Diamond Member
Mar 15, 2007
The OS will move threads around. Just because the average over a second is 50% does not mean a single thread isn't using 100% of a CPU somewhere.


Platinum Member
Jul 28, 2004
What am I supposed to do when this happens:

In Crysis I always see the culprit: one CPU core at 100%. But in Warhead?

46fps with barely 30% GPU load and every CPU core is below 50%.

That's not fair. Where is the bottleneck?
Crysis is not threaded for more than 2 cores, and even that I believe is with sound/network/small extra things being offloaded on the second core. It is very badly optimized for more than a single core because it is a product of its time, when CPU's were encouraged to be fast single processors for gaming.

Crysis will bottleneck with EVERY processor right now because of how it's designed, no way around it. Want extra FPS? Overclock your proc to 8.0 ghz with some Liquid Nitrogen.

Warhead is better threaded, and that's why you don't see 100% usage on one core at all times, so your FPS is probably being limited by some other component, most likely memory bandwidth or GPU processing power. Don't go by GPU load exactly, increase GPU memory and Core clock rates and see if your FPS goes up proportionally.

System specs would be helpful as well

P.S. - Pretty much every game now is being designed to take advantage of multiple cores, Battlefield 3's complexity is a great example. I get activity across all 6 cores and smooth 60 fps, the same game on my Dual Core was constant stuttering with both cores at 100% (MOBO, RAM, AND GPU identical in both tests).

Basically, you won't have this problem in the future. Multiple cores are the way to go now with gaming, you just happen to be playing two games that were on the bleeding edge of technology at the cusp of when the transition to multiple cores for gaming started to be recommended.
Last edited:


Platinum Member
Feb 13, 2008
Every core at 50% points to a dual core optimized game maxing out two threads on a quad core CPU.
An overall usage of 50% would suggest that (2 maxed, and 2 not), but not all 4 individual cores at 50%.

Some multi-threaded apps are just badly coded, they can share memory space between several thread and if one thread locks a piece of memory for editing the other threads can enter wait states where they're literally doing nothing while they wait for a piece of memory to be unlocked so they can modify it. This can lead to really big inefficiencies and much lower than 100% core usage. I can't say for sure that's what is happening here, but it's plausible.

Other explanations for both GPU and CPU under low load is artificial bottlenecks, is there a max FPS set, or are you using vsync?

One other possible explanation is memory size or bandwidth issue, you'll get low frame rates if your RAM or vRAM have to swap assets to disk because the settings are too high. You'll also get low usage if the GPU or CPU are starved of data because of memory bandwidth bottlenecks, not much can be done about this as it's a hardware limitation.

It's actually fairly common issue with SLI/Crossfire, because you don't double the effective amount of vRAM (vRAM is mirrored/cloned) and so you don't double the amount of effective memory bandwidth, but you do double the GPU power.
Last edited: