460m - 142Gflops? Performance issue

Hey all.
I am flat out of ideas here. I have 2 Alienware M15x’s with 460m’s. Both have the same bios. All of the parts are the same except the processor.

Alienware ‘A’ gets full computational performance. It has an Intel Core i7-620M.
Alienware ‘B’ gets 142Gflops. It has an Intel Core i7-740QM.

Next step…

I take the i7-620M out of the working laptop and place it into the low performance laptop (‘B’). Alienware ‘B’ now gets full performance.

There is definitely something going on with these processors. The problem is 100% reproducible.

I have verified that the ACPI tables are not only close, they are exactly the same. If I boot the good performing Alienware with kernel parameter acpi=off, then it goes down to being 142Gflops.

I am going to crack soon if we can’t figure this out :spaz:

More info…
Working processor is an Arrandale which has integrated graphics.
“Failienware” is a Clarksfield which does not have integrated graphics.

Side by side comparison is here

Thanks in advance for whatever help you can give.

Presumably, both systems throttle down GPU/VM clocks to a safe level when idle.

Does the failing system fail to raise the clock back to where they should be under load?

Could it be power? The obvious difference between the two CPUs is 10W TDP. That might mean that there isn’t enough power to allow the GPU to run at full clocks.


I was thinking it might be power too, but the system came preconfigured with the CPU, which leads me to believe that there would have at least been some sort of testing done before releasing it. I will look into it further though.

Correct, running a CUDA benchmark I wrote that gets 99.998% theoretical on PCIe cards should bring it up to load, I know it does on desktop cards. But even with this it can’t raise the performance above “Performance Level 1” in nvidia-settings.

Additionally, I was playing around with the working system… the one that gets full performance, and I noticed that if I use the kernel parameter acpi=off, the full performance machine drops down to getting 142GFlops…

When I boot with kernel parameter debug, I see “Performance Events: Westmere events, Intel PMU driver.” on the one that fails, and “Performance Events: Nehalem/Corei7 events, Intel PMU driver.” events on the machine that works.

if you want me to add kernel parameters or post dmesgs let me know, its no problem.

Thanks for the help,



*reason for edit: trying to fix quotes


I hate to bump, but any suggestions would be extremely helpful!

Replying rather wildly, but just with a thought… I think the alienware with the gtx460m is one of the only notebooks to use optimus. I am wondering if the problem isn’t caused by that. Can you try disabling the intel graphics in the notebook’s bios?

Oops - reading your first post more carefully, I see it is the other way around…