P45 ( ICH10 ) Poor Raid Performance ?

[EvanaTm]

I recently installed a new Asus P5Q motherboard based on the P45 chipset.

I setup two identical 160GB SATA Maxtor drives (8MB cache) in a Raid 0 configuration with the block size set to 128KB. I also have a Seagate 750 GB SATA drive (16MB cache).

Normally, I would expect the performance of my raid drives to at least be on par with my larger Seagate SATA drive. However, I've noticed that unpacking zip/rar files is much faster on my 750 GB Seagate drive, and two benchmarks verify my observation.

On PCMark05, my Maxtor sata raid 0 array scored 6350.
My Seagate sata drive scored 7393


I also ran HD Tach and noticed that the raid 0 array suffered major drops in the read speed. This was very noticeable on the 8MB file test (versus 32 MB), but was masked when I enabled the write-back cache option. However, the 32 MB file test still had major problems regardless of this option.

RAID HD TACH RESULT
NON-RAID HD TACH RESULT

Does anyone have ideas of why my Raid performance is worse than the single drive? The only thing I can think of is that the Raid drives have only 8 MB caches, while my larger Seagate has a 16 MB cache.

I am running Windows XP Pro w/ sp2, QX9650, 2GB 667 MHz.

 

[chizow]

Hmm that looks about right actually. You're comparing a considerably faster large platter drive to two older generation small platter drives in RAID. Judging from your average read speeds the RAID array does look to be faster, ~120 to 80MB/s. Do you remember what a single 160GB drive scored? If it had average reads of ~60MB/s it looks fine for synthetic results.

My own results are similar, with my Samsung 750GB with an average read of 83MB/s and my Raptor X RAID 0 array at 152MB/s. Before I set the Raptors up in RAID they scored about the same as the Samsung.

 

[EarthwormJim]

Those results seem perfectly normal, you're comparing two older drives in a raid to a fairly new drive.

Transfer rates have gone up by 50% (or more) in modern drives compared to the older sub 300gb drives from two or three generations ago.

It's not hard to believe those 160gb drives cap out at 70mb/s in single drive operation.

 

[EvanaTm]

Well, that may be true, but I really doubt it.

They are 7200RPM SATA-II compliant drives. They should be quite fast.


However, I think I found a more logical explanation for my troubles.

I was extracting a file from an archive on drive A and was placing it on drive A.
Whereas, when I compared it to drive B, I was extracting from A to B.

And, from B to A proved to be much faster, but still slightly slower than A to B.


This does not explain the PCMark05 HD performance...unless I need to install the PCMark05 application on drive B, and use the target as drive A. However, I would be a little surprised if the HD benchmark wasn't first loaded into memory prior to execution. It would seem like a skewed test if they are benchmarking the drive for read/write while also loading it with read tasks for the benchmark itself.

 

[EarthwormJim]

Originally posted by: EvanaTm
Well, that may be true, but I really doubt it.

They are 7200RPM SATA-II compliant drives. They should be quite fast.


However, I think I found a more logical explanation for my troubles.

I was extracting a file from an archive on drive A and was placing it on drive A.
Whereas, when I compared it to drive B, I was extracting from A to B.

And, from B to A proved to be much faster, but still slightly slower than A to B.


This does not explain the PCMark05 HD performance...unless I need to install the PCMark05 application on drive B, and use the target as drive A. However, I would be a little surprised if the HD benchmark wasn't first loaded into memory prior to execution. It would seem like a skewed test if they are benchmarking the drive for read/write while also loading it with read tasks for the benchmark itself.

Why not? My hd501LJ 500gb Samsung drives capped out at around 80mb/s and those are only 1 generation behind now. Those weren't slow drives when they came out either.

Those Maxtor drives are only 100gb per platter density. The Seagate drive you're comparing it to has nearly double the platter density.

You're really expecting too much, RAID doesn't always equal better performance. The bus that they're on doesn't matter either, SATA 300 didn't increase drive performance.

I guess you can try messing with the acoustic management on the Maxtor drives with Hitachi's Feature Tool.

For comparison, here's 4 of those Samsung drives in a RAID 0. Scaling is similar to your drives if you had two more imo (taking into account the Samsungs are faster).

http://smg.photobucket.com/alb...Intel___Raid_0_Vol.png - 4 Drive RAID 0.

http://smg.photobucket.com/alb...rk_SAMSUNG_HD501LJ.png - Single Drive.

You are essentially getting double transfer rates of one drive.

 

[The Odorous One]

The P5Q series also utilizes Drive Expert Raid. There are 2 special SATA ports on the board that run off of a SIL5723 controller on board. Just plug 2 similar hard drives into it and select Drive Expert Speeding in the bios. The drives will be automatically configured for Raid 0 and the OS will load with no need for drivers. Setup couldn't be easier and just as fast as ICH10R from what I have read. I am using it on P5Q-E and performance is excellent.