Thanks for the details! While I can't personally speak to the validity of any particular memory benchmark tool, I believe there must be a significant performance difference in memory access given the CPU differences. You can compare them side-by-side at ark.intel.com/.../37103,52207 The 2.4GHz E5530 memory controller operates at a maximum rate of 1066MHz. This is going to account for significant performance difference, and to a lesser extent the the fact that the R710 memory is ECC while the desktop memory is not. I hope that helps to clear things up!
There are a number of hardware and configuration factors that can impact the result of benchmark software. Did I understand correctly that you are testing 6x 4GB DIMMs in the R710 and then testing those same DIMMs again in another machine? Or is it a proper "desktop" machine that does not support registered DIMMs and has its own 6x 4GB DIMMs? If so, that is probably why you are seeing such a difference, these would be different architectures. In any case, in the absence of any other indicators the benchmark number for the R710 does not sound unreasonably low to me, though I understand your curiosity about the benchmark reporting difference. I would probably look at architecture differences to account for this. The R710 has a X5500 or X5600 series CPU and Tylersburg chipset. What CPUs and chipset are in your other test machine? In general I would also recommend updating the R710 BIOS to the latest (as of right now, it's 6.2.3; RHEL update package at downloads.dell.com/.../R710_BIOS_3PVDX_LN32_6.2.3.BIN; changelog: downloads.dell.com/.../R710-060203CBIOS.txt) to ensure the latest CPU microcode from Intel is present. Let us know about your other machine and what CPUs are in the R710 too and hopefully we can make some sense of it.
DELL-Jonathan S
153 Posts
1
July 27th, 2012 16:00
Thanks for the details! While I can't personally speak to the validity of any particular memory benchmark tool, I believe there must be a significant performance difference in memory access given the CPU differences. You can compare them side-by-side at ark.intel.com/.../37103,52207 The 2.4GHz E5530 memory controller operates at a maximum rate of 1066MHz. This is going to account for significant performance difference, and to a lesser extent the the fact that the R710 memory is ECC while the desktop memory is not. I hope that helps to clear things up!
DELL-Jonathan S
153 Posts
0
July 20th, 2012 08:00
Hi,
There are a number of hardware and configuration factors that can impact the result of benchmark software. Did I understand correctly that you are testing 6x 4GB DIMMs in the R710 and then testing those same DIMMs again in another machine? Or is it a proper "desktop" machine that does not support registered DIMMs and has its own 6x 4GB DIMMs? If so, that is probably why you are seeing such a difference, these would be different architectures. In any case, in the absence of any other indicators the benchmark number for the R710 does not sound unreasonably low to me, though I understand your curiosity about the benchmark reporting difference. I would probably look at architecture differences to account for this. The R710 has a X5500 or X5600 series CPU and Tylersburg chipset. What CPUs and chipset are in your other test machine? In general I would also recommend updating the R710 BIOS to the latest (as of right now, it's 6.2.3; RHEL update package at downloads.dell.com/.../R710_BIOS_3PVDX_LN32_6.2.3.BIN; changelog: downloads.dell.com/.../R710-060203CBIOS.txt) to ensure the latest CPU microcode from Intel is present. Let us know about your other machine and what CPUs are in the R710 too and hopefully we can make some sense of it.
tommo666
3 Apprentice
•
1.2K Posts
0
July 20th, 2012 09:00
There is a memory pdf here: support.dell.com/.../710_IU.pdf
It does state that 3x single or dual rank dimms per channel supports 800mhz but 2x per channel supports 1333mhz.
082net
4 Posts
0
July 23rd, 2012 18:00
Thank you for your answer tommo666
The PDF you've linked explained 'One and two memory modules per channel support up to 1333 MHz'
and our R710 has 1 module per channel ( A1, A2, A3 -- B1, B2, B3)
Channel 1 includes A1, A4, A7 and channel 2 includes A2, A5, A8 ....
082net
4 Posts
0
July 23rd, 2012 19:00
Thank you for answer DELL-Jonathan S.
I've already updated R710 BIOS to the latest(6.2.3) and there's no diffrerence.
As I mentioned above, R710 and desktop has same OS - CentOS 6 up-to-date - and same packages and same configurations
We have 6 more Dell servers, and four of them are R710, one is R810 and the rest is R610
Tested all servers with sysbench and all results are about 2MB/sec, I thouht that it's normal until I tested on my desktop local server.
The result on desktop was 3MB/sec and 2MB/sec on R710 servers.
I've tried another benchmark tool RAMSpeed ( http://alasir.com/software/ramspeed/ ) and desktop was also 1.5~2x faster than R710
Is there any other reliable memory benchmark tool?
R710
CPU: Intel(R) Xeon(R) CPU E5530@2.40GHz X 2
Memory: 4GB DDR3 1333Mhz X 6 ( 1 module per 1 channel - A1,2,3 B1,2,3 )
Desktop - local test server
CPU: Intel(R) Core(TM) i5-2400 CPU @ 3.10GHz X 1
Memory: 2GB DDR3 1333Mhz X 2
Same OS, same packages, same configurations
I just want to know that this is normal, or if it's not, wanna know that there's any solution for fix it.
Thanks!
quantranblog
9 Posts
0
July 25th, 2012 02:00
Hi,
R710
CPU: Intel(R) Xeon(R) CPU E5530@2.40GHz X 2
Only support DDR3 800/1066
Desktop - local test server
CPU: Intel(R) Core(TM) i5-2400 CPU @ 3.10GHz X 1
Support DDR3 1066/1333
This maybe root cause.
082net
4 Posts
0
July 30th, 2012 23:00
Thank you for kind response. You makes me clear ;-)