GA-X58A-UD3R - Random failure to wake from sleep after BIOS Update to FG1

[dikal]

when a PC drops down into "Hibernate" power is still being supplied to the Memory .......

You meant to say "Sleep" (S3), didn't you?  

[Aussie Allan]

   "Both"

  Aussie Allan

[dikal]

Although hibernation uses (the least) amount of power, according to microsoft (http://windows.microsoft.com/en-US/windows7/Sleep-and-hibernation-frequently-asked-questions), I have tested that if you put your pc into hibernation, then remove any power from it, there is no problem to resume into windows.

But your theory about the corruption after S3/S4 seems very logical.

[Aussie Allan]

 
      Dukal ....

              "I have tested that if you put your pc into hibernation, then remove any power from it, there is no problem to resume into windows."


     This is true for LAPTOPS and those lucky enough that have a UPS covering there desktop system .... in this scenario mains or battery power is still present in spades ...... what I'm refereeing to specifically is when the tower is connected to the mains only via the installed PSU .... think of it as a system , self induced "Brown out" ...... it might be only milliseconds long ..... but long enough to corrupt the volatile memory ...... it totally relies  on the PSU ability to also wake and supply enough current at the correct voltage almost instantaneously...... it might just mean a bigger capacitor on the motherboard as a voltage buffer for the memory .... interesting though anyway!

  Would be interest to hear what the boffin's at Gigabyte thought .

  Aussie Allan 

[Rolo42]

What?  No. You are thinking suspend.   Hibernation is to be used when you don't want power draw or when power will be completely removed.  Memory contents and CPU state are written to disk (hiberfil.sys) and the computer is shut down.  This is what you want your power settings to do before your battery (laptop or UPS) runs out of power to power down quickest without data loss.

[Aussie Allan]

  Rolo84 ..... Koo ...Koo ...Koo..... ..... perhaps I left the door open a tad! ..... the description was meant to be rather general so just about all understand where I'm coming from...

  most don't really know the difference between Hibernation, suspend, sleep, Hybrid sleep, the list goes on to the extent Windows 8 has a novel new way of doing things so the recovery time is shorter by compressing the cache data still further to be written to disc (or ssd)..... Greeeeeat ! ..... one more flaky mode that probable won't work out of the box!

  Perhaps I just should have left out "Hibernate" .... while on the subject .... one of my favorite reference points is Wiki ..... for the data to be excepted....it's checked, .... double checked ..... scrutinized and checked again by literally 100s of members ..... compare this with a review .... it's only as good as the the single person punching the keys and one proof reader if you're lucky.

  If like me ....you want to know a little more on the above subject .... I'll include two cool links below.

  My favorite extract that sums up how far manufactures have to go , and why most people on the forum recommend against, Sleep hibernate ext, is this statement .....

  Entering into hibernation can cause incorrect operation on restarting, due to problems with the hibernation software, or with devices or software which is not fully compliant. Hibernation will also "USUALLY" cause connections to "PERIPHERAL"  devices to terminate; this may cause problems for peripherals that were in use when hibernation started

  Sums it up nicely,  Doesn't it...

http://en.wikipedia.org/wiki/Sleep_mode

 http://en.wikipedia.org/wiki/Hibernation_%28computing%29

  Aussie Allan

[Nealeb]

I think that this is the most up-to-date discussion that I've found on this topic - and it's been going on for a long time...

I have a UD3R, rev 2, that I use as the basis of a system for recording off-air television, so it needs to reliably hibernate and resume. It has never done this reliably - as with the OP, it works a lot of the time but then fails to resume for no apparent reason. Failure mode varies from freezing with the Windows desktop (Win 7 Ultimate) displayed to BSOD - data in page or secondary processor failed to receive clock tick (or something like that). I have tried a number of fixes.

• USB3 hardware problem. Tried latest USB3 drivers; eventually disabled USB3 controller. No effect.
• SATA3 hardware problem. Disabled Marvell SATA3 controller. This stopped an occasional hang just after "DMI pool loaded" but this looks like a different problem
• Changed TV tuner card from Kworld device with flaky Win7 driver to Black Gold written-for-Win7 driver - no difference
• There is a theory that the X58 chipset is happy to drive DRAM at 1.65V (which my original memory required) but when resuming from hibernate it switches to 1.5V until memory has been rewritten from the hibernate file and only then switches to 1.65V (or whatever is selected in the BIOS). This could cause memory corruption in the restored image, and certainly the random hang symptoms are consistent with this. I tried running memory at 1066 and 1.5V where Corsair claim it should work (only needs 1.65V at higher speeds). That didn't make any difference. I have now installed 1.5V memory although again running at only 1066, just to stay well within spec. That doesn't make any difference. I've also played with QPI/VTT with no effect.
• Tried disabling ErP (some suggestions around that this can cause problems) but still testing

Using latest drivers for everything, and fairly recent BIOS (FH5, from memory). No SSD, which seem to give problems, but I do have a single system disk plus 4 disks in RAID5 on the Intel controller - needed for disk space for off-air recordings. Nvidia GTX460 with latest driver. No overclocking - I would be happy to underclock if it made the thing stable...

