Official GIGABYTE Forum
Questions about GIGABYTE products => Motherboards with Intel processors => Topic started by: styr on January 22, 2011, 12:07:00 am
-
I was trying to load windows xp64bit sp1 and sp2 without luck.
I'm getting BLUE screen (preventing screen) all the time after first reboot.
The way I'm loading windows are:
1.In My bios all SATA controllers are set to AHCI mode.
I connect my 2 HD to Gsata Raid Marvell controller and I did set to Raid 0.
My raid is recognize in bios and I'm able to load windows 7 (NO problem) through marvell controller.
The driver for Windows XP(1036 0.28mb release 2009-04-29)
I'm using Power supply 850 W 2x 300 gig black 10000rpm western digital drives and PCI card
Now I did more test and Of course Under AHCI mode I load proper driver 1.0.0.1036 for marvell raid controller and system recognize raid drive.
After first reboot when is lo-gin to windows xp installations screen I experience blue screen.
Also I did try to load windows in IDE mode on same Marvell controller on single drive and I have same blue screen.__Blue death screen !!!
The other 2 controllers (gigabyte and Intel) controllers are loading windows just fine.
Marvell controller is able to handle windows 7 installation no problem.
Any windows xp (sp1,sp3 )86 or 64 are the problem to install.
I'm coming to the conclusion this controller is defective.
I remove all components left only 2 raid drives and old PCI video card to make sure my video is not causing this trouble.
Please help I do not have any more ideas what to do.
Adam
-
*
Why you use not the INTEL ICHxx , that is the best port ?
Pierre
-
I initially tried to install Windows 7 on a RAID 0 using the Marvell SATA6Gb/s controller and I too experienced similar issues (different motherboard - mine is a P55A-UD4P).
In the end I gave up and used the Intel SATA ports. Afterwards I did some testing and discovered the disk I/O throughput is FAR better using the Intel controller. With the Intel the RAID 0 throughput is between 250-270 MB/s but with the Marvell controller the throughput using the same RAID 0 drives drops to about 130-140 MB/s.
In hindsight, especially after researching this on the Internet, I concluded that the Marvell controller is junk and the 6 Gb/s capabilities on these motherboards is more advertising hype than reality. Even with a 6Gb/s SSD most people report better performance with the Intel 3Gb/s controller and using the Marvell in a RAID 0 with dual SSD's produces a very poor 130-140 MB/s throughput. My guess is the Marvell's RAID 0 is limited to 130-140 MB/s so there is not much point in the 6Gb/s connection at all.
-
I did a thorough test of it when it first came out and came to the same conclusions in a nutshell. This Marvell controller just couldn't handle the throughput and in fact constantly dropped disks from the SATA6 RAID0 array. Just didn't work. However it isn't just on Gigabyte's motherboards that this is happening it is on all makes of motherboards that treid to utilise this chip. To be fair there was no option at the time if you wanted SATA3.
The fastest chip on these boards is the Intel controller by far.
-
I was under impression after my last conversations with Gigabyte support that they fix the speed issue on marvell controller with the new FE Bios.
The main reason why I did buy this bord was support Sata 3 drives.It is dificalt to pay the price and not use it.
How about trying to load beta Bios in my case?
Any help appreciate
Adam
-
I was under impression after my last conversations with Gigabyte support that they fix the speed issue on marvell controller with the new FE Bios.
Can you copy and paste the conversation here for us to see please?
Well if they have fixed the SATA3 issue with the Marvell chip I would have thought they would be making plenty of noise about it and I haven't even heard a whisper. As far as I am aware there is no BIOS update that cures this problem and personally I don't think it is even possible. :-\
-
Unfortunately that was conversation only and I don't record any of my conversations.
Anyways if some one try help me to solve my problem I will appreciate very much.
I would like to load windows xp on marvel controller.
Thanks
-
Can you copy and paste the conversation here for us to see please?
