Information

Dear forum members,
We are delighted to inform you that our forums are back online. Every single topic and post are now on their places and everything works just as before, only better. Welcome back!



Discussion

Discussion on Article:
RAID 0 of SSDs: Two Kingston HyperX 120 GB SSDs vs. Kingston HyperX 240 GB SSD

Started by: kokara4a | Date 01/17/12 08:02:09 AM
Comments: 7 | Last Comment:  01/20/13 07:33:57 PM

Expand all threads | Collapse all threads

[1-5]

1. 
I would have liked to see them compared against software RAID as implemented by the Windows Volume Manager, for instance. This has the advantage of exposing the TRIM command of the disks to the upper layers. The same goes for NCQ. I highly doubt that the impoverished RAID controllers like those built in the chipset expose NCQ for RAID arrays. But I may be wrong.
0 0 [Posted by: kokara4a  | Date: 01/17/12 08:02:09 AM]
Reply
- collapse thread

 
AMD's Raidxpert lets you set NCQ and a few other individual drive parameters. But I think by default they are switched off. I do remember getting superior performance with the Windows Volume Manager before adjusting the Raidxpert.
1 0 [Posted by: jonup  | Date: 01/17/12 01:31:04 PM]
Reply

2. 
Software Raid0 would enable TRIM, but would it make a noticable performance impact?
0 0 [Posted by: Simen1  | Date: 01/18/12 01:25:22 PM]
Reply
- collapse thread

 
Yes -- software RAID 0 of 2 disks is easily a benefit on most systems. It will partly be influenced by your CPU speed, but not nearly as much as a software RAID5/6 solution since no computation is involved.

The only downside is you will have some (likely insignificant) greater CPU usage to mangage the RAID 0 in SW, but on modern multi-core processors (espec 6 Core), I could easily see a software benefit for up to 4-5 disks in a RAID0.

I too wish this article had a 'clue' when it comes to testing -- they need to test like with like. Test the perf of 2x240's vs. 1-240 -- not 2x120's as the article points out -- ignoring the TRIM issue, the 120's use a different and slower technology. So the artile is very misleading about how close performance is.

Right now, I have a 4xRAID0, that doesn't know about TRIM, and my performance is about 1/2-1/4th what I would expect -- but if what they say about no TRIM support killing performance, I guess I'm lucky to be getting that -- my RAID0 is run through the BIOS which doesn't have a clue about TRIM AFAIK (was setup long before I knew what TRIM was or how important it is).

Another thing not tested -- some newer drivers Samsung and Plextors, have changed their algorithms to minimize the 'damage' of not using a TRIM -- resulting in performance that is near new levels all the time.

Compare this negative article on RAID0+SSD, with a Tom's article @ http://www.tomshardware.c...ssd-raid-iops,2848-5.html, where the author writes: "Each added SSD not only increases capacity, but I/O performance grows strongly and steadily as well. The test system does not seem to be nearing any upper limit of I/O performance, but scales with the same amount for each drive added."

This is a very different statement from what this article states -- completely opposite in fact. But noticed the article dates... The naysayer article was written Jan 2012 -- the article saying they scale linearly with # of SSD's with no downside was written March 2012.


Take aways:
1) Mileage likely varies on whether or not the SSD's are 'corporate grade' that are designed to deliver constant performance vs. consumer, 'peak performance' -- they'll be more likely to maintain performance in absence of TRIM usage (IMO).
2) Will be difference based on technology -- was it designed to compensate for non-trim usage?
3) a big issue -- are you comparing like w/like? same size disks to greater
numbers of the same-size disk -- NOT double # of half-size disks which will skew results.
4) if you can only afford 240 in 1 package or 2x120 in a RAID0, the 240 in 1 is your better bet - as later you can add another 240 and move to a RAID solution even if it is software. Looks like the best of all worlds is to buy an internal LSI controller to dedicate to those disks and skip the OEM BIOS managed raids -- I have a feeling those are the bottom end of performance (yeah -- like what I currently have ;-) -- hey win7 rates it a 7.9/7.9, BUT I know the difference -- and in my own benchmarks and useage 4x15K SAS drives beat my 4xSSD's in real performance (though Win7 only rated them at 6.3/7.9). But then I didn't put swap on those HD's, -- had a 5th HD (SSD) I used for swap).

I haven't switched back to the 15K SAS drives -- why? HEAT... they ran HOT...
using alot of power (I also, BTW, short-stroked them to get faster perf), -- but the new SSD's.. barely warm and no noise.... Now if I can just figure out a power solution for my T7500(Dell) and GTX-590 (it needs 2x8pin, Dell's highest end workstation provides 2x6+1x8, with 1 of the 6's on the same rail as the 8. So the Video card only gets 75% rated power. Dell should be drawn and quartered for designing a system that has cables on the same rail that would both be used for the video board ... LAME!

(my 2 cents and based on more common sense and limited Software RAID experience with HD's)....

Astara

0 0 [Posted by: Astara  | Date: 04/21/12 02:43:30 PM]
Reply

3. 
Great work.
One thing though - also testing a software Raid 0 would have been usefull. Because it is my understanding that a software raid 0 will be able to use the TRIM command and thereby avoid the harsh reality of degradation.

If you still have those drives pls do yet another test this time in software Raid 0 and then compare the results - that might just give us a whole new story alltogheter and might alter the conclusion part as well.
1 0 [Posted by: peter-pan  | Date: 01/18/12 01:56:36 PM]
Reply

4. 
Thanks for the tips. I made the mistake of trying to stripe Raid 0 in the bootup process rather than in Intel Rapid Storage itself. Took a few goes to work out my mistake.
0 0 [Posted by: manicsplendor  | Date: 04/27/12 05:37:36 AM]
Reply

5. 
It is worthy of note that Intel 7 Series chipsets with Windows 7 and Intel RST v11 now support TRIM on RAID 0. It doesn't exactly invalidate the results of this test, but for systems that support it, it does remove the "Doesn't support TRIM" caveat and tilts a bit more toward RAID 0 for SATA 3 SSDs

"Trim on RAID 0 for SSDs is supported in the Intel RST driver versions 11.0 and newer. Currently available for the general public on Intel’s downloads site is RST driver version 11.2 which offers TRIM support on RAID 0 compatible with MS Windows 7 OS on Intel 7 series chipsets (earlier chipsets NOT supported). Intel is also working on a future release providing support for TRIM on RAID 0 on Microsoft Windows 8 OS for Intel 7 series chipsets."
0 0 [Posted by: spiel  | Date: 01/20/13 07:33:57 PM]
Reply

[1-5]

Back to the Article

Add your Comment