unknownmember Posted January 10, 2012 Share Posted January 10, 2012 Hi there, This actually happened ever since I got them but never posted it coz I had no complaints about the speed but I'm curious tho as to why mine fluctuates a fair bit compared to Corsair's testing. Also if you've noticed my read vs write speed are.. weird I guess. The write speed surpasses read by miles at times and matches read at times. http://img214.imageshack.us/img214/9846/atto.th.png Uploaded with ImageShack.us http://www.corsair.com/blog/corsair-force-series-3-ssd-raid-0-performance-results/ http://www.corsair.com/blog/configuring-raid-0-arrays-with-solid-state-drives-intel-ich-and-pch-controllers Almost forgot, the drives have been flashed with 1.3.3 and the board is Gene Z Z68. Link to comment Share on other sites More sharing options...
Yellowbeard Posted January 10, 2012 Share Posted January 10, 2012 It's impossible to assist you without your system specifications. Please put them into your dropdown menu as requested at registration. Link to comment Share on other sites More sharing options...
unknownmember Posted January 10, 2012 Author Share Posted January 10, 2012 Soz, I did so just then but it doesn't seem to be showing up. Thanks Link to comment Share on other sites More sharing options...
parsec Posted January 10, 2012 Share Posted January 10, 2012 Very difficult to provide an answer without more information, and not just your PC specs, which I can see. How much space on your RAID 0 volume have you used to store software, etc? Or put another way, how much space is unused/available? Do you run Windows Disk Cleanup on that volume, and also remove all but the latest Restore point, if Restore is enabled? Did you keep some disk (really NAND) space unallocated out of the total amount available from the pairing of the SSDs? SSD garbage collection (GC) works better if it has more free space available to work with, even beyond what the manufacture (hopefully) over provisioned in the SSD's design. I don't know if GC is less effective when files are stored in RAID 0 format, or if it makes no difference. Of course, no TRIM for SSDs in RAID arrays (but there may be soon for RAID 0 arrays, go Intel!) Speaking of which, are you using IRST? Do you have Write-back Cache enabled? When you run the IRST software in Windows, does it say things are running normally, and all is well? Corsair's RAID 0 example might be on an empty volume, or one without an OS installed. If your's is the OS volume, that can make a difference. Do you consistently get the same results in the ATTO benchmark? Has it changed over time? About how many tests have you run? Sorry for all the questions, but at this level of technical complexity, simple answers about complicated situations are rare. Link to comment Share on other sites More sharing options...
unknownmember Posted January 11, 2012 Author Share Posted January 11, 2012 How much space on your RAID 0 volume have you used to store software, etc? Or put another way, how much space is unused/available? There's 136 free space out of 223. Do you run Windows Disk Cleanup on that volume, and also remove all but the latest Restore point, if Restore is enabled? I did ran it once after I finished with installing all the core components like drivers, updates etc. And no unfortunately all the restore points are removed and disabled. I've never really found much use for system restore, I find it lives to much mess behind. Did you keep some disk (really NAND) space unallocated out of the total amount available from the pairing of the SSDs? SSD garbage collection (GC) works better if it has more free space available to work with, even beyond what the manufacture (hopefully) over provisioned in the SSD's design. Nope, the array makes use of all the available space on both SSDs. I don't know if GC is less effective when files are stored in RAID 0 format, or if it makes no difference. Of course, no TRIM for SSDs in RAID arrays (but there may be soon for RAID 0 arrays, go Intel!) Only if the flux capacitor wasn't broken and if I had some plutonium....... Speaking of which, are you using IRST? Do you have Write-back Cache enabled? When you run the IRST software in Windows, does it say things are running normally, and all is well? Yep, it's 10.8.0.1003 and yep write back cache is enabled. Yep, the array has the green tick. Corsair's RAID 0 example might be on an empty volume, or one without an OS installed. If your's is the OS volume, that can make a difference. I guess in a way I do hope that's the case and on the other hand not because that means I can't do anything about it. Do you consistently get the same results in the ATTO benchmark? Has it changed over time? About how many tests have you run? Yep and surprisingly the results have changed a bit now. If I remember correctly I did about 3 tests at the time of posting the thread and just then, 4 tests. Here's a screenie of every second test http://img201.imageshack.us/img201/412/atto1.th.png http://img560.imageshack.us/img560/9117/atto2m.th.png Sorry for all the questions, but at this level of technical complexity, simple answers about complicated situations are rare. No probz lol, I understand where you're coming from. Thank you for taking the time out to help us out.:D: Link to comment Share on other sites More sharing options...