Well if they have fixed the SATA3 issue with the Marvell chip I would have thought they would be making plenty of noise about it and I haven't even heard a whisper. As far as I am aware there is no BIOS update that cures this problem and personally I don't think it is even possible. :-\
Although not Gigabyte themselves, in my scouring of web pages I found one review site where Gigabyte had sent them a beta BIOS (with new Marvell controller firmware), a new Marvell driver and a new Marvell RAID Utility and according to the review the speed improvement compared to using the Microsoft AHCI driver or the old Marvell driver was huge. They claimed the performance was what they were expecting from the 3rd generation SATA (although their testing was using dual high speed SATA 6 Gbps hard drives with around 150 MB/s performance each in a RAID 0 and NOT using an SSD with extremely high throughput; I believe the burst performance (using the hard drive caches) was very high though). What I found odd though is the new driver and RAID utility never was released on Gigabyte's web site as far as I can tell. Even today Gigabyte does not install the Marvell driver at all by default with Windows 7. Somewhere else I read that the reason Gigabyte recommends the MS AHCI driver is due to issues with the actual Marvell driver and SSD drives. Although the MS AHCI driver is apparently slower than the latest Marvell driver, apparently it does not corrupt SSD drives when used on the boot drive (although I can't see any reason to connect an SSD to the Marvell controller if the speed is totally crippled). Gigabyte may have incorporated the new firmware into their latest BIOS, but I don't know.
There is an old Marvell driver you can install if you want that is found on the current Gigabyte installation CD, but the only way to install it is to manually install it. If you click "install Marvell driver" using the graphical setup interface (that runs when the CD is inserted), instead it installs a virtual console utility that allows the Marvell Raid Utility to talk to the controller. It never installs the driver itself.
Unfortunately I discovered that if you do install the Marvell driver manually that the Marvell Raid Utility now fails to be able to communicate with the controller and you can no longer setup a RAID on it. I did download a newer driver from a french website, but I haven't tried it yet (installing it now).
One additional fact I found is unless the BIOS is configured to SATA3 Turbo mode, the Marvell controller is crippled due to using a single PCIe 1.0 x1 path to the controller. This bottleneck limits the maximum speed to 250 MB/s for the two SATA ports combined (this is 250 MB/s in each direction for a total of 500 MB/s if you add both directions together). Since each SATA 3.0 port is theoretically capable of 6 Gbps or 750 MB/s, limiting the pair of ports to a total of 250 MB/s combined is a major slow down and this apparently can cause drives to drop out of a RAID. By the time you add overhead and handshaking, the real limit ends up being somewhere around 150 MB/s.
When Turbo SATA3 is enabled in the BIOS, the Gigabyte motherboard reconfigures the PCIe 2.0 x16 path to the video card and uses 1/2 of it for the Marvell controller. Now the video card connects using PCIe 2.0 x8 and the other PCIe 2.0 x8 is used for the Marvell controller. None of today's single GPU video cards can use more than PCIe 2.0 x8 bandwidth so we should not notice any reduced video performance (unless running dual video cards or dual GPU cards). This solution provides a 4000 MB/s path to the Marvell controller, more than double what is required (PCIe 2.0 provides 500 MB/s in each direction per x1 path).
One unknown to me is the USB 3.0 controller. It looks like there is enough bandwidth available with the PCIe 2.0 x8 for both USB 3.0 and 2 SATA 6 G ports. I don't know if both work at the same time or if it is "Turbo SATA3" or "Turbo USB 3.0". The BIOS suggests it is one or the other...
-
I just finally found the solution, but I'm still confuse why the marvell driver is not loading properly in the begining.
I found this article: http://support.microsoft.com/kb/314859/
and I'm using windows xp32 to log in to the pre-install portion of windows and copy from my floppy 3 files belongs to marvel driver:
mv91xx.sys
mv91xxm.dll
mvxxmm.sys
to windows\system32\drivers dir even if I'm loading windows 64 bit.
and the system loading compliantly to the end.
