X25-M vs (2) X25-M RAID 0

[Griff805]

I am reconfiguring my system and figured I'd get some "real world" differences between a single X25-M and 2 X25-M in RAID 0. If you're interested, below are the results.

Windows 7 Installation process:
I used a 16GB SSD drive that I have to install windows 7 from USB.
I removed anything from my system that was not necessary. I only had a monitor, USB keyboard and mouse connected. I disconnected the CD-ROM drive and any spare hard drives. BIOS options were for the most part, default settings. I just disabled the floppy drive and disabled any kind of ASUS features that I never use anyway.

Hard drives were installed in RAID mode both during single and dual drive setups. For single drive tests, the second drive was disconnected. I ran HDDERASE on each drive before install. Firmware is the updated firmware from Intel on both drives.

After Windows 7 was installed. The only things I did were to disable page file, and turned off "Windows write-cache buffer flushing on the device". For dual drive setup, I installed matrix manager and enabled write-back cache, since that is how the drives would be used in a typical setup. I didn't install any drivers, as Windows 7 had drivers for everything in my system and they seemed to work fine.

The timer started right as I clicked the Next button to continue to the next step of "Copying Windows Files". It ended when I saw the network connection status change to "connected" on the taskbar. I subtracted the time during any kind of installation questions during first windows startup. There was only about 10-30 seconds worth of installation questions.

For the Applications-
I installed them in the same order and in the exact same way for both configurations. All programs except Adobe CS4 were installed from an image on a secondary hard drive that was installed after windows install. Only Adobe CS4 was installed and extracted from the main drive/drives.

Alright, if you have any questions, let me know. Extract from these results what you will.. I have my opinions about RAID-0 for SSD vs a single SSD. I'll leave it up to you guys to make your opinions. I didn't not run these tests more than once each time. I did not have the time to do this, but I think a single run should be sufficient for the purpose of this thread. By the way- Application load times were within .1 seconds of eachother for the most part, so it really wasn't worth creating a graph for that.

System
Q9450 @ 2.6 (stock)
ASUS P5Q-Pro (ICH10R is the controller used)
8GB DDR2
4870X2

[One_Hertz]

Thanks a lot for the testing!

[WhiteFireDragon]

thanks, reconfirms my though of how almost useless RAID is with SSD. double the cost, but get less than 5% performance increase.

[One_Hertz]

thanks, reconfirms my though of how almost useless RAID is with SSD. double the cost, but get less than 5% performance increase.

You forget the extra storage . Sometimes cheaper to buy two smaller drives for more storage than one larger drive.

But yes, performance wise it is not much help. There would be even less difference if Griff did some testing of game loading times; one drive would sometimes win over two in those scenarios...

[Griff805]

I figured that with the program installs, they would be similar in what would take place during a game loading. There's large files and small files for game loading and app installs like the ones I did, so it should be similar in results.

[WhiteFireDragon]

You forget the extra storage . Sometimes cheaper to buy two smaller drives for more storage than one larger drive.

what do you mean by storage? using two X25-M in RAID still only effectively gives you 80gb of usable space even though the two drives add up to 160gb. it's still the same amount of usable space as a single drive.

[One_Hertz]

I figured that with the program installs, they would be similar in what would take place during a game loading. There's large files and small files for game loading and app installs like the ones I did, so it should be similar in results.

Program installs have tons of writes and game loading hardly does any writes.

[Boogerlad]

no.. That's raid 1. In raid 0, you add up both hdd's storage.

[Griff805]

Program installs have tons of writes and game loading hardly does any writes.

Yes, you're correct. I didn't think before I typed- I did figure though that the app installs would be more SSD taxing than a game loading. You're tempting me to do it all again though... It does only take 7 minutes to install Windows 7.

[NapalmV5]

2.66ghz ? lol ok take it to 3.6 then to 4.6

and youll see widening gap before your eyes

7 min win7 install ?

3 min xp install here

[virtualrain]

Wouldn't the media from which the OS and apps are being installed be the limiting factor in either case? If the install media is slower, of course, there will be no difference. Unless you are installing from a RAID0 pair of SSD's I would expect these results.

[Griff805]

2.66ghz ? lol ok take it to 3.6 then to 4.6

and youll see widening gap before your eyes

7 min win7 install ?

3 min xp install here

I'll up it to 3.5Ghz and post the results.

