New Marvell driver

Discussion in 'SSD Benchmark Locker' started by Igor, Apr 3, 2012.

  1. Igor

    Igor Guest

    New Marvell driver came,so i try it,it seems stable.I been on msahci where"Turn off Windows write-cache buffer flushing on this device" didnt help anything so i try it now.That and enabling "bypass" under Marvell mode control in the BIOS give me this pretty,pretty good scores for Marvell 9128.
    as-ssd-bench M4-CT128 M4SSD2  2.4.2012. 20-44-13.png

    I would love to see any Intel chipset score in Marvell position on the motherboard.:p
    If somebody want to try driver
    Last edited by a moderator: Apr 4, 2012
  2. Igor

    Igor Guest

    Everything was beautifully till i decided to flash Marvell 9128 chip with new firmware.I made big mistake,forgot to disconnect SSD from Marvell chip.(read somewhere its got to be disconnect)After that my benchmark results are unstable,especially seq. Read which drops sometimes to 280mb/s.I reinstall,revert bios,but everything is the same.
    And what is very weird when i reinstall msahci driver everything is just fine as it was(same results,little smaller then with new marvell driver,when it work fine).I even reinstall marvell driver and flash SSD to new firmware 000F,because i was thinking that i screw firmware of SSD.It seems to me like Marvell driver for some reasons dont wont any more to implemented normally to Marvell 9128 chip(controller).I would appreciate any help
    as-ssd-bench M4-CT128 M4SSD2  20.4.2012. 14-56-11.png
    as-ssd-bench M4-CT128 M4SSD2  20.4.2012. 14-10-54.png
    as-ssd-bench M4-CT128 M4SSD2  20.4.2012. 14-07-17.png
    Last edited by a moderator: Apr 20, 2012
  3. Hameister

    Hameister Guest

    Thanks for the post Igor. I went to Asus support because I have the lousy 9128 Marvell controller on my Rampage III Formula motherboard.

    Asus is still recommending the original driver from March of mention of the new driver even though it is WHQL certified.


    I'll try this new driver this weekend when I return from vacation, and report back here on how well it works for me.
  4. The SSD Guy

    The SSD Guy Administrator Staff Member

    Likes Received:
    Trophy Points:
    Read it over Igor and thoughts here.
  5. OS-Wiz

    OS-Wiz Guest

    Igor, based on all I've seen, the msahci drivers would be best in your situation given your mobo and Marvel chip. You could do a DISKPART > CLEAN ALL of the SSD (given you have a backup), reflash in the 000F firmware, and then see how things go. You might wait for Hameister to get back and see how things go for him.
  6. Hameister

    Hameister Guest

    Sorry Igor, I have no better answer for you than Les.

    I have very little knowledge of the Marvell 9128 chip, other than the fact that I hate it. I haven't heard of an SSD issue when flashing the 9128 with the SSD connected.

    The only thing I can suggest is Google your issue. Perhaps someone has found a fix for it.
  7. Igor

    Igor Guest

    Ye,but before i flash Marvell chip,this new Marvell driver was very,very good.I know that this problem will be difficult to solve, but shoot at it that maybe someone has experienced this or something same.What do you think people about "must disconnect SSD before flashing Sata(marvell) chip"
  8. Bad_Machine

    Bad_Machine Guest

    I don't have a definite solution either Igor, I'm sorry but I'm quite new in the SSD world myself. If I were you I'd do what Wizzer said, do a full backup, secure-erase the SSD, re-flash it, then restore the backup and benchmark again. You could also try doing a fresh Win7 install (after secure-erasing and reflashing the SSD), and then bench again with only the Gigabyte recommended drivers installed and no other software added. It's just a hunch though.
    Last edited by a moderator: Apr 20, 2012
  9. Igor

    Igor Guest

    I am back with fresh install of Windows,and for now everything is good and numbers are stable again.I could not believe that flashing firmware could screw OS.
    Thanks to everybody that find time to reply.
  10. Bad_Machine

    Bad_Machine Guest

    Looks like my hunch was correct after all lol... I'm glad it's all sorted man!
  11. Igor

    Igor Guest

    O and i forgot

  12. Hameister

    Hameister Guest

    Glad you resolved the problem Igor.

    Computers are fun, are they not?
  13. Igor

    Igor Guest

    My wife start to say that I spin around a computer more than around her(true:eek:)
    And that she is going to kick my a.s:eek:
  14. Hameister

    Hameister Guest

    We've all heard that from our wives. They just don't understand. Computer hardware is to us, what shoes are to women. Only difference is, hardware makes sense, buying more shoes than they can wear, makes no sense at all.
  15. Computurd

    Computurd Guest

    well by the time i get here, the computer is fixed and everyone is drinking! ROFL, glad its fixed :)
  16. Hameister

    Hameister Guest

    I tried the new driver.

    The M4 scores were almost identical to the old driver.

    The Mushkin scores were slightly lower. I went back to the original driver.
  17. Igor

    Igor Guest

    But what is with that by-pass that i talk about,do you have that.Wait i second till i find picture on web.
    I could not find so i take pictures.
    P1010238 (640x480).jpg
    P1010239 (640x480).jpg
    P1010240 (640x480).jpg
    This Fw mode give me 20% lower 4k,and Auto is in the middle with results.
    Fw Mode is default setting.
    Last edited by a moderator: Apr 22, 2012
  18. Hameister

    Hameister Guest

    I'll have to look.....I don't think my BIOS has that option, not sure.

    Truth is, that for me it really doesn't matter. The drives are plenty fast just the way they are, and I wouldn't notice the difference even if they were 30% faster.
  19. groberts101

    groberts101 Guest

    all efforts to squeeze that last little bit out of that Marvell chip are an exercise in futility. No drivers or firmware are going to circumvent the real underlying issue here. Which is the Marvell's internal PCI-E x1 restriction and the halved write performance that results from the chips inner workings.

    Also consider that benchmarks can easily vary by 10 percent even on the best of systems so numbers can certainly be skewed quite easily.. especially with the benchmarks being used here. Disabling bios power mgmt, speedstep, and leaning W7 down to nearly the shell of a fully functional OS followed by the pre-requisite secure erase between driver tests will be about as good as you can get for closely repeatable results. Even the drives internal garbage collection from the trim commands can skew results with many controllers.

    My best advice is this. Don't screw around with firmware UNLESS you have a functionality issue on your boards Marvell chip because it's simply not going to net you additional speed like you're hoping for here and you can brick the chip completely. Drivers would be your best chance.. but as I already said above.. we've reached the point of severely diminished returns at this juncture(especially on such an old sata chip like that. If you want more speed?.. overclock the PCI-E bus since the Marvell chip runs directly off it.

    Here's my OC'd Marvell running my V3.

    And with 2 x Vertex 2's in R0.

    I'm done searching for more speed form that particular chip.
  20. Zaxx

    Zaxx Guest

    And anything squeezed will not be 'felt' in any way, shape or form...hence a waste of time imho.

Share This Page