Server
Hi,
We are using Geekbench 2.3.4 (I believe) to benchmark an HP server against an in-house built server. The latter uses the same CPUs, chipset and RAM as the HP but is running SSD rather than mechanical HDD. Unfortunately, despite the identical processing hardware, the results returned are showing the in-house server to be 31% slower. We are not sure what could be causing these results. Could you provide any ideas (we have a deadline of 24 hours so a quick response would be greatly appreciated) :)
Thank you.
- GeekBench_RAW_DATA.xlsx 16 KB
Keyboard shortcuts
Generic
| ? | Show this help |
|---|---|
| ESC | Blurs the current field |
Comment Form
| r | Focus the comment reply box |
|---|---|
| ^ + ↩ | Submit the comment |
You can use Command ⌘ instead of Control ^ on Mac
Support Staff 1 Posted by John on 18 Sep, 2013 06:56 PM
What operating system are you running on the servers?
Also, do you have the Geekbench 2 result files for both servers?
2 Posted by B Miles on 18 Sep, 2013 07:10 PM
Server 2008
3 Posted by B Miles on 18 Sep, 2013 07:11 PM
Unfortunately, I currently only have the raw data files.
Support Staff 4 Posted by John on 18 Sep, 2013 07:17 PM
Do you know if both servers are using the same power plan? You can find out which power plan is currently active by opening the Control Panel and selecting "All Control Panel Items", then selecting "Power Options".
5 Posted by B Miles on 18 Sep, 2013 07:23 PM
I think it's likely they are not - the Supermicro idles at around 1.2GHz, the HP at around 3.2GHz. Could this account for such a large discrepancy?
6 Posted by B Miles on 18 Sep, 2013 07:31 PM
Actually, my colleague has just provided the result files which I have attached.
I will only be able to get the power settings tomorrow morning. Could that explain the difference and in particular what settings could affect the score?
I assume you wont be online until later tomorrow morning (we are on GMT)?
In your opinion, what should the result be from these servers?Would the SuperMicro with SSD produce much higher results due to the high I/O?
7 Posted by B Miles on 18 Sep, 2013 07:42 PM
Hi John,
Did you get my reply with the attached data files as I posted over ten minutes ago and can't see it?
Support Staff 8 Posted by John on 18 Sep, 2013 07:52 PM
Thanks for the heads up. I didn't see the reply as it was flagged as spam.
Given the different idle speeds I think the power plan is the culprit here. I have seen similar discrepancies from power plans in the past. Try adjusting the power plans tomorrow and see what happens.
Also, Geekbench doesn't measure IO performance, so the SSD should not affect scores much (if at all).
9 Posted by B Miles on 18 Sep, 2013 08:02 PM
Thanks John, fingers crossed we will get a big uplift with a performance power plan.
I hugely appreciate your support!
Support Staff 10 Posted by John on 23 Sep, 2013 05:18 AM
Did you find out the source of the difference between the two machines?
11 Posted by Ben - Chillblas... on 23 Sep, 2013 05:19 AM
Thanks for your email, unfortunately I am on annual leave from Friday the 19th of September to Monday the 22nd inclusive. My email will be monitored periodically, but If your matter is urgent, please email [email blocked] or call the office on 0845 45678 31.
Best Regards
Ben
www.chillblast.com<https://amxprd0410.outlook.com/ecp/Organize/www.chillblast.com>
12 Posted by B Miles on 24 Sep, 2013 07:52 AM
It was the power scheme as you suspected. The variance is now just 2%. Out of interest, do you have an approximation for the statistical margin for error of the benchmark?
Thanks once again for your excellent support on this matter.
13 Posted by B Miles on 09 Oct, 2013 03:22 PM
Hi again John,
We have built a new server with the latest Intel Xeon E5-2697 v2 CPUs (x2). When we run the tests, the results seem a bit strange. The system info shows 2 processors, 24 cores, 32 threads - it should be 48 threads. The system RAM 32GB (1866mHz) is shown as 32734 MB -1MHz. Only Floating Point Performance is where we assume it should be, the rest are slower. Any ideas?
Support Staff 14 Posted by John on 09 Oct, 2013 03:35 PM
Hi Ben,
It looks like you're running an older version of Geekbench 2. Could you upgrade to Geekbench 2.4.3 and see if that makes a difference? You can download Geekbench 2.4.3 from the following location:
http://www.primatelabs.com/geekbench2/download/
Thanks,
John
15 Posted by Ben - Chillblas... on 09 Oct, 2013 03:38 PM
?Thanks for your email, unfortunately I am currently out of the office. My email will be monitored periodically, but If your matter is urgent, please email [email blocked] or call the office on 0845 45678 31.
Best Regards
Ben
www.chillblast.com<https://amxprd0410.outlook.com/ecp/Organize/www.chillblast.com>
16 Posted by B Miles on 09 Oct, 2013 05:33 PM
Hi John,
We have tested with the latest version and the same issues are still apparent. I noticed there is a similar thread http://support.primatelabs.com/discussions/geekbench/1519-impossible - do you have any further info on this issue? Also with the RAM not being recognised - could this explain the low performance?
Thx :)
Support Staff 17 Posted by John on 12 Oct, 2013 03:29 AM
Hi Ben,
Looks like you're experiencing the same issue. We're actively investigating the issue and I will let you know when we have a fix.
As for the low memory performance issue this could be an architectural issue with Geekbench 2 that was addressed in Geekbench 3. Have you tried running Geekbench 3 on the system?
Thanks,
John
18 Posted by Ben - Chillblas... on 12 Oct, 2013 03:29 AM
?Thanks for your email, unfortunately I am currently out of the office. If your matter is urgent, please email [email blocked] or call the office on 0845 45678 31 and one of my colleagues will be happy to help.
Best Regards
Ben
www.chillblast.com<https://amxprd0410.outlook.com/ecp/Organize/www.chillblast.com>