PassMark showing higher 2D and 3D Mark for overclocked CPU

[Salil]

Hello,
I had a overclocked my E8500 from 3.16 GHz to 4 GHz and ran PerformanceTest on it and it showed a higher CPU Mark from 1999 -> 2541. It also showed a higher 2D Mark from 447 -> 543 and a higher 3D Mark 5234 -> 5811. I then swapped my CPU for a Q9650 3 GHz. I see that the CPU Mark went up from 2541 -> 3846 but my 2DMark fell 543 -> 429 and 3DMark fell from 5811 -> 5534. Why did the stronger processor upgrade overperform the overclocked E8500 handily but underperform on 2DMark and 3DMark? My GPU is Nvidia 1050 Ti.

 

[Kenmitch]

My guess would be your running the Q9650 at stock clocks. The E8500 at 4GHz should be faster under certain circumstances as it would have better single core performance.

 

[HutchinsonJC]

Your CPU upgrade is from a dual core to a quad core both from ~2008. IPC should be similar enough between the two.

So in your CPU test, you should expect that 4 cores will outperform 2, even if the 4 cores has a small deficit in clocks compared to your dual core once it was overclocked. That's a fair and logical thing to conclude, I think... beginner or expert, right? This isn't a proper mathematical explanation, because it doesn't work this way, exactly, but 4 cores * 3GHz = 12 ... where as 2 cores * 4GHz = 8. Your quad will win.

In graphics test, if you're not testing at a high resolution to really stress the GPU, or the graphics test isn't really written to stress the GPU, then the CPU will have a big say in the final graphic score because the GPU isn't a limiting factor... it's breezing through the workload no problem and is waiting on the CPU. So in a graphics workload that's not stressing the GPU more than the CPU, the CPU will become a limiting factor. And because the workload is probably not really relying on 4 cores, and instead is relying on primarily a single core with high clocks/ipc, the graphic work load is easier to deal with on the higher clocked dual core for this test.

If the graphic test was more taxing on the GPU, the CPU's bottle neck might not have shown its face because the GPU would have been the limiting factor instead of the CPU.

 

[LTC8K6]

Hello,
I had a overclocked my E8500 from 3.16 GHz to 4 GHz and ran PerformanceTest on it and it showed a higher CPU Mark from 1999 -> 2541. It also showed a higher 2D Mark from 447 -> 543 and a higher 3D Mark 5234 -> 5811. I then swapped my CPU for a Q9650 3 GHz. I see that the CPU Mark went up from 2541 -> 3846 but my 2DMark fell 543 -> 429 and 3DMark fell from 5811 -> 5534. Why did the stronger processor upgrade overperform the overclocked E8500 handily but underperform on 2DMark and 3DMark? My GPU is Nvidia 1050 Ti.

The quad core beats the dual core in the CPU test because that test can use more cores.
The dual core's much higher clock beats the quad core in the graphics test, but not by much, because that test is likely not using the extra cores of the quad.

Overclock the quad core a bit and it will lead in both.

The basic answer is that some programs/benches do not take advantage of all the cores available, so that a quad core might score the same as an eight core chip, or even better if the quad is clocked higher.

Then when you move to a program/bench that does use all the cores, the eight core chip struts it's stuff, and leaves the quad in it's dust.