• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Intel SSD 730 Series 480GB benchmark request

E

EnderW

[H]F Junkie
2FA
Joined
Sep 25, 2003
Messages
11,464
I've been testing a couple of Intel SSD 730 Series 480GB drives over the weekend and have some interesting results on ATTO Disk Benchmark.

I'll be posting my findings soon, but until then, anyone who has a Intel SSD 730 please run ATTO and share your results, including your motherboard, ATA BIOS setting (IDE, ACHI, RAID), SATA controller and driver version.

I'm especially interested in RAID-0 configurations. Please include your stripe size if posting those. Thanks :)
 
I don't have ATTO or access to run it at the present as I spend most of my time in Linux doing development/virtualization work, however, I just purchased a set of these a couple weeks ago when Newegg was running a daily deal on them. Anyways, when I first installed them I ran a quick Crystalmark benchmark against my 3 or 4 year old (can't recall exact purchase date, but its been a long time now) OCZ Vertex 2 120GB's and this is what I got:



Using the Gnome bench under Linux, I got this:



And with a "real world" copy of my game data from the OCZ's to the Intels I was getting this:

37T5WuCm.png


...which I think was more of a limitation with the OCZ's read speed, but still not bad.

Both are RAID0, but stripe sizes are different. I believe the OCZ's are 128k and the Intels I left at the default 64k for this go around.

These drives are super nice, and Newegg is running another deal on them today starting at 10am PDT. I'm considering grabbing 4 more if they'll let me and toss in my VM server as a 1+0 array.
 
Last edited:
I've maxed out 12Gb/s raid card with 730s, they're awesome drives. (I have some for sale too).
 
Last edited:
EnderW said:
do you care to enlighten us as to why we shouldn't use ATTO?
Click to expand...

ATTO is only sequential performance, it doesn't really adequately relay the performance of an SSD vs a good hdd raid.

It is old as dirt and usually showcases best-case scenario performance.
 
Kaos_Drem said:
ATTO is only sequential performance, it doesn't really adequately relay the performance of an SSD vs a good hdd raid.

It is old as dirt and usually showcases best-case scenario performance.
Click to expand...
I'm aware of those limitations. I'm using it to explore the poor <4K block performance on SSDs with Intel's 3rd generation controller which can be seen here
http://www.anandtech.com/bench/product/1141?vs=428

This seems to be mitigated by using multiple drives in RAID-0 (likely due to controller caching), but results seem to vary from chipset and driver version. I'll post more soon.
 
A good way to measure just how many sub 4K accesses your drives do on a daily basis would be to use the drive monitor functions of HDTune 5.0 and higher to leave it running during your regular workloads. You might be surprised just how rare the accesses below 4K actually are.
 
EnderW said:
do you care to enlighten us as to why we shouldn't use ATTO?
Click to expand...

Synthetic SSD benchmarks are universally ridiculous, but doubly so for ATTO, with the highly-compressible workload that it generates- not specifically an issue for the 730, but it's a spurious test overall. It's beyond best-case scenario. It has no basis in actual workloads.
 
Kaos_Drem said:
A good way to measure just how many sub 4K accesses your drives do on a daily basis would be to use the drive monitor functions of HDTune 5.0 and higher to leave it running during your regular workloads. You might be surprised just how rare the accesses below 4K actually are.
Click to expand...
Thanks for the info. I expect it is a very small percentage of use for typical workloads including my own use. I just found it curious the Intel 3rd gen controller SSDs perform so poorly (significantly less than even a 5400 RPM hard drive) in comparison to other SSDs in this one area when they are nearly equal on all other block sizes.
 
Chris_Lonardo said:
Synthetic SSD benchmarks are universally ridiculous, but doubly so for ATTO, with the highly-compressible workload that it generates- not specifically an issue for the 730, but it's a spurious test overall. It's beyond best-case scenario. It has no basis in actual workloads.
Click to expand...
Yes I'm aware ATTO is not an appropriate way to measure a drive's overall performance. I'm using it specifically to see throughput at <4KB block sizes.
If you have a recommendation for another tool that measures that, please share, although I would note an advantage of ATTO being it has been widely used for several years and thus extensive previous data exists for comparison.
 
Also since I've been looking into this, I've become more interested in the difference in RAID performance between Intel chipsets and RST driver versions
 
EnderW said:
Yes I'm aware ATTO is not an appropriate way to measure a drive's overall performance. I'm using it specifically to see throughput at <4KB block sizes.
If you have a recommendation for another tool that measures that, please share, although I would note an advantage of ATTO being it has been widely used for several years and thus extensive previous data exists for comparison.
Click to expand...

Iometer.

Also, if you believe that data from ATTO is actually useful as a comparative tool, then good luck to you.
 
Chris_Lonardo said:
Iometer.

Also, if you believe that data from ATTO is actually useful as a comparative tool, then good luck to you.
Click to expand...
Thanks for the suggestion, I'll check it out.
Can you expand on what you said about ATTO not being useful for comparison? I've used it very little but seen it in most SSD reviews.
 
EnderW: It would be interesting to see if the physical sector size can be changed on your 730 like it can on the S3500, they are the same drives after all.
I don't have high hopes though, Intel is famous for not updating "old" drives firmware.
It's probably the best option for forcing Windows to not perform I/O transfers below 4KB in size...

http://www.intel.com/support/ssdc/hpssd/sb/CS-035064.htm
 
EnderW said:
Thanks for the suggestion, I'll check it out.
Can you expand on what you said about ATTO not being useful for comparison? I've used it very little but seen it in most SSD reviews.
Click to expand...

I'll be addressing this topic at length in an upcoming article/tirade, but in brief, the compressible workloads generated by ATTO make drives using write compression look way, way better than they should. Also, it's a sequential test, when you're really not going to have much sequential sub-4K I/O in the real world. To the extent that you have sub-4K I/O, it will probably be mostly random. Other synthetic benchmarks have other limitations, but ATTO is probably the least realistic of the big ones for these reasons.

Also, the fact that the test is so brief means that you're not really doing anything representative with the drive. Anandtech has had some great articles addressing performance stability over longer benchmark runs, and a quickie test like ATTO steps on exactly the landmines that they talk about.

ATTO is so widely used because it's quick, easy, and fills pages with rigorous-seeming graphs. Getting a meaningful assessment of SSD performance is far more nuanced than most reviewers want to admit. ATTO is lazy. ATTO is wrong.
 
You must log in or register to reply here.
Back
Top