Why I'm experiencing that issue, what can be wrong to not load the driver propertlley in the beginning?
any comment appreciate.
Thanks
Adam
-
Thanks for the extra info Slider, very interesting and maybe usefull if accurate. It must have taken a bit of digging to find some of that though. ;)
One fact that you omitted is that the Marvell contoller chip doesn't pass on the TRIM commands to the SSD anyway so that is another good reason not to use it for them. There have been several Marvell firmware updates and new drivers since inception but none seem to have had any real impact on the perfomance.
I will look into it some more and see what I can dig up within Gigabyte.
-
When Turbo SATA3 is enabled in the BIOS, the Gigabyte motherboard reconfigures the PCIe 2.0 x16 path to the video card and uses 1/2 of it for the Marvell controller.
I was looking for the TURBO option in SATA3 in the Bios but no luck.
I have FE bios version from Gigabyte site and no Turbo option under Sata3 controller
Please point where to look for
Thanks
Adam
-
Sometimes things get left out of the BIOS as they update them and that could well be the case here. I haven't anyway to check myself as I don't have the board but maybe someone else can see for definite. If it is missing it needs reporting to GGTS so they can issue an update to cover it.
-
When Turbo SATA3 is enabled in the BIOS, the Gigabyte motherboard reconfigures the PCIe 2.0 x16 path to the video card and uses 1/2 of it for the Marvell controller.
I was looking for the TURBO option in SATA3 in the Bios but no luck.
I have FE bios version from Gigabyte site and no Turbo option under Sata3 controller
Please point where to look for
Thanks
Adam
It should be in the "Integrated Periferals" section of the BIOS. Take a look at your owners manual. It should discuss it and confirm where it is found. By default my P55A-UD4P motherboard uses an "Auto" setting where it is supposed to automatically enable "Turbo SATA" whenever a 6Gb/s hard drive is detected. As far as I'm concerned "Turbo SATA" should be used whenever more than one modern SATA 3G hard drive is connected, but this needs to be done manually.
-
Thanks for the extra info Slider, very interesting and maybe usefull if accurate. It must have taken a bit of digging to find some of that though. ;)
I've been trying to figure out the "black screen" lockup when resuming from sleep with this motherboard for about a week and I've been reading just about everything I can find on the motherboard. I can't guarantee the accuracy of my information, but it is my "best understanding" currently and is based on a lot of reading (and discounting LOTS of inaccurate information being posted everywhere). Most of the information on the PCIe 1.0 x1 bus vs PCIe 2.0 x8 bus switch was from reviews on Anandtech and Toms Hardware. I can't recall which site was sent the updated firmware, driver and MRU from Gigabyte, but I believe it was NOT Anandtech nor Toms Hardware.
One fact that you omitted is that the Marvell contoller chip doesn't pass on the TRIM commands to the SSD anyway so that is another good reason not to use it for them. There have been several Marvell firmware updates and new drivers since inception but none seem to have had any real impact on the perfomance.
I actually left that out deliberately because there seems to be a lot of confusion as to if TRIM is or is not supported. I have read several times that TRIM is supported with the last two revisions of the Marvell driver and newer firmware as well as when the MS ACHI driver is used; although I have also read that even though it is supported, some argue that it still doesn't work properly. Some of the comments claiming that it does work include images showing the results of test software that states it DOES work so this makes me more inclined to believe that it at least partially does support TRIM. On the other hand, other discussions have many comments where people say it doesn't work at all so I'm not sure if it really does or doesn't.
I wish Gigabyte would take an interest in these issues and include in their motherboard FAQ if the SATA 6G ports (Marvell 9128 controller) properly support SSDs (TRIM support, etc.) as well as include information on what driver is best to use for which scenario (perhaps MS ACHI for SSDs and Marvell driver for 6G hard drives for example?) Note I'm not saying this is the best configuration at all; one would "hope" that just using the latest Marvell driver would be the best - of course Gigabyte would then actually have to include the driver on their download pages...
