Sign in with
Sign up | Sign in
Your question

SATA 3 Marvell controller have issue with SSD?

Last response: in Storage
Share
November 10, 2011 7:36:21 PM

Hi guys,

i just purchased Motherboard and SSD but it seems the performance is poor comparing to Data sheet specification.

lets talk in details:

first im using motherboard GA-990FXA-UD3 (rev. 1.0) which use SATA 3 Marvell controller and SSD Patriot wildfire which rated 520MB write/Read .

all the SATA in the board is 6Gb/ Sata 3 and i installed windows 7 64Bits in AHCI Mode not IDE.

and im getting 470~490MB read with SA SSD benchmark and 170~250MB write.

how to improve the speed to reach almost factory specification ? do i need to replace the controller with PCI-E one?

any recommendation ?
a c 523 G Storage
November 10, 2011 7:55:31 PM

The Marvell controller is only used on the 2 eSATA 6Gb/s ports on your back panel.

The other 6 SATA 6Gb/s ports are run by the native AMD SB950 controller.

Your SSD is SandForce based so you need to benchmark with ATTO software to determine if you are getting advertised Read/Write speeds.

ATTO uses compressible data to test Read/Write speeds.
AS-SSD uses incompressible data to test Read/Write speeds.

You will always get lower Read/Write speeds on SandForce based SSDs with data that cannot be compressed.
m
0
l
November 11, 2011 6:18:04 AM

Dereck47 said:
The Marvell controller is only used on the 2 eSATA 6Gb/s ports on your back panel.

The other 6 SATA 6Gb/s ports are run by the native AMD SB950 controller.

Your SSD is SandForce based so you need to benchmark with ATTO software to determine if you are getting advertised Read/Write speeds.

ATTO uses compressible data to test Read/Write speeds.
AS-SSD uses incompressible data to test Read/Write speeds.

You will always get lower Read/Write speeds on SandForce based SSDs with data that cannot be compressed.



i will try to use the ATTO driver , but question does the controller make a big different to get the advertised speed?

there is any better than sandforce based SSD?
m
0
l
Related resources
a c 523 G Storage
November 11, 2011 1:49:14 PM

Yes, you will have better performance on the AMD controller than you will on the Marvell controller.

SandForce SSDs generally have the highest Read/Write speeds, but other brands such as Crucial & Intel have better reliability.
m
0
l
November 11, 2011 6:51:57 PM

Dereck47 said:
Yes, you will have better performance on the AMD controller than you will on the Marvell controller.

SandForce SSDs generally have the highest Read/Write speeds, but other brands such as Crucial & Intel have better reliability.



do you recommend to buy another controller? or stay with AMD?
m
0
l
a c 523 G Storage
November 11, 2011 6:58:06 PM

Run ATTO benchmark software and see what your maximum Read/Write speeds are.
m
0
l
a c 353 G Storage
November 11, 2011 7:16:15 PM

Dereck47 is correct about the "advertised" Sequencial performance. SF2281 controllers do very well with data that is readily compressable, such as what ATTO uses. - BUT THIS is not real life. AS SSD used compressed data which is closer to real life.

Some general comment.
(1) Sequencial read/write perfomance has only a small impact on "real life" OS + Program drive. What is important is the Random 4 K read/write performance.
(2) If you not, most reviews for sata III SSDs use a Intel based system - reason Sata III SSDs perform better on the Intel iaSTor driver compared to default msahic drive.


Bottom line - I do not look at Sequencial performance, I first look for Benchmark using PCMark vantage then AS SSD
m
0
l
a c 523 G Storage
November 11, 2011 8:12:09 PM

RetiredChief said:
Dereck47 is correct about the "advertised" Sequencial performance. SF2281 controllers do very well with data that is readily compressable, such as what ATTO uses. - BUT THIS is not real life. AS SSD used compressed data which is closer to real life.

Some general comment.
(1) Sequencial read/write perfomance has only a small impact on "real life" OS + Program drive. What is important is the Random 4 K read/write performance.
(2) If you not, most reviews for sata III SSDs use a Intel based system - reason Sata III SSDs perform better on the Intel iaSTor driver compared to default msahic drive.


Bottom line - I do not look at Sequencial performance, I first look for Benchmark using PCMark vantage then AS SSD



Yes, you’re correct that ATTO uses highly compressible test data, but I always suggest users use that to benchmark SandForce based drives so that they can verify that they "got-what-they-paid-for", and that their drive is not faulty.

But AS-SSD (and CrystalDiskMark) is not "real life" either. It’s the opposite end of the spectrum from ATTO.

If you’re a DJ working with only .wav files, or a video editor working with only .mpeg files, then you should not purchase a SandForce based drive. But most people don’t work strictly with files that are incompressible.

A way to check how compressible your files are is to right-click on a folder ("My Documents", "Program Files", etc.), select Properties, and write down the file size. Then use a file compression software (WinZip, WinRAR) to create .zip or .rar file of the folder. You can then divide before and after file sizes to see what percentage of your data is compressible. The higher the percentage, the better your performance will be with a SandForce drive.

PCMark Vantage is in-between ATTO & AS-SSD and a little more realistic.

There’s a new benchmark software being beta tested on XtremeSystems.org.
It’s called Anvil Storage Utilities. It’s new so it’s designed more with SSDs in mind.
Hopefully it will be used by SSD review sites in the near future.
Here’s the link: http://www.xtremesystems.org/forums/showthread.php?2736...
m
0
l
a c 353 G Storage
November 11, 2011 8:45:03 PM

You're correct on AS SSD, but it is closer to realistic for a OS + program drive as majority of files on the "C" drive have a pretty low compressability ratio.

And, as you indicated, when used as a data drive then it becomes a function of indivdual's files. Ie bitmap photos compress a lot more than jpeg photos.
m
0
l
!