Part 2 of my SSD SATA III performance tests.
If you are here and haven’t seen part 1 “Sata 3 SSD running on onboard SATA II benchmark” go here >>>
Is there any reason to buy a SATA III expansion card to run your new SATA III SSD if your MOBO only has SATA II?
So I have a shiny new Corsair Force GT 120GB SATA III drive and an old SATA II motherboard.
Last article I benchmarked the performance to see what kind of experience you get when running on your onboard sata II connectors.
This article is the part 2 where I am going to benchmark the relative performance with a SATA III expansion card to see if there is any point is spending the money on one.
The Test Card – ASUS U3S6
It was a bit hard to track one down in Australia so I had to buy this second-hand off the overclockers australia forums, but I wanted this card for its x4 PCI-e v2.0 connection.
Details of the U3S6 here.
The U3 = USB 3.0 with a NEC chip set; and
The S6 = Sata III 6.0Gb/s Marvel 88SE9123 controller.
The card has yet to be flashed from its original bios which appears as 188.8.131.522.
I decided to play it safe and rather than flash the bios and destroy the card I would start from scratch, benchmark it first. Later I will play and see if I can improve on stock performance results, buy flashing to later versions of the firmware.
At on OS software level I tested with 2 driver configurations:
with some interesting results..
The latest drivers are best found on station-drivers. It’s french but version numbers can be read easily, go to the bottom for the MV91xx drivers and firmware.
The test rig
|Mother Board:||Socket 775, Gigabyte EP45-DS3|
|CPU:||Core 2 Duo E8400 3.0Ghz, running overclocked @ 4.0Ghz.|
|Memory:||4GB (2x2GB) Corsair CM2X2048-8500C5C (1066Mhz)|
|Video:||MSi R6850 Storm II 1G OC (R6850 PM2DIGD5)|
|SSD:||Corsair Force GT 120GB, (1.3.3 firmware on SATA II in AHCI mode)|
|SATA III Card||Asus U3S6 (firware 1012)|
|OS:||Windows 7 SP1 – 64 bit|
The U3S6 is a x4 card and will run on PCI-E v2.0. The EP45-DS3 only has 3 x PCI-e 1.0 x1 ports, and has a x16 and x8 PCI-e v2.0 slot.
The x16 slot is currently running my MSi HD 6850, and the x8 slot is free.
The U3S6 is therefore running in the x8 slot.
I was interested in seeing what might happen when running something like crysis whilst having the U3S6 trying to use the same bus.
My original SATA II Benchmark score on SATA II was 480! I did this benchmark on day 1 of a clean Windows 7 install. Since then I have been using the computer for a few months and the drive is about 50% full so I benchmarked with AS SSD again.
I did numerous benchmarks with in multiple different driver and SATA configurations. The Read speed seemed to give consistent results across similar benchmarks but the write performance (particularly the 4K-64-Thrd) was a bit all over the place and seemed to influence heavily the overall score giving results between 395 and 450.
The following results are the best of each run:
The U3S6 seems to have better sequential read of approx 340MB/s. This is only a 75MB/s improvement.
I noted that generally the benchmark results for the earlier 184.108.40.2066 driver were slightly better and more consistent than the later 220.127.116.116 driver.
Also the Read performance was generally consistent but the 4k and 4k-64Thrd write times were up and down.
Crystal Disk mark
Given the fluctuations I also tried Crystal Disk mark which seemed to give similar results without the randomness.
Again the 18.104.22.1686 driver seemed to out perform its older brother.
Windows Start Up Times
Same as last time, the benchmark recorded the time from when windows started loading after the POST screen, until a working desktop was available. This was identified as the time at which the Gadgets appeared on the desktop. This was split into 3 timing points
- time to login screen
- time to login
- time to see the desktop and the gadgets loaded.
I didn’t bother with the load time tests as I figured they would be comparable.
I did however run the Benchmarks for both Crysis and Crysis 2 in the various configurations, with some interesting results.
The crysis benchmark was run @ 1920×1080 – 64-bit and DirectX 10
The benchmark was run with the following SSD connections
- SATA II
- U3S6 22.214.171.1246
The results were almost identical. No change here.
Crysis 2 Benchmark
The benchmark was run on DirectX 11 on the Times Square Map.
The most interesting thing here is that when the video card was being pushed by crysis 2, the PCIe channel appears to be compromised by the SSD drive running on the x8 port.
This can be seen in the poor fps on the two higher settings.
Just to see how the USB 3.0 worked I hooked up my Western Digital 1 TB external hard drive and did a 4.14GB and 15.5 GB file copy both up to and down from the External drive.
I found the 15.5GB copy onto the external drive took 10m 1s @ USB 2.0 but only took 3m 52s on USB 3.0. The copy down from the external driver took 9m 11s @ USB 2.0 and 3m 43s @ USB 3.0.
The copy rates were approx 27.1 MB/s combined for USB 2.0 whilst USB 3.0 achieved 67.9 Mb/s or 2.5 times faster. Nice.
Why so slow?
I have gotten around to flashing the Card. At first it felt like “flushing” not “flashing” as the firmwares I tried off “Station-Drivers” broke the card. After a few hours of flashing I finally got the driver working with Firmware revision 1028 “Firmware pour U3S6 Rev 0 (MV-9123) Version:126.96.36.1998) Mod by Daoud333”
So when I start it up I see the following
PCIe x1 5.0Gbps. x1 why x1, its a x4 card running in a x8 slot?
After some digging I found this spec document on the controller, Marvell 88SE91xx Product Brief. If you read this the spec fo the controller shows this..
The controller only supports a single x1 connection.
I feel a bit duped. I specifically got this card for its x4 connection, but given the fact the marvel controller only accepts a x1 connection the SATA will only ever be able to run at this speed. I have read other people suggest that the x4 is used to split the channel in half and use some of the x 4 bandwidth for the USB 3.0 controller, which would make sense but it seems like false advertising to me.
Overall I think if you are out of SATA ports on your PC and need a few more, then the card is OK as long as you are NOT using it for gaming.
The sequential read is slightly better but this is only a benchmark figure as it did not translate in the real world.
The only other nicety is the USB 3.0 which will come in handy in the future.
I think in the next few months The Marvel 92xx controllers
will start to appear and these may allow my SSD to run to its full potential as this supports PCIe x2.