CORSAIR Force SSD 40GB RAID 0 Hi-Speed test in dual PC platform

[windwithme]

SSD full name is Solid-State Drive.
It’s new booming hardware within 2 years. It really enhances the PC/NB transfer performance.
The main reason is SSD using NAND Flash instead of diskette design.

SSD normal capacity are 32/40/64/80/128GB. Recently, SSD market price is dropping a little bit.
The main reasons for users to buy are transfer performance; seek speed, anti-shock ability and MTBF.
Especially for NoteBook users, SSD is also lower power consumption and lighter.
However, SSD price and capacity are still the key factors to impact the selling ramping speed.



Related to SSD products, USA brand, CORSAIR, also put lots of efforts. They have 6 SSD product lines.
Per performance, it divides into
Force/Nova/Reactor/Performance/Extreme/Legacy series.
This review is CSSD-F40GB2, belongs to Force series. It has 3 years warranty.
CORSAIR F40 SSD position is similar to Intel X25-V 40GB. This review will focus on RAID0 in PC.

First to check the packing, CORSAIR FORCE SERIES F40


Product size is 2.5”, code is F40, and the highest performance is 280/270 MB/s.
Random 4K writing performance is 50000 IOPS. 4K writing performance is the weakness of most SSD.


Black metal shell and brushed finish surface look great.
Official weight is 80g which is lighter than 2.5” HDD.
The website showed the controller IC is SandForce SF-1222TA3-SBH.
MLC NAND FLASH is Intel 29F32G08AAMDB, total 10 pieces 4 GB chip.


The sticker above is warranty notice. The warranty will be invalid if it’s broken.


SATA supports SATAII which is for PC or NB.


This review uses desktop with different SB chipset to test RAID0 performance.
I believe this combination is very popular now and using SSD for booting drive.
And use big capacity 3.5” 7200rpm HDD for data drive.


Desktop uses Intel platform, GIGABYTE X58A-UD5.
SB is ICH10R, the most high end Intel chipset currently. I use it to test performance.


Built-in USB 3.0 and SATA3, 2 latest MB technologies.


Test Configuration
CPU: Intel Core i7-930
MB: GIGABYTE X58A-UD5
DRAM: CORSAIR DOMINATOR-GTX3 DDR3-2400
VGA: ATI Radeon HD 5830 CrossFire
HD: CORSAIR CSSD-F40GB2 Raid0
POWER: ANTEC HCG 620W
Cooler: Mega Shadow Deluxe Edition
OS: Windows7 Ultimate 64bit

[windwithme]

HD Tune Pro 4.60 Benchmark
Read - Average 520.0 MB/s Access Time 0.085ms
Write - Average 488.6 MB/s Access Time 0.017ms


Random Access


Extra tests
Sequential middle
Read - 522.673 MB/s
Write - 528.175 MB/s
Random seek 4KB 62779 IOPS


File Benchmark
ATTO DISK Benchmark 8k above transfer rate is 563.1 Mb/s at reading and 531.5 MB/s at writing.


CORSAIR F40 in HD Tune Pro test also has very high performance.
Both read/write are all over 500 MB/s which are very impressive transfer rate.
The seek time is below 0.1ms and even achieve 0.017ms. This is also very outstanding.
ATTO DISK Benchmark can also achieve over 560/530 MB/s, read/write performance. It really shortens the PC access time.

EVEREST Read Test Suite - Linear Read(Middle) 527.0 MB/s.
FDEBENCH gets the reading at 247.6 Mb/s and writing at 136.5 MB/s.
CrystalDiskMark gets the reading at 308.0 Mb/s and writing at 89.52 MB/s.


EVEREST Linear Read Average 526.2 MB/s
AS SSD Benchmark
Seq Read - 296.41 MB/s Write - 92.24 MB/s
4K-64Thrd Read - 198.54 MB/s Write - 86.68 MB/s


EVEREST Random Read 515.5 MB/s
AS SSD Benchmark
4K-64Thrd Read - 51293 iops Write - 20921 iops


CrystalMark Sequential Read 521.37 MB/s Sequential Write 495.65 MB/s
SiSoftware Sandra 481.72 MB/s


EVERSET many items are over 500 MB/s.
The others like CrystalMark、SiSoftware Sandra are also similar.
The most surprise is some items of AS SSD Benchmark and FDEBENCH didn’t reach the highest performance.
After repeating test and check the internet, I found it is caused by controller chip.
SandForce SF-1222 in other SSD products is also having this issue.

