OK Jafa i nabbed a copy of ATTO to run some tests. One thing i noticed though using different programs to bench, my read speeds appear to be capped by my raid. I am just using the onboard controller on my motherboard so reaching it's cap is to be expected honestly. Some programs seem limited to 100 MB read speeds, however ATTO is showing almost 120.

Here is a Screenshot of the results, hopefully there is something useful for you. If there is anything more i can do let me know.

Void4ever

Hi,

With a single drive, factory firmware, and write-cache enabled you should see >100MB/s sustained sequential write speed.

http://techreport.com/articles.x/15730/10

I ran a series of sequential read/write speed tests when I first received one of these drives... including a full-disk wipe. Now the drive has been updated I need to retest. I will post the results shortly.

Nick

The plot thickens...

Test machine has two drives, one with the OS and the other the DUT.

The DUT is a 1.5TB Seagate. One large partition but and no filesystems mounted.

Read test results are good: 129MB/s sequential read speed across the first 64GB of the disk.

Write test: (WARNING - do not run these commands!)

hdparm -W0 /dev/sdx

dd_rescue /dev/zero /dev/sdx -b 1M -m 64G

hdparm -W1 /dev/sdx

dd_rescue /dev/zero /dev/sdx -b 1M -m 64G

(ie write zeros over the first 64GB of the drive, first with drive write-cache disabled, then again with drive write-cache enabled)

With drive write-cache disabled the drive averages 101MB/s. 64GB in 11m08s

With drive write-cache enabled the drive averages 86MB/s. 64GB in 12m58s

ie 2 minutes slower to copy 64GB with drive write-cache enabled?

I have run the test multiple times with similar results.

Before the firmware upgrade I ran a similar test (full disk wipe)... it sustained >100MB/s with write-cache enabled.

Weird.

Shouldn't it be the other way around? That is odd.

Void4ever

I added the drive back into the RAID array (RAID 1 mirror) not reailizing that I left write-cache enabled...

After 20 mins I stopped it and disabled write-cache. The estimated numbers suggest it will finish 2 hours faster with write-cache disabled!

Nick

I'll run a test with that same program with write cache off, but i might not be able to get to it until later tonight, possibly tomorrow. Still it will be interesting to see the results.

Void4ever

Hmm less speed with cache on - not sounding good.

Both your test and v4e's are with a raid right?

Any chance for a single drive test with new firmware directly on the motherboard?

And if i remember jafa has a different firmware than v4e?

I could test with a nonpached SD17 edition... but which program is best? (winxp sp2 here)

---

Starlifino wrote:

Hmm less speed with cache on - not sounding good.

 

Both your test and v4e's are with a raid right?

Any chance for a single drive test with new firmware directly on the motherboard?

 

And if i remember jafa has a different firmware than v4e?

 

I could test with a nonpached SD17 edition... but which program is best? (winxp sp2 here)

---

No RAID.

The device under test was a single drive with no mounted partitions.

Nick

Your results does not match my "real world" numbers.

I realize my tests aren't optimal, as they have raid levels and file systems, but here they are:

All tests use the command "dd if=/dev/zero of=/path/to/fs/file bs=1M count=3000

Raid 1 - ext3 - two disks with the updated firmware SD1A

Cache on = 3145728000 bytes (3.1 GB) copied, 33.8445 s, 92.9 MB/s

Cache off =  3145728000 bytes (3.1 GB) copied, 86.5976 s, 36.3 MB/s

Raid 0 - xfs - two disks with updated firmware SD1A

Cache on =  3145728000 bytes (3.1 GB) copied, 19.2063 s, 164 MB/s

Cache off =  3145728000 bytes (3.1 GB) copied, 43.6608 s, 72.0 MB/s

I didn't bother rerunning the tests several times, or use a bigger test file,  since the results were so obvious and what I expected.

The disks weren't completely idle when running the tests either, but close enough.

I have hooked up SD17 and SD1A drives to Gigabyte mobo with Intel ICH8R SATA controller, port 2 and 3. Ran the ATTO benchmark and noticed no real differences. Drives maxed out at 130MB/s after 16kB stripes. What I did notice is actually improvement for 1k blocks, read/writes improved almost by 2x (from 7MB/s to 14MB/s)

I don't have the screen shots here so I can't post them.