As I mentioned before I was not able to get Windows 7 to install when I had the two Seagate SATA 3G drives connected to the Marvell controller. I tried using both Microsoft's own built in driver as well as the Marvel 9218 driver version 1.0.0.1036. I used the BIOS to configure the drives in a RAID 0 and attempted to install Windows 7 64-bit on the RAID. The RAID 0 was detected during the install, but the drive kept "disappearing" or I would experience a BSOD during the install or during the very first boot. Unfortunately I did not try enabling "Turbo SATA" in the BIOS (at the time I didn't understand it was required in order to support the throughput for the RAID 0 and this may have been the reason for the failure). I also did not try the newer Marvell controller driver (latest I'm aware of is 1.0.0.1051 and can be downloaded from http://www.station-drivers.com/page/marvell.htm). It might be worth trying to enable Turbo SATA as well as use the latest driver. This might allow Windows 7 to install on a RAID 0 on the Marvell 9128 controller.
To elaborate a little on the Turbo SATA BIOS setting and why not using it might "cause" the data corruption and BSODsg... When it is not used, the Marvell controller is connected through a PCIe 1.0 x1 bus with a total maximum directional throughput of 250 MB/s. By the time overhead and other inefficiencies are included this drops down to around 150 MB/s (based on me trying a pair of high speed drives now that the OS is installed and only seeing 150 MB/s instead of 250+ MB/s like the Intel SATA controller does; there were also lots of I/O errors reported in the event log and the system kept "pausing", like it does when attempting to handle disk I/O errors). Since the pair of drives in RAID 0 had a throughput that is higher than the bus, the Marvell controller would need to handle some form of handshake flow control with the PCIe bus. One possibility for the data corruption, BSODs and crashes is that flow control through the bottle-necked PCIe bus is not handle properly and data is lost. Perhaps there is no flow control at all and the extra data is simply lost, without an adequate buffer. Once the Turbo SATA setting is configured in the BIOS the Marvell controller now connects using a 4000 MB/s PCIe 2.0 x8 bus. The Marvell controller should now be able to pump data through the bus without worrying about flow control and this might resolve the data corruption issue.
I might give this a try, but if any of you have a chance to test it before me please post your results here. It would be nice to discover that using the combination of the Turbo SATA option in the BIOS and the latest Marvell controller driver provides reliable SATA 6G performance (since we did pay for this when we bought our motherboards). I am also very interested in knowing the burst I/O performance to a pair of 6 Gb/s SATA drives in RAID 0. If the Marvell controller actually works properly we should see numbers in the 1000-1500 MB/s range in theory.
-
I searched a little and found the article on the updated firmware. It can be reviewed here:
http://www.tweaktown.com/articles/3122/gigabyte_sata_6g_update_raid_0_just_how_we_like_it/index.html (http://www.tweaktown.com/articles/3122/gigabyte_sata_6g_update_raid_0_just_how_we_like_it/index.html)
Here is another article discussing improved firmware:
http://thessdreview.com/latest-buzz/new-marvell-sata-6g-firmware-spotted-for-better-ssd-performance/ (http://thessdreview.com/latest-buzz/new-marvell-sata-6g-firmware-spotted-for-better-ssd-performance/)
From what I can tell the latest updated firmware and drivers were initially "leaked" to the Internet last October. The very latest driver was uploaded on December 22, 2010. The updated firmware and drivers were provided to Tweak Town for review last February (although they may not be the same ones that were made their way to the Station-Drivers website initially).
Again, the latest firmware and drivers can be found if you scroll down a little on this web page:
http://www.station-drivers.com/page/marvell.htm (http://www.station-drivers.com/page/marvell.htm)
It is very odd that Gigabyt doesn't have even a single driver available on their website for the Marvell 9128 controller and that their "driver" is actually just an interface to the MS ACHI driver for the Marvell Raid Utility (MRU).
