MARVELL SATA3 problem

[runn3R]

On a tip off from a friend I went to the Gigabyte site this morning and guess what I discovered. A little change has been made to the blurb about the capabilities of my motherboard that weren't there before! (http://www.gigabyte.com/products/product-page.aspx?pid=3263&dl=1#ov)

It now says:
* SATA3 RAID 0 4x performance is a maximum theoretical value. Actual performance may vary by system configuration.
* SATA3 SSDs are not recommended for use in RAID 0 mode on Marvell SE9128 ports.
(...)

Hi

FYI, the same change has been made to all Marvell SATA 3 MB's descriptions and this 333 website too:
http://www.giga-byte.co.uk/microsite/81/data/tech-091020-333.htm

Newer technology, better able to utilise the SATA3 interface will become available but, in the short term, unless you have bottomless pockets and can afford the High Spec SSD's then don't expect miracles.

I would like to comment on what absic has just said. According to the new "writing on the wall" even the new SATA3 SSDs won't work properly in RAID0 with the Marvell 9128 controller as configured on the motherboards. I can see no reason why this should be. If this is the case then how does Gigabyte get the advertised 4x SATA2 throughput?  I can understand the problem with getting the Western Digital Blacks to work in RAID0 as they do not support TLER but that shouldn't affect the SSDs. I wish Gigabyte would comment on this officially and clear it up.

I'll check why SSDs are not recommended for RAID 0 usage on Marvell.

[runn3R]

(...) how does Gigabyte get the advertised 4x SATA2 throughput? (...)

Regarding the bandwidth:
Single SATA2 is 3Gb/s
Single SATA3 is 6Gb/s, so it is 2x comparing with single SATA2 (3Gb/s)
SATA3 in RAID 0 mode is 12Gb/s, so it is 4x comparing with SATA2 (3Gb/s)

SATA3 RAID 0 4x and SATA3 2x performance is a maximum theoretical value.

P.S. It is the same like USB 3.0 10x performance (comparing with USB 2.0) which is just a maximum theoretical value calculated from the bandwidth. As you can see here:
http://forum.giga-byte.co.uk/index.php/topic,975.msg12153.html#msg12153
the current real performance ratio is around 3x while copying the files and 5x in burst speed (HD Tach)

[absic]

Thanks for the further info runn3R but with which HDD's or SDD's can we get this improvement as Western Digital Caviar Black are obviously not the ones to opt for! 

You also say the SSD's should not be used in a RAID 0 Array on the Marvell chipset. Any update forthcoming on this issue?

[absic]

UPDATE:

Today, I received from Western Digital, what I think is one of the best responses from a hardware manufacturer. They sent me the following e-mail:

Thank you for your reply.

I have received an answer from my support line.

Could you be so kind to send us test results with HD Tune when you connect them to the sata 3 ports?

- Test 1 Sata 3 drive in sata 3 Connection
- Test 1 Sata 2 Drives in Sata 2 Connection

Please send us a print screen of these results in order to continue investigating the issue for you.

If you have any further questions, please reply to this email and we will be happy to assist you further.

Sincerely,

So, not only do they not know why their Caviar Black SATA3 drives are performing so badly on the Marvell 9128 Chipset but now I have to invest my hard earned cash on different software that I don't have, to do further tests for them!!! 

Not only that but I also have to re-configure my PC which is now running in RAID 0 on the Marvell 9128 SATA3 Ports, quite happily with Hitachi HDD's as well. And another thing they haven't asked me to test the SATA3 drives on the SATA 2 ports where they actually perform better! 

To get to this stage has taken 3 e-mails and 11 days.

Next time you want to complain about Gigabyte's Technical support, maybe you should remember this!!


Rant over.......... for now!!!! 

[Dark Mantis]

Yes I totally understand your frustration with the answers received from this company. I have been using them for years and always reckoned them as being one of the top manufacturers of storage devices.  As you know I also tried to get a definitive answer from Western Digital regarding this problem and met with much the same response(after the usual inane questions like can I supply the serial numbers so they can look into the problem).
If I wanted to be a tester for WD I would have applied for the job! I don't expect to have to supply all the test data for them to then tell me the answer. It is just not good enough  
To cap it all if you don't reply within 4 days they call it fixed.

[absic]

Well, in the interests of fair play I have suggested they supply the software ($35) then I will do the testing. I await their response to that one!!

[F5BJR]

New firmware 1.0.0.1025 for Marvell 88SE9128

extracted to the new ASUS BIOS with MMTOOL for P7P55D-E-ASUS-1207

http://station-drivers.com/page/asus/asus%20p7p55d-e.htm

Download-Link: http://rapidshare.com/files/416237735/1.0.0.1025.rar

Send this to GigaByte suport !!

Pierre

[Dark Mantis]

Well found Pierre. Is it possible to use this as standalone software or does it have to be incorporated into the BIOS?

[F5BJR]

*
I have found the ASUS firmware version 1.0.0.1025 for 88SE91xx in a new BIOS

For the Gigabyte motherboards the firmware has 2 possibility selectable in BIOS setup :

1 - in BIOS

2 - in a separate eeprom  Macronics MX25L4005

That is not easy because requiert a new BIOS and a Service Pack for the Macronics eeprom

Pierre

[Dark Mantis]

Thanks for clearing that up for me Pierre, I had a feeling this was the case. I will pas this on to Gigabyte and hopefully they will make use of it in upcoming BIOS updates.

