Gigabyte P55-USB3 Intel RAID vs. GSATA2 or JMicron RAID in Windows Home Server

The following is a test of the Gigabyte P55-USB3 motherboard’s RAID capabilities.  The hard drives in use are not speed demons.  They are the Seagate Momentus 5400.6 ST9250315AS 250GB 5400 RPM 2.5″ SATA 3.0Gb/s Internal Notebook Hard Drive.  The real test is to see how the different RAID chipsets on the board perform.  These tests are merely a curiosity.  I doubt that there will be any noticeable difference on a functioning Windows Home Server but you never know.  As long as the RAID performs well and consistently I think I’ll be happy with trying this setup.  I will no doubt use the Vail option of removing data from the OS partition and only use this RAID as OS.  I have not researched this setup at all and have no idea if it is advisable or not.  Remember, it’s a curiosity test for the most part.

Intel vs. GSATA(JMicron)

image

A little information to begin with.  The Vail build for these tests is 7600.

image

Just for clarification, the Intel P55 chipset is running the 6 SATA ports shown here in light blue.  The Gigabyte SATA2 chip is running the two white connectors.  The GSATA is a rebadged JMicron chip.  The P55 has support for RAID 0/1/5/10.  GSATA RAID 0/1/JBOD.

image

Here you can see how VAIL detected the Intel chipset and loaded the appropriate drivers for it.  The install went perfectly and did not require an external disk or supplemental drivers.

RAID testing the array with no other hard drives present.

image

image

image

This test is on the Intel SATA ports, RAID 1 using ATT0246.

image

This is the Intel application to see what’s going on with the RAID.

HD Tune Test

image

There were no other drives attached to the system during these tests.  66.1 MB/Sec Max, Average of 45.3.

I then attached a 1 TB Hitachi drive and a 500 Gb Western Digital and added them both to the Vail drive pool.  I’m curious if having other drives attached to the same Intel SATA system will effect RAID performance or not.  I added the drives to the pool and am not doing any other requests to them.  I am not sure if Vail is doing anything on the SATA channels at the same time or not.

image

The Intel application showing the two attached drives.

image

The results are clearly slower compared to the first test.

image

54.6 MB/Sec with an average of 33.4 MB/Sec.

I’m not sure if Vail is doing anything in the background but it’s clear by these tests that having other drives attached to the same SATA system as the RAID is on has an effect.  I’m sure that there are more scientific tests that could be performed or tweaking of the system but this is how it loads up by default so I’m good with that for now.

Unfortunately, to test the JMicron/GSATA2 RAID capabilities I have to wipe everything and reload.  But before I do that I would like to test the RAID capabilities of the chipset so I popped one of the RAID drives out and brought it up on my workstation in order to format it.  The server booted up perfectly on the remaining drive.

image

Here is what the Intel RAID app reported.

image

The drive came up as two partitions.

image

H: Drive looks like the OS drive and I: drive the drive extender drive or the pool partition.  I fired up Computer Management, Disk Management and deleted both volumes.

image

Before

image

After

Just to make things more interesting, I’ll transfer some files to a WHS share on the server.  I’m not sure why, just to see what it will do when and if the RAID is rebuilt correctly.  I shut down the VAIL test server instead of hot plugging the drive.   Inserted the drive and powered it back up.  I see that the bios recognized the new drive and I could have rebuilt it in the RAID bios utility but I want to see the screen shots of what is happening inside VAIL so I let it boot up.

image

It immediately started to repair the RAID and although it took a while to finish it finally did without error.  I saw no issues with the shares and the data that was previously there or what was copied to the single drive during the RAID controlled failure.

Re-Install Vail on GSATA2 RAID

The install on the GSATA connectors as a RAID 1 was fairly uneventful.  I went into the bios and disabled the Intel RAID, enabled the GSATA RAID, deleted the RAID that it found on the drives and rebuilt it.  The VAIL install didn’t find any drivers for the GSATA so I had to use the driver disk that came with the motherboard.  This was the only difference between the two installations.

image

The ATTO Disk Benchmark was slightly better than the Intel.

image

HD Tune shows a 83.3MB/Sec maximum with an average of 61.9.  That beats the Intel RAID which had a 66.1 MB/Sec Max, and average of 45.3.  So much for all the reports that spoke of the GSATA2/JMicron being inferior.

