Single socket vs Dual socket submissions

[ADVenturePO]

Hello HWBOT Community!

 

Have anyone of You encountered a problem with single socket vs dual socket submissions?

I did.

I'm a Xeon overclocker and thus I'm submitting single and dual socket benchmark scores.

But only one of them can count. Only one of those two can be given points and only one of them will be shown on the table of the CPU under the amount of the sockets with which I had a better score.

 

Examples:

E5-2690 frequency

Dual Socket:

ADVenturePO`s CPU Frequency score: 3951.46 mhz with a Xeon E5 2690

 

Single socket:

Christian Ney`s CPU Frequency score: 3799.9 mhz with a Xeon E5 2690

 

But I did upload better Single Socket score:

ADVenturePO`s CPU Frequency score: 3815.65 mhz with a Xeon E5 2690

 

It doesn't counts. The Single socket 3799MHz for Christian is still in the Single Socket Page for this CPU. Despite in fact it's not the best score.

 

Example no 2

2x E5-2696 v4 PiFast

 

My Single socket result:

ADVenturePO`s PiFast score: 17sec 850ms with a Xeon E5 2696 v4

 

Dual Socket result goes for H20 vs LN2:

H2o vs. Ln2`s PiFast score: 20sec 500ms with a Xeon E5 2696 v4

 

But I have uploaded Dual socket score: 18s 280ms

http://hwbot.org/submission/3397588_

 

 

There is more of it od course.

 

 

I do find it as a problem because:

1. I'm not getting scores when I should.

2. The results on CPU pages aren't real - 10% faster result is a difference of a processor family.

3. Regulating single socket and dual socket isn't the same.

 

 

How can we change it? Who should I turn to?

 

Help me please.

 

Peter (ADVenturePO)

[mr.paco]

I too bench multi socket systems

With any single threaded benchmark including Cpuz freaquency no matter how many cpus used will only take your best run.

Only multithreaded benchmark will grant points / rankings per number of cpus used.

Hope that helps?

[I.nfraR.ed]

That is correct. All results in single-threaded benchmarks like superpi, pifast, cpu frequency go to the unified ranking and only your best counts, no matter how many cores you had enabled during the benchmark.

 

Btw, your 3952MHz result is invalid, because the validation file only shows x36 multi, not x38.

[ADVenturePO]

Mr.Paco - thanks for a reply. Do You support idea of this being not fair?

 

I.nfraR.ed, thank You for a reply.

(About validation: I disagree: can You look at the sceenshoot at a side? Well, it isn't my fault that CPU-Z cant read the max frequency in dual CPU systems. Many times I'm achieving much higher frequencies than shown in validation. I didnt put it from air. I run benchmark SuperPi and CPU got instantly at highest frequency)

You wrote:

"All results in single-threaded benchmarks like superpi, pifast, cpu frequency go to the unified ranking and only your best counts, no matter how many cores you had enabled during the benchmark." But it's not about cores but number of CPUs. Other boards, other RAM, other PSU... And why should the worse result stay at the main CPU page if the score was beaten?

Edited December 27, 2016 by ADVenturePO
minor errors

[mr.paco]

Mr.Paco - thanks for a reply. Do You support idea of this being not fair?

 

I.nfraR.ed, thank You for a reply.

(About validation: I disagree: can You look at the sceenshoot at a side? Well, it isn't my fault that CPU-Z cant read the max frequency in dual CPU systems. Many times I'm achieving much higher frequencies than shown in validation. I didnt put it from air. I run benchmark SuperPi and CPU got instantly at highest frequency)

You wrote:

"All results in single-threaded benchmarks like superpi, pifast, cpu frequency go to the unified ranking and only your best counts, no matter how many cores you had enabled during the benchmark." But it's not about cores but number of CPUs. Other boards, other RAM, other PSU... And why should the worse result stay at the main CPU page if the score was beaten?

Its not so much about being fare as it is about making sense.

When benching a single threaded benchmark like super-pi, if I bench 2 x E5520 that would equal a total of 8 cores/16 threads BUT the benchmark will only use one thread or core from one cpu only to make its computation (that's the way single threaded benches are meant to be, to test one thread or one core)

The only thing I can see that would any difference would be in OCability.

You can OC one cpu or one core much higher ndmore stable then OCin 2 cpus.

So it really wouldn't make much sense separating single threaded benchmarks by number of cores or threads or cpus used because it only runs on one thread on one cpu.

[Strunkenbold]

Well, it isn't my fault that CPU-Z cant read the max frequency in dual CPU systems. Many times I'm achieving much higher frequencies than shown in validation.

 

This isnt the fault of CPU-Z. You have to force the CPU to be in the highest state while you validate. We cannot accept results just because the user says "he saw that his CPU is capable of reaching 5Ghz" while he just provides a 2Ghz validation.

I edited your result.

[ADVenturePO]

Thanks Strunkenbold, agreed to that.