How are my results with X25-M RAID0?

[zlojack]

I am under the impression that my write times should be higher? Or is this what I am going to get now that my drives are about 80% full.

I'm running Windows 7 RC1 (Build 7100) x64

Write back caching is enabled.

System specs in sig.

[One_Hertz]

It is quite a lot slower than it should be. I don't know why. Stripe size?

[zlojack]

Stripe size is 128

Writeback cacheing is on.

I'm not running anything during the test and I've got about 32GB of free space on the RAID array.

I haven't really done much tweaking. I suppose I should? I read something about realigning? Should I look into it further?

[Savuti]

For x25 m raid0, the writes portion seems low. It might have to do with the remaining space, the degree of use the drives have undergone. As a reference perhaps you should consider doing a secure erase to return the drives to fresh state and do an image of fresh OS install then bench from there.

[pphx459]

You may just need the latest firmware



Think mine are a little low too, 128k stripe, about 90gb left. Using Windows 7 RC build.

[ACE76]

It's due to not having optimized drivers for Windows 7...I have X-25M drives and experienced similar problems...I just put my drives onto an Areca 1230ML controller and my scores shot up significantly in Win7 using Vista64 drivers...change your stripe size to 64 and try again in Vista64 to see how well they perform.

[zlojack]

Thanks for the tips, guys. I'm not sure I want to return to using Vista. I've been using Win 7 for a while now and I really like it and I'm used to it. Perhaps I just need to be patient and wait for updates?

[lowfat]

It's due to not having optimized drivers for Windows 7...I have X-25M drives and experienced similar problems...I just put my drives onto an Areca 1230ML controller and my scores shot up significantly in Win7 using Vista64 drivers...change your stripe size to 64 and try again in Vista64 to see how well they perform.

Never had that problem. My X25-M performed better on Win7 than it did on Vista.

[zlojack]

I guess I'll have to pull them and install them in the backup rig and run the benchmarks there.

[Ender17]

did you update the firmware?

[zlojack]

Yes, the firmware is updated.

[rge]

I am running same windows 7 build RC1, no raid just one x25m, it is about 60% full...your write speed is def low. Did you run iometer for long time?

[zlojack]

I haven't run IOmeter yet.

I'll be home in a short time and I'll try that next and post up the results.

[Ourasi]

I haven't run IOmeter yet.

I'll be home in a short time and I'll try that next and post up the results.

I don't think he ment for you to test it, only if you have done alot of IOmeter runs, wich usually slow down the drives...

[zlojack]

I don't think he ment for you to test it, only if you have done alot of IOmeter runs, wich usually slow down the drives...

Ah ok...

Well I've done exactly none

[rge]

yeah, sorry, should have been more clear, I accidentally ran a very long iometer run and killed my write performance...but that was about time when RC1 came out, so secure erase and reinstall fixed it. Hopefully trim support will be coming soon, maybe that will fix your write speed, especially if it is from being 80% full/use habits.

[zlojack]

Well, it looks like I've made some progress here.

I found a setting in Intel Matrix Storage Manager that I had not played with previously, so I ran CrystalDiskMark with four different configs and these were the results:

Config 1 (This is what I had been running) - Low write performance across the board



Config 2 with write-cache buffer flushing turned off. Pretty similar results to Config 1



Config 3 Much better results with Hard Drive Data Cache Enabled in Storage Manager (right click on the Array and Enable it) and write-cache buffer flushing turned off



And Finally, Config 4 with write-cache buffer flushing turned on and Hard Drive Data Cache Enabled in Storage Manager



I'm going to do some more tests between Configs 3 and 4 to see where it stands with other benchmarks.

EDIT:

More benchies
Alright! Now I think Config 4 will be the winner here

Here are a couple more benchmarks


Config 3





Config 4

[zlojack]

ATTO writes are still very slow, right?

[Rhys]

ATTO writes are still very slow, right?

No they are fine!
They are about double the speed of one drive. That slight blip at the end is probably a touch of degradation.
It should be cured with a new FW update before it becomes a problem.

x25-m's don't have very fast sequential rights anyway, so perhaps that's why you thought it looked slow, but your mainly going to be taking advantage of the drives read speed when loading apps ect. which the m will generally do slightly faster than the e, but you do suffer a sequential write penalty, which I have only found annoying if I am writing a large file to the drive, but i'm only using a single x25-m, so you will have double the throughput.

[zlojack]

Thanks Rhys. So it looks like I just had to enable that one setting in Intel Matrix in order to get things "up to speed"

[Rhys]

Thanks Rhys. So it looks like I just had to enable that one setting in Intel Matrix in order to get things "up to speed"

Yeh, write back cache got me to, when I first put my couple of Vertex's in raid 0.
Well glad you got it sorted.
You should stay tuned for when Trim will be released and more importantly (for you) does it support raid 0.
It should do, and when it does performance degradation should be a thing of the past.

Note: the x25-m is fairly good anyway at trying to combat degradation.

[zlojack]

Yeah, I am eagerly awaiting the new firmware. I was really hoping to avoid going through the whole OS rigamarole to test these drives, so I'm glad I was able to find the real problem. It's funny, because I enabled write back cache in Device Manager and also in Intel Matrix, but in that one spot on the "Array"

The two circled in Red were on but the one circled in Black was the one that helped make the difference for writes.

Interesting.