Hey guys. Long time lurker, first time poster here.

 

I have been experimenting with some equipment that I have slowly been collecting for an unraid build. Overall, I am very happy with how things are working out. Here is the hardware I have.

 

MOBO : Z10PE-D16 WS

CPU : Xeon E5-2667v3 ES QEYA

Memory : 24gb 2133mhz Kingston ECC

GPU1 : GTX 950

GPU2: Radeon HD7850

 

I currently have 2 gaming VM's setup with no dockers at the moment.

 

Heres what happened today. I recently came across a good deal on a E5-2630v4 ES M0 stepping so I decided to pick it up to try out. Initially I just installed it and let my system boot up like normal, including one VM. Surprisingly, it booted up without an issue coming from the 2667v3. I decided to run CPU-Z and Passmark to see how it compared to my 2667. To my surprise scored the following:

 

CPU-Z

Single  -  1832

Multi  -  7512 (3core / 3HT)

 

Passmark

Overall  -  10,163

Single  -  2097

 

These numbers were quite a bit higher than the 2667. One thing I did notice in CPU-Z though, was that the frequency kept jumping up to 4.1-4.2ghz. I thought maybe I received some magical stepping processor.  I then noticed that my CPU assigment in the VM was using the wrong pairs since I had upgrade from a 8 core to a 10 core. I corrected this to see if I could squeeze out even more performance. To my dismay, the numbers dropped quite a bit. CPU-Z single thread was around 1350 now and the multi was about 4,000. Passmark showed similar drops.

 

Of course changing the cpu assignments back to the old setup did not restore this performance. Could this have just been a freak occurrence? My current numbers seem lower than they should, at least compared to the bare metal benchmarks.

a single benchmark to determine a score is never a good idea. 3-5 tests minimum and then average.

 

now, why did it do that? many factors could contribute to it, from artificially created score to freak solar flare.

 

post your cpu pairs, what the pinning was on the high test, and what it was on the low test. CPU-Z may treat the test differently if the processor is presented in a different topology.

 

are you using an emulator core on your vm? are you isolating cores from unRaid? you should probably read the cpu pinning thread at the top of the forum if you haven't already.

Not sure if it's necessary, but I'm wondering if it's a good idea to reset the BIOS back to defaults, when you switch CPU's, just to make sure it reconfigures itself for the new CPU.

Each time I booted up, I ran each benchmark at least 3 times. I also made sure to do it on a fresh boot and checked the task manager to ensure no background processes were hogging all of the resources.

 

CPU pairs are as follows

 

0,10

1,11

2,12

3,13

4,14

5,15

6,16

7,17

8,18

9,19

 

I currently have 2,3,4,12,13,14 assigned to the VM. I have 9,19 emulator pinned. I also have 2,3,4,12,13,14 isolated as well as a usb controller passed through via syslinux.

what was the original pairing and xml cpu assignment?

 

my guess off the top of my head was that you were running the first test on non-ht paired cores with the previous assignment, meaning they could all (or more of them than not) run at 100% utilization. You can't run 100% utilization on HT pair threads, so lower scores vs. running all single threads. 

 

if you want to see the difference, isolate 2,3,4,5,6,7, give those to the vm, rerun the test and see if it's closer to your first scores. We'll either know that is it, or isolate out a variable.

Original pairing was 4,5,6,12,13,14 with no emulator pinned.

 

Isolating 2,3,4,5,6,7 yielded the following results.

 

CPU-Z 

Single  :  1340

Multi  :  6809

 

Passmark

Overall  :  8593

Single  :  1382

 

Although the multi scores are much closer, the single threaded scores are a ways away. I am feeling like it was some sort of strange glitch as I was seeing core speeds of over 4ghz initially, while this CPU's max turbo speed is only 3.1ghz.