/ Sign-up
Your question

Raid 0 of 4 SSDs

  • NAS / RAID
  • SSD
  • Storage
Last response: in Storage
May 28, 2012 2:25:17 PM

Hello I Built a system with Raid 0 hopping to have a very high speed for files transfer ( and running apps in short time)

My system configurations:

CPU: AMD 8150 FX 3.6ghz 8Core voercloacked at 4.2 GHZ
MB: ASUS CROSSHAIR V FORMULA ( latest firmware )
GPU: Geforce 210 ( just to test the transfer going to add double sli of GTX 670 later )

and of the 4 SSDs of 120 GB CORSSAIR FORCE Saata 3 as Raid 0 ( total of 480 GB )

i set the raid as 120 bit, and installed windows 7

I am going to install windows later with Users folder on a regular raid 5 HDD to let my SSDs live longer

So i Tried to Copy with windows 7 a file of 12 gigs
the transfer Speed was about 350 - 600 MBs

and when i tried to copy the same file 4 times in the same time it gave me average of 600MBs

As i know each SSD speed of 550 MBs Read, 525 MBs write so i assume the raid 0 should be 2 Gig /s

also i did test with ATTO BENCHMARK:
- With Direct I/O result was:

-Without Direct I/O Result was:

Also I used HD TUNE PRO 4.61 and result was:

so as we see for atto benchmark read speed was about 1200MBs with I/O and 2100 MBs without I/O
and for HD tune Result was between 550-1000 MBs for reading

I saw a video on youtbe before talking about building 24 SSds to reach copy speed of 2gig/s

So my questions are:
1- Did i do something wrong or everything is fine and that how windows copy speed is...?
2- If 4 ssds wont go over 500 MBs then should i just go with 3 ssds ?
3- My ASUS MB only have 6 Raid and extra 1 Sata and 1 e sata but i it's a different raid controller can i link both of them? because what i want is to have raid 0 off ssds and raid 5 of 3X3 TB HDD for storage
4- if i have raid 5 of 4 SSDs will it work as fast as 3 SSDs in a raid 0 or not?

thank you for replying for my questions

More about : raid ssds

a c 352 G Storage
May 28, 2012 4:29:12 PM

First off, ATTO is a lousy bench mark for SSDs, was developed for HDD, and worst yet it uses data that is highly compressable. Yes This is what manufs use as it it gives an inflated (Higher) number, espacially for SFxxxx based controllers based SSDs.
PS I use AS SSD, which was designed for SSDs, and uses compressed data for it's benchmark.

If your file that you transfered consisted of a File that does not compress then you will not get near advertized performance. To check transfer speed you need to transfer between SSDs not between a HDD and SSD. The do NOT scale, that is two drives do not = 2 x one drive, maybe 60% and as Number of drives goes up scaling decreasaes.

Raid0 will NOT improve an OS + program drive in terms of loading OS not loading Programs.

NOTE you have an AMD chipset, so NO TRIM now, nor in the near future (intel should be coming out with a driver to support passing trim when in a Raid0 config shortly.

Just posed this about 10 mins ago inresponse to a raid0 question
As jitpublisher, I used Raid0 in all my IDE system ever sense they incoprorated it on MBs. It was great, and I never lost a single drive. I still have an E6400 (OCed to 3.2) that uses raid0 (Sata HDDs). Keep it for the Kids to play on when they vist. Stopped using raid0 when I got my first SSD.

Raid0 still has a limited place, and that is if you do ALOT of work with large file structures.
By nature these files take a lot of space and SSD are generally to expensive for most, ie 500 gigs -> 1 TB space. The Newer HDDs, with their much higher magnetic domain density, and in many cases the higher failure rate than there older models means that you must use HDDs made to work with raid.

Trim support, is probably the biggest negative for raid0. However Intell should be releasing RST ver 11.5+ which is suppose to allow trim to pass when drive is a member of a raid set-up (Back of mind in a month or two). AND THIS only applies if using a intel chipset!!

.. Improves Sequencial performance, BUT this is the Least Important parameter for an OS + Program drive. It is the small 4K random performance that is IMPORTANT and raid0 does very little to no improvement on this catagory - HENSE Raid0 does Very little to improve performance on a OS + program drive.
.. The improvement in sequencial performance is why I stated it would boost performance when working with LARGE file structures - But here it is limited to the read and write portion - as NO in-program performance is not dependent on SSD, or HDD.

Side comment, One reason larger SSDs are faster than their smaller sibling, is that the larger SSD uses two Little brothers in raid0 internally. (other reasons also exsit).

I have NO need for SSDs in Raid0, and I think in the vast majority of users a larger SSD is more sensible. SSD speed and HDD speed have the same IN-PROGRAM performance, and gamers see No Improvement in FPS when game is running - SO Placinging SSDs in Raid0 would do What.... That Said, I'm not totally opposed to RAID0, when most usesage is with large files, such as enconding/converting large Vedio files (.vobs are 1 Gig files, blue-ray up to 40 gigs for a single file.), Very large spreedsheets, Cad/Cam drawings, and working with hundreds of them 10 meg Photo bitmaps/jpegs. - ALL these, unlike os and program files, are Very large sequencial files - even a "Small" 10 meg Photo is 2,500 sequencial clusters. For Raid0 stay away from SF based Controllers, Recommend the marvel M4, Samesung 830, and Intel 510 (510 uses the marvel controller - the 520 uses the SF controller. - My reasoning is not just from reliability, but SF controlled get high Seq marks because of compressing the data, if data is non-compressable, performance drops off considerably.
End quote: Ref
a b G Storage
May 28, 2012 4:42:26 PM

Raid 0 with SSDs? pointless imo ... without raid 0 you can copy from 1 SSD from another with 500MB/s (depending on the files, speaking here about i.e DVD movie). AND considering you have chosen the Corsair Force 3 Series i have some questions about reliebility...
What would i do? well in worst case i would make 2 arrays (2 x 240GB) or i would buy 2 x 240GB SSDs ..
Related resources
Can't find your answer ? Ask !
a c 288 G Storage
May 28, 2012 4:48:38 PM

Synthetic benchmarks are designed to exercise the device to their maximum capability.
The benchmark app drives the device in a multithreaded manner and develops high queue lengths.
The i/o is overlapped with processing.

Unfortunately, that is not what the user app normally does. Windows copy, and other apps will do one thing at a time, checking that the results are comitted to the destination device before proceeding. That is for recovery purposes. You can disable that checking if you want to accept the risk.

In the end, I think you would be better off to use the ssd's as individual devices.
Can you try that for us?
Time how long it takes to do the copy with raid, and again just copying from one device to another.
Your best benchmark is YOUR app.
May 28, 2012 6:04:07 PM

I forgot to mention I was doing the copy to the raid 0 to it self,
does this means the speed was split to half ? due to copy and writting in the same time?
anyway I have an extra 240 gig SSD Force 3 with the same speedSSD i will try to copy it to and lets see
a c 288 G Storage
May 28, 2012 6:09:34 PM

does this means the speed was split to half ? due to copy and writting in the same time?

a b G Storage
May 28, 2012 6:30:22 PM

4 SSD's with raid 0? why not just go a level of raid that has some redundancy. i can understand using RAID 0 if you have 2 drives but with 4 it just seems like a bad move.