CPU Rocks... except in physics and strings?

[Richard Parker]

This is strange.

I have an overclocked Q8200 (@ 2.87 GHz). My CPU scores all exceed comparable CPUs (Q8200s, Q9400s), except for Physics and String Sorting, which are both less than 40% of the comparable computers.

This is strange, right?

What could be making my Physics and String results so out of whack with the rest of my CPU results, and other comparable CPUs?

[David (PassMark)]

What are the actualt results?

I think it is these 2 tests which make the most use of main RAM. Some of the other tests use a small enough amount of data to be held entirely in the L2 / L3 cache.

[Richard Parker]

Thanks for the reply.

The overall CPU Mark is 4178.
Integer is 1794.
FPU is 2758.
Primes is 1012.
SSE is 10.1
Compression is 6264.
Encryption is 18.
Physics is 63.8.
String Sorting is 1046.


In all but the last two categories, I'm at (and usually well above) the comparable systems. But I get blown out by a scale of magnitude on the last two...

[David (PassMark)]

Yes, those 2 results look a bit low.
I would start by looking at your memory access speeds. Maybe the memory timings have got worse as you have overclocked the CPU.

Try going back to stock settings and see if these 2 benchmarks improve.

[Richard Parker]

Yes, those 2 results look a bit low.
I would start by looking at your memory access speeds. Maybe the memory timings have got worse as you have overclocked the CPU.

Try going back to stock settings and see if these 2 benchmarks improve.

Well, I got rid of my overclock, and little changed. Physics = 81, String = 1252.

Any other ideas?

Also, in terms of real world performance impact (e.g., games)... is there likely to be much, as a result of this?


Thanks again for your help!

[Richard Parker]

One other thing: my MB doesn't technically support my processor. I have a Q8200 on an ASUS P5N-E SLI.

I say "technically" because the 45 nm quads aren't on the official CPU support list. I haven't had any trouble with the pairing, however I understand that some others have. The notional incompatibility is due, I've read, to the board not being able to pump out the required amount of juice at all times.

Could this be the cause, or is it more likely a separate memory issue?

[David (PassMark)]

and little changed

Performance improved around 25% by slowing down the CPU. As this is the opposite of what one would expect, I think this is signficant.

While I can't tell you the exact cause of the problem. I think this might have a real world impact.

We actually had a similar issue in house with a i5 CPU and a Foxconn motherboard. Memory access speeds were very bad. It affected some of the CPU results, the memory benchmarks and the disk benchamrk to some extend. After investigating the problem for half a day we gave up and threw away the motherboard. Replacing the MB fixed the problem for us. We won't be buying another Foxconn motherboard in the future.

Try some other benchmarks, have a closer look at memory performance & timings.

[Richard Parker]

Well, I fixed it. The Passmark results were what sealed it--I decided to get a new MB, and my results in the test and performance (especially in modern, CPU-intensive games like Battlefield BC2 and Assassn's Creed 2) more than doubled.

Thanks!