Slow Performance on a RAID 1

15s isn't too bad. Try using Windows built-in RAID rather than the RAID controller's RAID functions. Windows built-in RAID always performs faster than that of fake-RAID controllers, and usually also faster than that of real RAID controllers, as it can read part of the data from one disk, and part of the data from the other simultaneously, which RAID controllers usually can't do, or not as well.

Remove software that isn't needed, and check what is autostarted, and disable what doesn't need to start automatically. You can use msconfig for that. Remove crappy AV software like Norton (Symantec) which wastes lots of your resources, and replace it with a good quality product, like Panda AV. If the PC is for private use you can use the free version. When installing, make sure you don't install extra stuff like add-on toolbars and firewalls. Your Windows firewall is good enough.

Make sure you have no Virus or other malware.

In addition to the above comments about RAID, remove one memory stick and run on one, then the other to see if there is an issue with memory.

It is Windows 7, so shut down, start up and press F8 at startup to start in Safe Mode (minimal items) and see if Safe Mode is faster.

I have a ThinkPad laptop, no raid, Symantec Corporate (not Norton Consumer) and 8 GB of memory. It is very fast and opens programs quickly.

The drives are ST2000DM001 I believe these are 4k drives.

Yes, they are 4K drives.  That raid controller does have 4K sector firmware update, but I did not see Win7 compatible 4K firmware update drivers for Win7.  (But you would have had to install the firmware before loading the O/S and used proper drivers  on initial installation).  

In any event, your O/S will do a much better RAID1, especially with 4K drives then that so-called controller.   So if problem is the RAID as you claim and not something else then my official recommendation is to go native software RAID and let the O/S give you the read load balancing and more caching then you can ever git with that so-called RAID controller.

According to this it looks like the drives are 512k. But the mfg specs say 4k
Screenshot-2015-07-26-10-08-13.png

They emulate 512b.  But they are 4KB.  Give the manufacturer of the HDD the benefit of the doubt that they are right ;)

The fact that the program says they are 512b supports the fact that the program you use isn't capable of discerning the difference.

You'll note also that that POS fakeraid controller is running the disks at a whopping' 3Gbit instead of 6gbit.  All the more reason not to use the fakeraid.

I have this exact same Raid Card in a Intel server that has been running fine and no performance issues for the last 6 years. The reason i used this card is cause i had a spare. Since it was working fine on the intel server i would assume it would be fine on this. My other Option was to use the Asus Pike Controller that is designed for this board. I appreciate all the info as im learning alot about this as i never really got into this area before. I was always against software raid as i hear bad things about it. But for the sake of learning i will try a software raid 1 as see if i notice a big difference. Also Before i go to a software raid I want to use Iometer to get a baseline test do you know what i should enter in iometer to run a test?

6 years ago you didn't plug in an AF disk.  The AF (4K) sectored drives require a read-modify-write if doing anything other than multiple of 4KB that are aligned properly.

Use true 512 byte sectored disks and your performance problem relating to AF disks goes away.

Every write I/O you do that is not aligned properly (which is going to be nearly 100%) requires reading other 1 or 2 4KB chunks, then modifying it in HDD cache, combining with the write request obtained from the initiator, then writing it back out to the target drive.  Think about that .. just write 512 bytes requires reading an extra 4KB and then writing an extra 4KB.  That HAS to hurt you.

so i ran iometer using the All in One Access Specification with 4 workers and default settings on the Intel server and im getting 187 iops i ran it on this ASUS box that uses the same raid card and im getting 53.55 iops. So since the Intel Server has true 512b disks that is why im getting better iops? and on the Asus box this is probably the issue "Every write I/O you do that is not aligned properly (which is going to be nearly 100%) requires reading other 1 or 2 4KB chunks, then modifying it in HDD cache, combining with the write request obtained from the initiator, then writing it back out to the target drive." As you stated. So if this is true then if i were to get disks that are true 512b sectors i should get better performance?

THanks again just trying to wrap my head around this.

So to close this. We can say that in order to get maximum performance, that Using the Intel SRCSATAWB is not optimized to work with the Seagate ST2000DM001 drives as they wont be seen by the OS as 4K Drives.

Because the disks are designed to emulate 512b, and the controller's firmware isn't even aware that there is such a thing as 4KB, so it doesn't know any better.

Basically the disks negotiate a fallback position of 512 for ignorant controllers.

So i ran a bench mark on all 3 scenarios. Can someone tell me which is the best option to use.
HDTune-Benchmark-ATA-----ST2000DM001-1ER
HDTune-Benchmark-INTEL---SRCSATAWB.png
HDTune-Benchmark-LSI-----MegaSR--.png

None, really,  as you can't run a benchmark independently of your O/S, so all results are flawed.   You need to take 2 disks, separate them from your O/S, and configure the disks to match the benchmark.   You don't have a megaraid, so the onboard LSI one is wrong.


Run the individual SATA benchmark off a single disk, then configure widows software RAID1, run the same bench on the logical RAID1 device, if it will let you.

Or run the HDTune for the intel as it is now, but configure the 2 disks as a separate RAID1.