More cores better results?

[jamesschroer]

I have two different chips here one is a 4 core and the other is an 8 core. The 4 core has a higher speed while the 8 core has a lower speed. How does this software work with the multiple cores? The CPU tests are coming out fairly close dispite having twice as many cores. Do all of your CPU tests take advantage of the multiple cores or are there some more based off the pure speed and not the core numbers?

[David (PassMark)]

Yes, the CPU test will use all available cores. That is why (at least at the moment) the 32 core machine is leading on the CPU charts.

But also check under the Edit / Preferences window to ensure the number of processes is at least equal to the number of cores.

[jamesschroer]

Ah thanks I'll give that a try. So is that total number of cores? For example if I have 2 chips with 8 cores should I make that number 16?

Also We have a script set up so that we use the same settings every time. Is there a way to set that before running that script?

Thanks for you help I'm excited to see what the difference is.

[David (PassMark)]

if I have 2 chips with 8 cores should I make that number 16

Correct. But the value should set itself correctly. If it doesn't let us know.
Also if you have a CPU with hyperthreading (2 virtual cores per physical core) then you need to double the figure.

[jamesschroer]

Okay I looked and you are correct it does configure it's self. So I ran the tests again. Here is what I came up with. To me it doesn't seem quite right. Let me know what you think.

System 1 has 2x8 core chips running at 2.4GHz and got a CPU rating of 7533.7

System 2 has 2x4 core chips running at 3.16GHz and got a CPU rating of 8076.5

System 3 has 2x8 core chips running at 3.2GHz and got a CPU rating of 10229.0

[David (PassMark)]

The Mhz clock speed is no longer a good indication of performance. For example, the new Core i7 chips are much much faster than the old Pentium 4 chips even when the clock speed is the same.

I would select the 2x4 system, becuase not all software can run on 16 cores. So you also need to think about the performance of software that runs on a single core, or just 2 cores. The 2x4 system will be faster than the 2x8 system with software that runs on 1 core.

[jamesschroer]

Thanks but we will be using a in house app. We are trying to find out if it would be worth while making the changes from 4 cores to 8 cores. But according to the test the 8 core even if it's being taken advantage of is no better then the 4 core running at a higher GHz. To me I would think the system with 8 cores even though not running as fast should still be able to at least keep up with the 4 core chip that's running at a higher GHz just because it got twice as many cores to do the same amount of work just a bit slower.

[David (PassMark)]

As already pointed out, the Mhz clock speed (and even core count) is not a good indication of performance by itself.

[jamesschroer]

So does the passmark software take advantage of everything else that adds to the performance of a chip? Like you said the performance of the chip is no longer best measured by the MHz and Cores. I'm trying to compare two chips and the one that is said to perform better is not performing much better at all according to the tests ran from the passmark software. That's telling me that the older chip is actually better then the newer chip.

[David (PassMark)]

I won't claim that every feature of every chip is used. But the tests performed as part of the CPU tests are close to, or identical to, algorithms used by real life applications.

So the CPU result should be refective of real application performance. Except we uses all cores, and not all real application use all cores. And PerformanceTest only has 8 different CPU tests, which can't fully represent the performance of all the possible applications available. (i.e. each application is unqiue).