That’s where the report ends.  JMicron edged out the Intel chipset.  There is more however.

I powered the server down in order to attach some drives to the Intel set of 6 SATA sockets just like I did above and turned it back on.  When I got back into Vail I noticed that the mirror was not up.  Only one drive was active.  I shut it down and reseated both drives and cables in case I bumped one while plugging in the others.  I reseated the 2.5” drives on another attempt. Still nothing.  I rebooted back into the RAID bios and tried to rebuild the mirror and it gave an error on several different attempts.  Tried to boot into Vail and it failed as well.  What happened?  I have no clue but it didn’t boost my confidence in the RAID or my 2.5” drive caddy.  Instead of re-installing again I choose to wait out the next Vail build and try again.

UPDATE: I think I found the problem.  When I loaded the two 2.5″ drives I grabbed two SATA cables from another motherboard box.  They came from the Zotac board that I used for my HTPC build.  I found that they wiggled around in the SATA connector on the motherboard.  Quite a lot in fact.  I have replaced them and hope that was the cause for the mirror to get out of sync and ultimately fail.

I’ve reloaded VAIL and am re-running the hard drive benchmark tools.  This shot above is against the RAID set on the JMicron/GSATA2 chipset.  I also have other drives connected to the Intel SATA chipset.  This is the test I didn’t get to run due to the issue above.  It’s in line with the first test meaning the drives installed on the Intel SATA connections have no bearing on the performance of the JMicron/GSATA2 chipset.

Here is the HD Tune test.  It shows 83.3MB maximum and a 65.8 Average.  This average is slightly better than before.

Share

Tags: , ,

Category: User Builds, Windows Home Server

Comments (10)

Trackback URL | Comments RSS Feed

  1. edcasati says:

    I’m always wondering just how useful any drive speed test is for a WHS. After all, it is only going to be accessed by the network, and even a slow drive is faster than the network under real world conditions.

    Another thing is that RAID, although designed to provide more reliability, is highly dependent on the controller, or in this case the motherboard. A hardware failure in a controller motherboard is typically a fatal error, resulting in unrecoverable data. Hardware upgrades are also nearly impossible to do.

    I would not suggest using RAID on a WHS system, unless it is RAID-1 (mirroring), because a WHS system is typically the primary storage, and backups of a WHS system are rarely done properly.

    Vail introduces all of the above problems into WHS, since it is no longer file based, becoming now ‘block’ based, and spreading single files across multiple disks, making recovery of a failed Vail WHS system quite difficult.

    • AlexF says:

      1Gb LAN (1,073,741,824b/s) has a theoretical throughput of 128MB/s. Practically, MAC, LLC, IP and TCP headers (and trailer in case of MAC) does reduce this, but the point is that RAID isn’t useless on a server.

    • AlexF says:

      Forgot to include SMB header in the list below.

      But you know, SMB (ie. file sharing) isn’t necessarily the end of it. If VAIL supports using its pools as iSCSI targets, (SMB headers can be replaced by iSCSI headers which I’ll guestimate are less) things get more interesting.

  2. pcdoc says:

    Nice write up Dave. Interesting comparison. Might be worth a shot to use two other drives straight off the controller to see if the issue repeats.

  3. Homeserver says:

    The speed tests were to test the RAID performance and not meant to disqualify one or the other due to speed issues. There are a lot of reports and posts out there about the difference between the two and the JMicron chipset always lost.

    I agree with all your statements. It's hard to go with a RAID setup on a system that is supposed to be redundant to begin with. This is just testing however. I have a new install on the JMicron SATA in a mirror and will continue to mess with it.

  4. [...] that are also configurable as RAID.  I’ll be comparing the Icy Dock to those tests and you can find that review here.  The summary of those RAID speed test were simple.  The JMicron outperformed the Intel set.  It [...]

  5. [...] Gigabyte P55-USB3 Intel RAID vs. GSATA2 or JMicron RAID in Windows Home Server [...]

  6. [...] did a head to head RAID comparison of the Intel ports vs. the JMicron ports on the motherboard in this article.  It is also required [...]

  7. [...] did a head to head RAID comparison of the Intel ports vs. the JMicron ports on the motherboard in this article.  It is also required [...]

  8. I wish I had PC now after reading your comparison, unfortunately I gave up to PCs and switched to the competition.

Go Ahead, Speak Your Mind

You must be logged in to post a comment.