[Ao1]

Nice work Griff. What stripe size did you use?

Edit: did you use the 16GB SSD for OS and application installs?

[Ourasi]

Nice work Griff805. What stripe size did you use?

If he is on ICH controller@128kb stripe, it would be fun to see those load times on a 32/16kb stripe, all though install times might be hindered by other stuff, starting of services CD/DVD media etc.. If you got the time, a 1 vs 2 in raid0 IOmeter bootupconfig would be nice....

[PeterStoba]

thanks, reconfirms my though of how almost useless RAID is with SSD. double the cost, but get less than 5% performance increase.

Is this true for the lower end drives? I've seen people get better gains by raiding two OCZ drives than these..

[Griff805]

These results just reflect my setup, which I would think is a somewhat typical setup. I would also think that most people don't have multiple SSD in RAID 0 to install from. As far as raiding two OCZ drives showing better improvements, that is understandable because the performance of 1 OCZ (core series probably) drive may not be able to deliver the data fast enough for the intended program, so when you add a second drive it would show a gain(or larger gain).

[Ourasi]

These results just reflect my setup, which I would think is a somewhat typical setup. I would also think that most people don't have multiple SSD in RAID 0 to install from. As far as raiding two OCZ drives showing better improvements, that is understandable because the performance of 1 OCZ (core series probably) drive may not be able to deliver the data fast enough for the intended program, so when you add a second drive it would show a gain(or larger gain).

You will see some nice improvements in IOMeter bootup config, and some heavy app loading/games/utilities etc. Did some testing for myself with the stopwatch a while ago, and there was indeed some pretty good improvements...

[Griff805]

I used a 128K stripe on the ICH10R.

Which apps/games/utilities did you see "pretty good improvements"?

[Ourasi]

I used a 128K stripe on the ICH10R.

Which apps/games/utilities did you see "pretty good improvements"?

Tested with Photoshop loading (good improvement) + some I can't remember, and a few offline webpages I was working on (very good improvement). Loading of Eventwiewer (fully loaded with same amount of full logfiles, modded to be identical) (good improvement), Cod4 (*small* improvement), Crysis (do not recall) and America's Army mapload (good improvement), IOMeter Bootup config (only tested stripe vs stripe on this one, but comparing my 6800 total IOPS with single results, it looks to be very good improvement).... This testing was just for my own curiosity, and not for publication on any forums, but might do it all again if the "spirit" suddenly appear. Tested with 16/32/128kb stripes... Just added what I remember in (), the textfile with the numbers is long gone I'm afraid, appart from IOMeter and other benches, those I still have.....

[virtualrain]

These results just reflect my setup, which I would think is a somewhat typical setup. I would also think that most people don't have multiple SSD in RAID 0 to install from. As far as raiding two OCZ drives showing better improvements, that is understandable because the performance of 1 OCZ (core series probably) drive may not be able to deliver the data fast enough for the intended program, so when you add a second drive it would show a gain(or larger gain).

I'm still confused... what's the point of this?

As far as I can see, you've basically shown that copying files from a single drive to another single drive or a RAID0 stripe is the same. What am I missing here? I'm not trying to be a jackass, I just don't see the point when the source of your installs is the limiting factor in either case.

[Rhys]

I'm still confused... what's the point of this?

As far as I can see, you've basically shown that copying files from a single drive to another single drive or a RAID0 stripe is the same. What am I missing here? I'm not trying to be a jackass, I just don't see the point when the source of your installs is the limiting factor in either case.

Some people like to backup and store Installation ISO's on their hard drive.
But still, boot/load times might be useful even if they are the same.

[Ender17]

do some tests like this
http://www.anandtech.com/storage/sho...spx?i=3535&p=4

[lowfat]

Are you doing any more tests? I'd love to see some application launch times

[virtualrain]

Are you doing any more tests? I'd love to see some application launch times

Yes, that would be more useful in my opinion.

[Griff805]

Been at a business conference the last 2 days- I'll do some more benching today. The whole point of my testing was just to show that for my usage patterns, nothing is improved by using 2 drives in RAID-0. Even the application load times for the apps that I use. Everything is either instantly opened or takes at most 2 seconds, such as Photoshop CS4.

I am not trying to find the max throughput of 2 drives vs. one. Just trying to show average system usage with 2 drives vs. one. I'll do some more testing so we can pull some more conclusions out.