12-24-2012 08:55 AM
Hi all,
I have some performance issues that I can't seem to straigthen out.
These screenshots should give you an idea what I mean:
The SSD is connected to a GigaByte GA-770TA-UD3 through the AMD SB710 host (the Marvell 9128 controller performs even worse).
Latest BIOS update, AHCI enabled before installing Windows 7 Ultimate 64 bit. Used all the latest drivers.
What I've tried so far:
Any ideas on where to go from here? I think I have done all the obvious things, but might have missed something?
Any help would be much appreciated.
Regards,
Dieter
12-28-2012 05:01 AM
Update:
After playing around with the power settings in the Windows power options, I went to the maximum performance setting with no speed increase. When switching to power saving and after a reboot I got the almost double the 4K throughput and about half the access times (both read/write +/-0.200 ms).
A second benchmark however didn't confirm this and ever since most of the time I get about the same results as the first screenshot in the first post, but sometimes access times as low as 0,200 ms. These better results seem to come at random times. Doing the benchmark twice with only the access time test enabled can yield 4 times 0.4ms and then one time 0.200 ms.
All the time I kill as many processes as possible, watch the resource center to check that there aren't any unneccesary services or processes writing to the disk. No swapfiles on the SSD.
The SSD is on the first SATA port. Connecting my 4 other (non-ssd) drives doesn't seem to produce a noticable difference.
Is there a reliable alternative to AS SSD Benchmark that runs under Linux (command line or GUI) to check if that gives any difference?
I have no idea where to go from here. Any thoughts?
Regards,
Dieter
12-29-2012 04:58 PM
What causes this is found here and applies to desktops.
http://www.storagereview.com/how_to_improve_low_ssd_performance_in_intel_series_5_chipset_environmen... http://www.storagereview.com/how_to_improve_low_ssd_performance_in_intel_series_5_chipset_environmen...
Mostly it down to the CPU idling as the 4K test is done.
01-12-2013 02:04 PM
@PeterUK: Finally had some more time to try it out.
The tweak for the drivers didn't make any difference.
The tweak for enabling the option in power settings and then preventing the CPU from going into idle made a big difference. I now have these values:
(the language is in German because the language files didn't get copied properly).
Much better. Still not great, but much better. Seems the CPU going into idle makes a big difference.
Then did some measurements with a decent power meter. Before changing the setting, the computer used about 130 Watts in idle.
In the same situation but with the setting changed, it averaged out around 200 Watts.
I still wonder about why the values are rather bad (especially the access times), even after preventing the CPU from idling.
Any more thoughts?