parsec Posted January 11, 2012 Share Posted January 11, 2012 Sure thing, thanks for the answers. Most of my questions were about the free space you have on that volume, and you have plenty. The Restore point thing was part of that, but if you don't use Restore, it's not an issue. I wasn't implying you could restore to an earlier point, that wouldn't do anything in your case. I wanted to see how much free space you had since that allows GC to do it's job easier. All the settings and status I asked about are fine. The new tests look better, which could mean that GC has had a chance to work it's magic. Or your PC was doing something during the previous test that caused it to show reduced performance. IMO, your results are great. Are you still concerned about it? I'm not sure if you are or not. The high write results, higher than the reads, which is unusual as you know, is probably due to the way the SandForce controller works. I'm not saying it is bad, just different than other SSD controllers. With all compressible data used in the ATTO benchmark, plus the SandForce controller, plus the RAID Write-back cache enabled, the write results are actually not that odd at all. But now I'm gonna stir things up a bit (not really) and suggest you try testing with AS SSD. Yes, the SSD Drill Sargent, that has no mercy and pulls no punches with all incompressible data, and displays the average result of several tests, not just the best one :evil:. For the "one number crowd" (ie, a SSD reads and writes any and all data in any size file under any circumstances at the same speed of 550MB/s) it is a spoiler, evil, and just wrong. I know that you know better than that, so for fun give it a try. I'm sure the results will be a little different in the writes (lower), but it will give you another perspective on your RAID volume that you can use for comparison purposes, if need be. I forgot to ask earlier if you are seeing or feeling the performance changing on your PC that would be related to your disk I/O performance. Link to comment Share on other sites More sharing options...
unknownmember Posted January 12, 2012 Author Share Posted January 12, 2012 Thanks parsec, It's probably is GC at work, because every time I do any bench on my PC I always make sure nothing other then what's set to load on start up is running and they're not doing anything. But no, I'm not really concerned about it. If there's a fix I'll jump at it but if not that's fine. I'd just like it to be more consistent that's all. I'm guessing write back cache is not enabled in the tests done by Corsair or else my results will be alike. Most probably is tho. Thanks for the heads up, I'll definitely give it a try. On that matter, can it be used for stability testing? Coz that's really the reason why I've updated the firmware to 1.3.3 just recently, which lead me to posting this thread after seeing the unusual results. Tho the results now is alike when they're running 1.3.0. I had random BSOD despite my OC passing 20+ runs of IntelBurnTest at ~97% used memory and the ram is set to XMP profile also...:sigh!: So fingers crossed the firmware cured it. Not really to be honest, I do noticed boot to be a wee bit faster. I mean I had to go from 3.3 base clock to at least 4+ to notice considerable speed boost lol. Link to comment Share on other sites More sharing options...
parsec Posted January 12, 2012 Share Posted January 12, 2012 AS SSD is not made for stability testing, it's purely a drive benchmark tool, low CPU usage, it hits the SATA interface and drives hard. Intel Burn Test or Prime95 target the CPU and memory, the only I/O they do is the test results, so they aren't stressing the drives and SATA interface at all. Since you have a Sandy Bridge CPU PC, your OC is not changing BCLK much if at all (IMO, leave it at 100 and just increase the Turbo multiplier) so the stability of the SATA interface that is using the BCLK frequency is unaffected. Changes in BCLK to OC could affect the SATA interface in the previous generation Intel CPUs, the i7-900 series, etc. So IMO, with this PC, BSODs during OC testing with IBT are caused by the CPU, but BSODs at idle are much more likely to be drive related. I was asking if you felt the PC was different due to the changes shown by the ATTO benchmark, but at those levels of performance, losing 25MB/s out of 400+ MB/s speeds is hard to notice. Such hardships... ;): I'm waiting for Intel to release the IRST driver that will allow TRIM in RAID 0. I sure hope it works. Link to comment Share on other sites More sharing options...
Garvin Posted January 12, 2012 Share Posted January 12, 2012 Looking at your screenshots, it looks to me like you have a program running in the background that's performing a scan while you're running Atto, problably an anti-virus rootkit scan or anti-malware scan. If Atto and the other program are running at the same time sharing bandwidth, your read results will be scewed just as you're seeing. Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.