Have any of you tried the updated firmware, driver and MRU?
-
It would be nice to discover that using the combination of the Turbo SATA option in the BIOS and the latest Marvell controller driver provides reliable SATA 6G performance (since we did pay for this when we bought our motherboards). I am also very interested in knowing the burst I/O performance to a pair of 6 Gb/s SATA drives in RAID 0. If the Marvell controller actually works properly we should see numbers in the 1000-1500 MB/s range in theory.
Nice idea! ;D It would be good to find out that this part of the motherboard wasn't just wasted after all, but i won't hold my breath just yet.
Have any of you tried the updated firmware, driver and MRU?
If I remember correctly Pierre checked this out a little while ago but it was before the newest firmware/drivers that you refer to.
-
Although I haven't tried it (other than installing the latest 9128 driver just to see if it fixes my "black screen" crashes issue - it didn't; my next step is to outright disable it, along with everything else I don't use at the moment), I have some potentially bad news:
http://forums.tweaktown.com/gigabyte/42075-my-p55a-ud4p-hdisk-transfer-rate-become-slow-after-firmware-update.html (http://forums.tweaktown.com/gigabyte/42075-my-p55a-ud4p-hdisk-transfer-rate-become-slow-after-firmware-update.html)
Apparently he installed the firmware and updated drivers and his 9128 performance decreased to even worse levels. He isn't running SSDs, but regular hard drives (the firmware update and latest drivers are supposed to improve SSD performance I believe).
So - even if the new code improves SSD performance (unproven at this point), it definitely does not look like it turns the 9128 into a top performing SATA 6G controller...
Here is another web page I stumbled upon where a guy who "sounds like" he might be fairly familiar with Gigabyte's implementation of the 9128 controller discusses it and the issues with it a bit:
http://www.overclock.net/11468940-post5.html (http://www.overclock.net/11468940-post5.html)
Ultimately he says the one and only situation it "might" be worth considering using the 9128 controller (other than where you are totally out of SATA ports and are forced to use it) is where you have a single C300 SSD, you only have one of them (no RAID 0) and you are willing to give up small file read/write performance for a small improvement in large file throughput. I've always thought if I was to go the SSD route I would pick up a pair of smaller SSDs and set them up in a high speed RAID 0 so the 9128 would be once again totally useless to me (unless Gigabyte manages to magically fix the horrific performance of the 9128 in RAID 0).
-
I just finally found the solution, but I'm still confuse why the marvell driver is not loading properly in the begining.
I found this article: http://support.microsoft.com/kb/314859/
and I'm using windows xp32 to log in to the pre-install portion of windows and copy from my floppy 3 files belongs to marvel driver:
mv91xx.sys
mv91xxm.dll
mvxxmm.sys
to windows\system32\drivers dir even if I'm loading windows 64 bit.
and the system loading compliantly to the end.
Why I'm experiencing that issue, what can be wrong to not load the driver propertlley in the beginning?
any comment appreciate.
Thanks
Adam
I did successfully load xp64 after almost give up.
I've notice now windows xp won't load(constant reboot) if I change from IDE mode to AHCI and I will need AHCI mode for second partition for windows 7.
What is the difference if the Bios is set to IDE mode not to AHCI in xp and 7?
I'm still looking for Turbo SATA3 enabled option in the BIOS.I did check manual and all my X58-UD3R Bios FE rev and is not there.
If someone confirm please.
Thanks
-
What is the difference if the Bios is set to IDE mode not to AHCI in xp and 7?
The main difference is that it doesn't support Hot Swapping which is fairly essential for ESATA drives. The IDE mode is slightly slower also but not by a massive margin.
-
Can you copy and paste the conversation here for us to see please?
