GA-x58a-ud5 rev 1 Mushkin Chronos 120gb ssd slow speed

[4trees]

Hello GB community.  I have been happy to date with my GB mobo but recently, while trying to setup my first ssd, i came across an issue i wasnt to happy about. 
Setup:
Ga-x8a-ud5 rev1
i7 920
OCZ 6gb ram
corsair 850tx
Win7 64bit
Mushkin Chronos 120gb SSD
Western Digital 1tb Black hdd

I found upon installation of the ssd that my speeds were not as high as anticipated or advertised by Mushkin (my best bench results using ATTO read/write 240/405 mb/s)  The advertised benchies by product are more like 515/550.  I contacted Mushkin support and this is what they said in the reply. 


1:37 PM (9 hours ago)

to me
Hi ,

You are using an X58 mobo, all the Intel SATA ports are SATA2(3Gb/s)
The SATA3(6Gb/s) SATA port uses the Marvell 9128 controller, that is as fast as that controller is capable of going.
The limitation in read and write speeds is caused by the available SATA controllers on your mobo.

Thanks
Sean

What this tells me is if I want to use my SATA3 I will be bottlenecked as a result of the Marvell controller. Write speed of 240 mb/s and read 405 mb/s.  Regardless, the performance of my Mushkin ssd is still far better than my previous disk hdd so its not like I wasted my money. I do feel that the Marvell controller Gigabyte is using is something that should be addressed in reviews and benchmarks. This is a rather serious limitation to the SATA3 capability.

If I am wrong about my assumptions please let me know. Any corrections or suggestions could only improve my situation. I am curious if any Gigabyte users out there are able to install the Intell Rapid Storage Tech driver instead.  Do I have alternatives that would speed up the Mushkin SSD. 
Thanks for reading this and any replys you might offer.

[Fatman]

Yeh there was issues with the marvel controller and most just used the intel port....The speed is a theoretical speed only but you wont notice the difference as it just comes down to milliseconds.

[4trees]

So the intel SATA2 ports are faster than the Marvell SATA3 ports on this board!!?  I now know the Marvell SATA3 ports share the PCI-E bus, but with only 1 SSD being utilized (no raid) I hoped the shared pathway wouldnt be so much a bottleneck as compared to the SATA2.  I mean, if it is, what's the point in having SATA3 at all?  This leaves me the impression that the Marvell 9128 Controller is more like a SATA3 "look-a-like" than a real SATA3 port. 

On that point.... Why is Gigabyte claiming 6gb/s in the specs on this motherboard (and many other GB motherboards with the Marvell 9128 SATA3 controller?  Clearly it is not anywwhere near 6gb/s and it would be impossible to do so.  I think this seems like false advertising.  It is more upsetting that Gigabyte knows about this problem. 

I assume in 2 years or so, when I do another build and am considering which motherboard manufacturer to purchase from, this misinformation on the Gigabyte specs page is going to play a significant part in my decision.  I mean what else do I have to go on as  a consumer other than the hardware specs on the product page?  If that information (in the specs) is false then Gigabyte is standing knowingly selling snake oil.  Two years ago the benchmarks on the reviews of this board weren't recognizing this problem because the ssds then weren't nearly as fast as they are today. 

[4trees]

The speed is a theoretical speed only but you wont notice the difference as it just comes down to milliseconds.

I agree the results will have very small differences one way or the other.  But (and this is going to be a big but) I didn't buy a $280 motherboard for results that were just satisfactory or adequate.  And I know Gigabyte was not advertising this board, or any other, as just minimally as good or better than any other board on the market.  They wanted this board to appear "top of the line" or a premium board.  At the time I bought this board it was among the top 3-4 mainstream boards on the market.  I bought it because it wasn't $400+ like some but it advertised really nice features that I wanted and the reviews and benchies were positive. 

I mean i could spend $100 on a motherboard and show only minimal real time changes in the speed of my pc.  Heck, I could just go to Microcenter and spend $400 on  a decent desktop!  But that's not what  my pc is to me. 

Ultimately the point and facts are...  I bought a motherboard that claimed to have SATA3 ports at 6gb/s speed.  The capable speed of these SATA3 ports on the GA-x58a-ud5 (and any other motherboard with the Marvell 9128 controller) falls significantly short of this speed.  Gigabyte is/was aware of the inability for these SATA3 ports to perform as spec'd.  Gigabyte is caught lying to its consumers. 

[Aussie Allan]

       Morning Gentlemen ...... To answer your question strictly .... switching the SSD to the Intel ICH10R Sata ported controller will improver your reads and writes substantially..... they won't get to the levels that the SSD is capable of , but it will be far closer then you are now.

  Alternatives: .... Aria have the Rocket Raid 640 PCI-E SataIII- 4 port Raid card on special presently for about £89 ..... these cards are bootable out of the box .... will get you to the 500Mb/s+ you're after .... The down side is it will add about 30 seconds to your boot time and it also carries the dreaded Marvell 9128 controller......although these PCBs carry not one but two controllers ...... this combination has proved to fix the bottle neck issue that plagues this controller on the 58 platforms on all boards, not just Gigabyte boards ..... I know this for a fact because I went through this myself and now own and use one myself to overcome this limitation.

  If you require anymore information or clarification ........Just ask.

  Aussie Allan

[Rolo42]

SandForce SSD controllers have had compatibility issues with Marvell HDCs.

Additionally, ensure you are using the same test the manufacturer is using in its rating (usually sustained rate since its most favourable on paper).

[4trees]

Thanks for the comments and info Rolo42. 

I have read the sandforce controller, especially the 2281 because it is relatively new, has had issues with compatibility.  Hopefully that will work itself out in firmware updates.  Aside from the speed the Mushkin ssd seems to be working normally despite Sandforce.  I used HD tune and ATTO to test for "health" and speed in a variety of tests. 

Mushkin uses ATTO for its speed tests.  I know ATTO has been used for a long time by manufacturers and it often results in high numbers (which makes the products look good of course).  I ran the ATTO standard test at least 10 times to get the best possible results, as shown in the OP.  So for better or worse I am using the same testing software for comparison sake.   

[Aussie Allan]

   ATTO was designed and developed for SSDs ..... but keep in mind ..... SSDs as a rule don't take kindly to being benched and benched and benched.....try to use it sparingly to ensure a long life ......  secure erase occasionally to keep it running optimally.

  Aussie Allan

[Fatman]

AS SSD is another benchmark that is used, however it is brutal for the sandforce.

   ATTO was designed and developed for SSDs ..... but keep in mind ..... SSDs as a rule don't take kindly to being benched and benched and benched.....try to use it sparingly to ensure a long life ......  secure erase occasionally to keep it running optimally.

  Aussie Allan

Yep that's bang on AA 

[4trees]

To Aussie Allan & Fatman....  That is news to me.  I am glad to know this and will limit anymore stress tests/benchmarks for the future.

I am left scratching my head though.  Why would it be damaging to the ssd to run multiple benchmarks?  Im trying to think of potential impacts, but aside from cluttering the write space on the drive itself with needless info, I am stumped.  Trim is enabled but trim doesnt delete unwanted info it only consolidates unused space.  Are the remnants of the writing the problem?  Basically I create a more dirty ssd?

If this is the case I would hope or expect that every once in  awhile I could effectively do a reformat of the entire drive?  Something like HDD Erase maybe?

[Dark Mantis]

SSDs have a limited lifespan unlike magnetic drives and so the more you do superfluous reads/writes the quicker it will wear out. Each area can only be written to a limited number of times.

A secure erase does help to keep the drive in prime condition.