CrystalDiskMark 50/100/500/1000MB comparing test


According to the different file size, you can see the improvement of CrystalDiskMar Seq performance.
F40 high light point is having better transfer rate in smaller file. This is normal for high performance 4K file type SSD.

IOMETER 2008
4KB test - 100% Write 100%/Random 4KB
Total I/Os per Second 68653.32


4KB file IOPS writing performance is COSRAIR F40 strength.
The test is not achieve double performance as Raid0. It’s also possible the limitation of SB.
However, the benchmark is still over 68600. I will test with RAID card to compare this in the future.

[windwithme]

Another desktop PC test is in AMD platform.


Using GIGABYTE 890FXA-UD5, the SB is SB850, the most hi-end AMD Chipset so far. It also supports SATA3.


Supports ATI CrossFireX, On/Off Charge, USB 3.0 with 3X USB Power


Test Configuration
CPU: AMD Phenom II X6 1090T Black Edition
MB: GIGABYTE 890FXA-UD5
DRAM: CORSAIR DOMINATOR-GTX3 DDR3-2400
VGA: ATI Radeon HD 5830 CrossFire
HD: CORSAIR CSSD-F40GB2 Raid0
POWER: ANTEC HCG 620W
Cooler: Thermaltake V1AX
OS: Windows7 Ultimate 64bit


ATTO DISK Benchmark 64k above test can achieve reading at 446.3 Mb/s and writing at 445.2 MB/s.
EVEREST File Benchmark64k above test can achieve over 400 MB/s.


EVEREST Linear Read Average 422.3 MB/s
CrystalDiskMark reading at 240.9 Mb/s and writing at 106.6 MB/s.


AMD Raid0 has no “Write-back cache” item, so the performance is lower.
I hope AMD can add in this feature in Raid tool to enhance own chipset performance.
Of course, if we also disable “Write back” in Intel platform, the performance will be similar.

I also have some friends use SSD as NoteBook system drive and use 2.5” external HDD as data drive.
For desktop users can have higher performance with SSD as building Raid0 by 2 pieces 40GB SSD. The performance will be much higher than single 80GB.
However, the previous SSD version has lower performance issue after using for long time.
The most new versions SSD all support TRIM to solve this issue, but it’s only for single drive now.
Due to Matrix Storage Manager doesn’t support Raid0 TRIM instruction, the performance dropping issue is still exiting.
The solution is clear SSD Raid0 mode and all data. Then re-build the Raid0 to recover the original performance.
For Raid0 not support TRIM issue has none solution now. We only can expect Intel launches new Matrix Storage Manager to support it.
Just can let Intel chipset with SSD Raid0 users to really enjoy Raid0 performance and worry free for performance dropping issue.



Intel SSD product has very high level standard. The 4K writing performance is almost the best.
CORSAIR F40 uses Intel 29F32G08AAMDB and new controller, SandForce SF-1222TA3-SBH.
Except the high 4K writing performance, it also improves the overall performance of Intel NAND FLASH IC.
Force series can be the new generation high performance SSD spec. It’s also the key factor to ramp up the SSD market share.



CORSAIR Force series has 40/60/80/120/160/240GB.
The users can buy by budget. More SSD capacity provides more flexibility.
Force 40GB price is around 125US. Comparing to the other brands, the price is not too high.
There are more and more controllers in the market. It also improves the performance. At least the different chipset has different strength.
In the future, the price and capacity still impact the SSD coverage. I do want to see it happens soon.

Recently, I got a DC - Canon IXUS 200IS, official from Taiwan disti.
If I have time, I will make a preview. The content has outlook and experience sharing.
Thanks for your time to read this SSD review by windwithme.


It also post in my blog.WIND'S 3C Welcome you to come and comment.

[n-ster]

Nice intel performance! like my Vertex 2's

[mlee49]

Thanks for the details. These sandforce ssd's are crazy fast. I'll have two Nova's in Raid0 here in a few days to which I'll compare numbers with ya. I think they will be within 10-15% of yours.

Did it make any difference on you UD5 with the SATA6 slots? I would think that should help the throughput right?

[n-ster]

Aren't these Sata 2 SSDs? which would mean no difference?

[btarunr]

A friend tells me that the test between the two platforms isn't fair. The AMD setup is having 10~12 GB of random data preloaded into it (which cripples synthetic read-speed tests on SandForce-based SSDs), whereas the volume was run empty on the Intel platform.

So you connected just the RAID 0 volume on the AMD rig, installed the OS, and tested the same volume where the OS was residing, whereas the OS was installed on a different volume on the Intel setup, and you tested a completely blank RAID 0 volume there. How is that fair?