[F5BJR]

*

GA- X58A-UD5 Bios F6 released ..

http://station-drivers.com/page/gigabyte/ga-x58a-ud5.htm

-Enhance CPU, DDR, PCIex16/x8 compatibility
-USB3.0 chip legacy support
-Enhanced SATA3 RAID mode
Fix SMART Command Fail mit OCZ Sandforce FW 1.11

but not with 88SE9128 version 1.0.0.1025 Marvell firmware !!

Pierre

* some others new BIOS for GigaByte motherboards

http://station-drivers.com/

[Lsdmeasap]

*

GA- X58A-UD5 Bios F6 released ..

http://station-drivers.com/page/gigabyte/ga-x58a-ud5.htm

-Enhance CPU, DDR, PCIex16/x8 compatibility
-USB3.0 chip legacy support
-Enhanced SATA3 RAID mode
Fix SMART Command Fail mit OCZ Sandforce FW 1.11

but not with 88SE9128 version 1.0.0.1025 Marvell firmware !!

Pierre

* some others new BIOS for GigaByte motherboards

http://station-drivers.com/

The new "Enhancement for SATA 3 RAID Mode" isn't anything new, same old firmware from the older beta BIOSes, only in the new BIOS the users have have not already updated their firmware it's set to auto/force update it.   

That firmware in itself isn't very good anyway, the controller still freezes or drops drives (SSD's), nothing Gigabyte will ever be able to do to fix this, it's a Marvell chip issue.   The chips just can't handle the speeds drives are now offering.

Also, all these latest BIOS releases are flawed anyway, so it's suggested you wait to update for them to be corrected.   You can no longer enter the Marvell RAID BIOS, it keeps telling you that you are in IDE mode (No matter if you are in IDE, AHCI, or RAID).   You can build an array from windows using the MRU though, but even after that you still can't enter the Marvell RAID BIOS in the BIOS.   

So huge bug IMO, not an enhancement, and the error you are presented with is actually worded kind of rude too!  They definitely need a better translation for this error, and to correct when users are presented with it.

[Dark Mantis]

Hi and welcome to the forum. You are a long way from home though. What brings you to this side of the pond?

That firmware in itself isn't very good anyway, the controller still freezes or drops drives (SSD's), nothing Gigabyte will ever be able to do to fix this, it's a Marvell chip issue.   The chips just can't handle the speeds drives are now offering.

If the Marvell chip is pretty much cr*p then why did Gigabyte decide to put it on the boards anyway? I know there wasn't a lot of choice and I am sure they wanted to rush the product to market as is always the case but to include something that is known to be unsuitable and then extoll it's virtues in the advertising is ludicrus.

[ExOblivione]

I am interested in the UD3R board for my upcoming build. One of the main reasons I chose this is because it has SATA III support, something I want for the Intel G3 SSD I am planning to buy (upon release).

I hope to make a RAID 0 array with two of those, but reading the posts here has me worried. Is a solution to the Marvell problems even possible with a software update?

BTW, what about the AnandTech review of SSD SATA III speeds?
http://www.anandtech.com/show/2973/6gbps-sata-performance-amd-890gx-vs-intel-x58-p55/5

It shows the Marvell controller on the GB UD3R board to have great performance. Top in sequential read, and nearly so in others too.

[Lsdmeasap]

Hi and welcome to the forum. You are a long way from home though. What brings you to this side of the pond?

That firmware in itself isn't very good anyway, the controller still freezes or drops drives (SSD's), nothing Gigabyte will ever be able to do to fix this, it's a Marvell chip issue.   The chips just can't handle the speeds drives are now offering.

If the Marvell chip is pretty much cr*p then why did Gigabyte decide to put it on the boards anyway? I know there wasn't a lot of choice and I am sure they wanted to rush the product to market as is always the case but to include something that is known to be unsuitable and then extoll it's virtues in the advertising is ludicrus.

I was following this thread a few weeks back and just remembered to post my thoughts here on the testing I just completed, so hopefully Runn3r can report the RAID BIOS settings issue to GBT HQ.

I believe they used it because it was the only solution, and still is.    I am not sure if they were aware of the issues at the time or not?   Maybe they were and expected a firmware or driver update could fix them, maybe they didn't know at all?

I am interested in the UD3R board for my upcoming build. One of the main reasons I chose this is because it has SATA III support, something I want for the Intel G3 SSD I am planning to buy (upon release).

I hope to make a RAID 0 array with two of those, but reading the posts here has me worried. Is a solution to the Marvell problems even possible with a software update?

BTW, what about the AnandTech review of SSD SATA III speeds?
http://www.anandtech.com/show/2973/6gbps-sata-performance-amd-890gx-vs-intel-x58-p55/5

It shows the Marvell controller on the GB UD3R board to have great performance. Top in sequential read, and nearly so in others too.

The Marvell is ONLY good at Sequential read/writes with a single drive (And it can struggle with that under heavy load), other than that the Intel ICH10R is much better overall.  If you are going to RAID SSD's, the ICH10R will give you the best speeds for everything