Dual Core HAL

[ctmiller]

I recently started a new job where our team evaluates commercial laptops. I've just learned that performance scores from our image were consistently higher than on factory loads for dual core machines.

I found this rather odd since our baseline load seems very bloated and far from optimized. Then I discovered something interesting. Someone did not properly sysprep the machine our baseline image was taken from and as a result the ACPI Uniprocessor PC HAL was being installed instead of the ACPI Multiprocessor PC HAL. (For clarification - the multi is also used on the factory load.) I have since fixed this issue.

Benchmarking the Uniprocessor HAL against the Multiprocessor HAL on identical machines with identical loads I found an unexpected surprise. The uniprocessor HAL creamed the multiprocessor HAL by almost 200 points. Any thoughts?

HP NC6320
Intel Centrino Duo T2600 @ 2.16Ghz
4 GB DDR2 Dual Channel RAM
Windows XP SP2
Performance Test v5.0

UNIPROCESSOR HAL
http://home.comcast.net/~christmiller/uniprocessor.jpg

MULTIPROCESSOR HALhttp://home.comcast.net/~christmiller/multiprocessor.jpg

[passmark]

My guess is that, you are using the oldf V5 release of PerformanceTest which doesn't know anything about dual core machines or Duo CPUs. So probably both of your tests runs were only using a single CPU core.

When you are only using 1 CPU/core the Uniprocessor HAL is probably more efficient.

If you retry the tests with PerformanceTest V6, I think you will see a big difference as PT6 can detect and use multiple cores.