Well if they have fixed the SATA3 issue with the Marvell chip I would have thought they would be making plenty of noise about it and I haven't even heard a whisper. As far as I am aware there is no BIOS update that cures this problem and personally I don't think it is even possible. :-\
Although not Gigabyte themselves, in my scouring of web pages I found one review site where Gigabyte had sent them a beta BIOS (with new Marvell controller firmware), a new Marvell driver and a new Marvell RAID Utility and according to the review the speed improvement compared to using the Microsoft AHCI driver or the old Marvell driver was huge. They claimed the performance was what they were expecting from the 3rd generation SATA (although their testing was using dual high speed SATA 6 Gbps hard drives with around 150 MB/s performance each in a RAID 0 and NOT using an SSD with extremely high throughput; I believe the burst performance (using the hard drive caches) was very high though). What I found odd though is the new driver and RAID utility never was released on Gigabyte's web site as far as I can tell. Even today Gigabyte does not install the Marvell driver at all by default with Windows 7. Somewhere else I read that the reason Gigabyte recommends the MS AHCI driver is due to issues with the actual Marvell driver and SSD drives. Although the MS AHCI driver is apparently slower than the latest Marvell driver, apparently it does not corrupt SSD drives when used on the boot drive (although I can't see any reason to connect an SSD to the Marvell controller if the speed is totally crippled). Gigabyte may have incorporated the new firmware into their latest BIOS, but I don't know.
There is an old Marvell driver you can install if you want that is found on the current Gigabyte installation CD, but the only way to install it is to manually install it. If you click "install Marvell driver" using the graphical setup interface (that runs when the CD is inserted), instead it installs a virtual console utility that allows the Marvell Raid Utility to talk to the controller. It never installs the driver itself.
Unfortunately I discovered that if you do install the Marvell driver manually that the Marvell Raid Utility now fails to be able to communicate with the controller and you can no longer setup a RAID on it. I did download a newer driver from a french website, but I haven't tried it yet (installing it now).
One additional fact I found is unless the BIOS is configured to SATA3 Turbo mode, the Marvell controller is crippled due to using a single PCIe 1.0 x1 path to the controller. This bottleneck limits the maximum speed to 250 MB/s for the two SATA ports combined (this is 250 MB/s in each direction for a total of 500 MB/s if you add both directions together). Since each SATA 3.0 port is theoretically capable of 6 Gbps or 750 MB/s, limiting the pair of ports to a total of 250 MB/s combined is a major slow down and this apparently can cause drives to drop out of a RAID. By the time you add overhead and handshaking, the real limit ends up being somewhere around 150 MB/s.
When Turbo SATA3 is enabled in the BIOS, the Gigabyte motherboard reconfigures the PCIe 2.0 x16 path to the video card and uses 1/2 of it for the Marvell controller. Now the video card connects using PCIe 2.0 x8 and the other PCIe 2.0 x8 is used for the Marvell controller. None of today's single GPU video cards can use more than PCIe 2.0 x8 bandwidth so we should not notice any reduced video performance (unless running dual video cards or dual GPU cards). This solution provides a 4000 MB/s path to the Marvell controller, more than double what is required (PCIe 2.0 provides 500 MB/s in each direction per x1 path).
One unknown to me is the USB 3.0 controller. It looks like there is enough bandwidth available with the PCIe 2.0 x8 for both USB 3.0 and 2 SATA 6 G ports. I don't know if both work at the same time or if it is "Turbo SATA3" or "Turbo USB 3.0". The BIOS suggests it is one or the other...
Hi,
My X58A-UD3R working now without any issues (so far so good )I successfully load windows 7 and Xp64.
I tested the Raid0 Raptors 2x300gig in windows 7 and I'm getting min 160mb/s and max 225mb/s.Average is around 190.
I would like to try this option to enable Turbo in Marvell controller but can't find in My new Bios FE.
If someone conform what bios revision has that option.
Thanks
Adam
-
It doesn't seem to be part of the standard BIOS theat the board shipped with but it is possible that it has been incorporated into a newer version.