Jump to content
AIDA64 Discussion Forum
Zakkorn

Benchmark results lower than expected

Recommended Posts

My PCs spec is the following:

 

Dual Xeon 2690 V3 (2600MHz)

64GB RAM (8x8GB) (2133MHz)

ASUS Z10PE-D8 WS

Corsair 1200W

Quadro K4200 (Primary)

Titan X

 

 

As I expected my scores should have been higher than the Dual Xeon 2660 V3.

 

But of course are lower in some benchmarks

 

                                              MINE (Dual Xeon 2690)             Dual Xeon 2660

in CPU Queen                                  120220                                  147245                                

in CPU PhotWorxx                              56419                                    62186

in FPU VP8                                          6045                                      6650

 

 

I have tried many installation of Win 8.1 and Win 10

 

I have even changed the motherboard to a SuperMicro DAi and I got almost identical results + 1-2% faster due to less lag in memory and plus 1MHz more in Bus

Thats it.

 

Why is that?

2 Different motherboards many installations of Windows 8.1 or 10 and still behind the 2660???

Share this post


Link to post
Share on other sites

Please try the following, and let me know what results do you get. Go to the CPU Queen in AIDA64, push the Parameters button on the toolbar, and select 20 CPUs. Then start the benchmark, and note down the score you get. Then do it for CPU PhotoWorxx as well.

Thanks,

Fiery

Share this post


Link to post
Share on other sites

I did the change to 20 cores.

In Queen I went higher in score 155764

but in PhotoWorxx I went lower 49783

In VP8 I went higher as well 7667

 

I want to add that in PhotoWorxx that the score when I have 24 cores (the normal ones) is between  50160 - 50400 and I dont why I wrote 56419 yesterday.

So the change to 20 cores pushed the score a little downwards.

Share this post


Link to post
Share on other sites

It's definitely an odd issue. I can think of 2 issues that might explain this:

1) Turbo Boost and power limits are configured in a way that they kick in too much with all cores utilized, preventing your CPU from running at high clock speeds.

2) Haswell-EP processors may not scale perfectly when going from 10 cores to 12 cores. Maybe 10 cores mean a more ideal configuration than 12 cores, and in certain situations utilizing 2 more cores could mean a regression in overall performance or per-core performance.

We've done a test run on our 10-core reference Haswell-EP Xeon system to see how it scales when going from 1 core to 2 cores, then 4 cores, 6 cores, 8 cores, 10 cores, and the scaling didn't show any regression or bump in the performance. The pecularity that you're experiencing may well be related to the 12 cores your CPUs feature.

Share this post


Link to post
Share on other sites

Hmmm. Strange

So I had bad luck that chose 2690 over 2680 (10 cores)?

 

Very weird.

 

Because I know that I get the maximum MHz boost per core for my CPU (3100MHz)

Thats why I believe that the 1st case is not possible to be the answer.

Share this post


Link to post
Share on other sites

No, I'm not saying 2690 is worse than 2660. What I'm saying is that we've already seen such regression in performance before, although it was more about overclocking vs. memory bandwidth and latency. Sometimes adding more resources do not mean an increase in performance.

Of course, it would be best to confirm those results by running other benchmarks as well. Of course it has to be such a benchmark that could scale up to 12 cores / 24 threads per socket, and can also be limited to use only 10 cores.

Share this post


Link to post
Share on other sites

In ZLib Hash Julia Mandel and SinJulia I am faster

 

So that means that these benchmarks are more Core friendly, right?

It means your CPU scales better when AIDA64 is running those benchmarks. VP8 and PhotoWorxx are different beasts anyway. VP8 because it doesn't scale with all cores in the system perfectly, so its behavior is different to all the other benchmarks. And PhotoWorxx relies heavily on the memory and cache subsystem, so in many cases it will not scale well after a certain point because of the narrow memory bandwidth (or cache bandwidths).

Share this post


Link to post
Share on other sites

OK.

Thats new "territory" for me. (Dual Xeons), so I didn't think of that kind of situations.

I was expecting my system to be faster than a lower Xeon in every possible situation.

But as I see now, thats not true.

 

Anyway thanks for your time.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...

×
×
  • Create New...