CPU load nox at max during Fire Strike physics?

[pegnose]

Hi,

6700k @ 4.6 GHz on Win10 64 bit
3000 Mhz dual-channel DDR4 (CL15)

Neither in Fire Strike nor in 3DMark 11 the physics test is able to load my CPU to its maximum. I see an average total CPU load of 91% (OSD from RTTS) with 95% on rare short occasions, as well as 84%.

Is that typical? And if not, how do I fix it?

I probably don't need to say that but other tools are able to max out the CPU just fine (Prime95, e.g.).

 

[Ratman6161]

I don't use any of those tools, but here are a couple of thoughts.
1. Some other bottleneck kicking in resulting in the CPU sitting their waiting?
2. Other processes running on the system? For example when running something like Intel Burn Test and watching CPU utilization, if I do some other task, I can clearly see the cpu drop below 100%. So could you have something running in the background that's effectively doing the same thing?

 

[BonzaiDuck]

Do you let your graphics card(s) select one of themselves for PhysX, or do you manually assign PhysX to the CPU? Maybe my question in naïve, but then again . . .

I really wanted to say that the only reason I run FireStrike is to benchtest and stability-test my graphics configuration. So I never thought about this that much.

 

[pegnose]

Another bottleneck? Entirely possible.

Other processes should result in the overall CPU load as reported by the Riva Tuner Statistics Server to increase. And this is the case if I put additional load onto the CPU. I get 100%. But not by the physics test alone. Maybe it is how the test works. I am just curious.

I tried all three combinations: physics to CPU, to GPU2, or to AUTO. Does not matter. I think 3DMark does not respect these settings as the physics test is designed to test the CPU, no matter what.

 

[pegnose]

Could anyone perhaps just run the physics test and tell me how it is with your rig?

 

[pegnose]

So no one is running Fire Strike here?

 

[Hitman928]

I'll run it for you at some point today and let you know.

 

[pegnose]

Thanks, that would be great!

 

[Hitman928]

Looks like it maxes out my intel 4690k just fine, but that doesn't have hyperthreading. I'll see if I can run on a machine with SMT later.

 

[pegnose]

Utilization is definitely higher without hyper threading. I even saw the occasional 100%. But then, CPU load with HT always puzzles me.

 

[Burpo]

93 to 94% across all threads & hit 56 watts (Xeon E3 1280V2)..

http://www.3dmark.com/3dm/19462064

 

[Ratman6161]

Another bottleneck? Entirely possible.

Other processes should result in the overall CPU load as reported by the Riva Tuner Statistics Server to increase. And this is the case if I put additional load onto the CPU. I get 100%. But not by the physics test alone. Maybe it is how the test works. I am just curious.

I tried all three combinations: physics to CPU, to GPU2, or to AUTO. Does not matter. I think 3DMark does not respect these settings as the physics test is designed to test the CPU, no matter what.

Its not as obvious as it may appear..Ill give you an example. Last night I was running Intel Burn test and getting the same issue as you. In task manager, I sorted the process tab on cpu utilization and as expected, Linpac (what IBT is actually using) was using most of the CPU. But my crashplan backup was trying to run and using 2% CPU. I killed the task and cpu utilization went up to 100% total.