX9 to DX10 so you can be taking advantage of your hardware.  By using it on DX9 you are underitilizing your hardware. 

2.  Do not lower resolution, run native for your monitor which you already are, but turn AA up to 4x. 

3.  Try using the CCC 4+ autoexec.  I think you said last night that you were running it.  Its good.  I run it 24/7.

My system is a 955BE @4Ghz with 4Gb of DDR3 1600 memory and a 460GTX 1Gb card.  I run at 1920x1080 with the same tweaks you mentioned, using CCC 4+ autoexec and 4x AA.  I get perfectly fine performance.  You should get at least the same at the same settings.  If not you might need to start over from scratch with the software.  Something is messed up if you can't run it at least that much.

[Freeborn_Toad]

Occt is a pile of shite. That about sums up everything about it. Gpu-z for me works flawlessly even when i thought it wasnt, it was me not the program!

OCCT is one of the most useful GPU stress tests I've ever encountered.

Regardless, I can get GPU-Z to falsely register 0% utilization in a number of other ways. Sometimes forcing AF is enough to do it. This happens in FurMark too.

There are far better graphic card stress tests just I have no pc to find the one I use. I think its furmark is what i used after asking my friend.

FurMark has no error checker, so only errors serious enough to be visually apparent can be detected with it. This is not good enough if stability is a serious concern. I've had GPU OCs that were stable enough not produce visible artifacting but which were unstable to the point they could cause game crashes. Also, I run several OpenCL apps that have no visual output. I need to know the calculations the GPU is performing are accurate, not merely good enough not to cause artifacting.

OCCT also stresses most cards (especially ATI cards) harder than FurMark, making it a better representation of worst case scenarios.

Of course, no single stress test can cover all scenarios, and OCCT only one of many I use.

[Mjolnir]

I am measuring GPU usage with EVGA precision, which seems pretty accurate. The times when I get the slowdowns are not when my GPU is taxed, but when it is apparently waiting for something else (like the CPU). Half of the time I get 99 percent GPU usage and good framerates (45-65 fps) even when graphically intensive things are going on). Its when the GPU usage drops that I lose performance. If I didn't know what CPU I had I would think the CPU was not powerful enough, but the i5 2500K is just about as powerful as a CPU can get nowadays (except for the 2600K),  and it seems like only one core is really being used fully. This happens on maps like Sandblasted with as low as 20 players.
I have switched between CCC 4 and 4+ multiple times with no significant difference; the GPU can handle it just fine. Running in DX10 mode with crysis has always resulted in lower FPS, and running MWLL in DX10 doesn't seem to definitively help my framerates at all. I have the latest nvidia drivers and I believe I have the latest chipset drivers.
In all honesty my first thought is that in the middle of a battle there is simply too much going one for one core to handle, and crysis is not multi threaded enough to take advantage of the other 3 cores I have in a way that would make a difference. However, overclocking my CPU didn't seem to help much, so I am really not sure.

[sleepysheep]

I am measuring GPU usage with EVGA precision, which seems pretty accurate. The times when I get the slowdowns are not when my GPU is taxed, but when it is apparently waiting for something else (like the CPU). Half of the time I get 99 percent GPU usage and good framerates (45-65 fps) even when graphically intensive things are going on). Its when the GPU usage drops that I lose performance. If I didn't know what CPU I had I would think the CPU was not powerful enough, but the i5 2500K is just about as powerful as a CPU can get nowadays (except for the 2600K),  and it seems like only one core is really being used fully. This happens on maps like Sandblasted with as low as 20 players.
I have switched between CCC 4 and 4+ multiple times with no significant difference; the GPU can handle it just fine. Running in DX10 mode with crysis has always resulted in lower FPS, and running MWLL in DX10 doesn't seem to definitively help my framerates at all. I have the latest nvidia drivers and I believe I have the latest chipset drivers.
In all honesty my first thought is that in the middle of a battle there is simply too much going one for one core to handle, and crysis is not multi threaded enough to take advantage of the other 3 cores I have in a way that would make a difference. However, overclocking my CPU didn't seem to help much, so I am really not sure.

Ccc might not be sticking physics onto another core, can't remember the lines to put in cfg thought nor can I remember if ccc does this.

While I have yet to test crysis/mwll performance with my 2500k, if anything could be said from my other game tests is your cpu, as long as its clocked over 4ghz should be having zero problems with this game as long as the servers are under 24 people. Thus something is either horribly wrong software or hardware wise.

First thing is your going to need to check temps of the cpu and gpu while under load, however unlikly, its worth checking. Your need to update dx since you haven't said you have to the latest update.

To make sure it is the cpu, change ccc down to level 1, delete the shader folder,  and then try running the game. This should take the gpu out of the equation.


Also have you tried running any other game to se if you get por performance with those games as well? You should not be losing performance running in dx10, defiently not any noticable fps loss.

@free born, while you should always use more than one gpu test to test a card properly, occt isn't able to read anything correctly about my pc except the gc name and cpu name. Almost everythimg else is incorrect not to mention occt crashes if i try and change most the settings and fails to even get my cards warmed up. Playing solitare stresses my cards out more.

And yes I do have the very latest version. Im not saying it doesn't work for you, just for me I might as well use crysis benchmark than occt. If occt worked it would be a very useful program that's why i downloaded it and tried it in the first place.

[whatever]

there are alot of settings in the crysis engine, my point is you can adjust your settings and see if you like how the game looks

the idea to lower your in game graphics resolution will add more fps to your game. if it doesn't work,  well turn it back to your resolution what ever it was.

also lowering your AA will give more fps as well.. along with lowering your shaders and shadows off will increase performance as well..

another method is to set all the graphics to low and raise each option up to the minimum eye candy to performance ratio..

another thing is to look for a crysis warhead tweak guide. set the settings one level lower than the settings for crysis warhead and it should make mwll performance better..

[GB-72AT Talon]

there are alot of settings in the crysis engine, my point is you can adjust your settings and see if you like how the game looks

the idea to lower your in game graphics resolution will add more fps to your game. if it doesn't work,  well turn it back to your resolution what ever it was.

also lowering your AA will give more fps as well.. along with lowering your shaders and shadows off will increase performance as well..

another method is to set all the graphics to low and raise each option up to the minimum eye candy to performance ratio..

another thing is to look for a crysis warhead tweak guide. set the settings one level lower than the settings for crysis warhead and it should make mwll performance better..

yes, however the problem is his system should be able to handle running at the settings he is running, if not even higher, with perfectly fine performance

if hs has to lower resolution and settings there is something wrong, so instead of lowering the game experience, he should be trying to find the solution

everybody here knows how to dumb down a game to make it run smoother, what we are trying to do here is figure out why his hardware isnt running like it should be, not dumb down the software to match his gimped performance

[Mitchpate]

yes, however the problem is his system should be able to handle running at the settings he is running, if not even higher, with perfectly fine performance

if hs has to lower resolution and settings there is something wrong, so instead of lowering the game experience, he should be trying to find the solution

In my experience, situations like this often have a temperature component to the problem.

[GB-72AT Talon]

yes, however the problem is his system should be able to handle running at the settings he is running, if not even higher, with perfectly fine performance

if hs has to lower resolution and settings there is something wrong, so instead of lowering the game experience, he should be trying to find the solution

In my experience, situations like this often have a temperature component to the problem.

no no no, that goes against what most people have told me here, that having better cooling isn't worth spending any money on.....