Happy to compare notes with anyone else chasing this problem.

[Aussie Allan]

  Keep in touch Nealeb and the link active .... we have been researching this and the occasional new theory pops up and thanks for your's

  I'm starting to suspect it could be something thats already been discovered but a combination of two or three .... really frustrating that a pretty good platform can get this old but a few niggely things persist with no end in sight ... as with anything , eventually someone will put there finger on it.

  Also nice to find someone willing to just shoot the breeze and talk about it rather then turn every second word into a "Battle of the Keyboards."

  Going back to what you said about it hoping from 1.68 to 1.5 IS very interesting and potentially could be one of the two or three I was talking about.....I'll be putting it into the file I keep to test a few Ideas with this new information.....

  Aussie Allan

[Rolo42]

Yeah, I thought he had it with the 1.5 vs. 1.65v there.

Hibernate has been flaky since Win95 and because it's been that way for so long, a fair amount of acceptance may have precluded its resolution.

(I have nothing technical to add, so there's some philosophy.) 

Fortunately, ours works (I inadvertently tested it by accidentally leaving the wife's battery config to "Critical = 100% and hibernate" and power went out for about a second.); let me know if there's any way I can help.

What about disabling hybrid sleep..?

[Dark Mantis]

Yeah, I thought he had it with the 1.5 vs. 1.65v there.

Yes me too! It really seemed to make sense there and I was expecting a "Eureka" moment.

[Nealeb]

Every few weeks I've been doing some googling around the subject - just surprised that I didn't stumble across this thread before! I should point out that none of the things I've tried are my own idea - found them all by endless reading. The ones that I've tried are the ones that at least sound technically plausible - even the BIOS DRAM voltage switching which seemed to be based on someone's measurements. But there is an awful lot of nonsense talked and often by people with "loud keyboards" and little thought process.

I get the feeling as well that there are a few problems that give similar end-results. Some of these are the USB3 and SATA3 problems, which are hardware and/or driver related. Then there's the Windows issues, like PCI-e power saving settings and hybrid sleep. SSDs seem to have their own problems, and if you look very carefully through the Intel release notes for the RAID controller driver, there is mention of something done to improve S3 resume reliability. Graphics cards aren't above suspicion - I have a plan in place to borrow an older lower-spec graphics card just for testing, if my current test doesn't show up anything. Oh, and I forgot the bunch of Windows patches dealing with SCSI miniport issues, large disks, and so on - and the fact that a lot of people have claimed that the problems started with Win7 SP1. Dunno about that one - I started this game post-SP1.

I shall watch this space, and report back anything that I find.

[Aussie Allan]

 
  As will I Nealeb

  For what it's worth .... all Raid or non Raid HD or SSD aside ..... the biggest overall stability improvement I have had with several small issues including a couple you just mentioned ......was going over to a big, high quality PSU in conjunction with a quality UPS

  Aussie Allan

[Dark Mantis]

I never let my systems sleep or do anything other than fall back to a screensaver! My attitude is that I pay enough for my hardware and don't want it kipping on the job! Besides as for what is saved in electrical costs are negligable and just not worth the trouble.  Of course not everyone has the same attitude and I agree that these things should work fine if wanted.

[Nealeb]

Just gone back to check when I started this game. I built the system back in August 2010. I started with a Radeon ATI graphics card and as I remember (this was before I started writing things down) had hibernate/resume problems so I swapped the card for an Nvidia GTX460. I had had success with Nvidia on a previous machine running XP - although I had problems with hibernate/resume with that one until I eventually tamed it.

I also have an Antec CP850 power supply, one of the bigger-than-ATX-standard supplies that only fits a few Antec cases. Seems to have had good reviews for stability, ripple and so on. No UPS, though. Additionally, in the last 18 months or so, I think that I might have had just one unexplained BSOD while the machine was running. It has been pretty stable by Windows standards. It is only at resume-time that I see problems. What's different about resume that doesn't happen at boot? Is the Vdram thing relevant? Dunno - haven't actually hung a voltmeter on the motherboard test point to check.

Once past the "verifying DMI pool" stage where I did have problems with (I think) the SATA3 controller, the hang or BSOD only seems to happen once Windows has been reloaded into memory and started running. Is there a driver somewhere that does not do the right things on resume when it comes to reinitialising the hardware? Is the memory image itself corrupt (and the symptoms are consistent with this)? Does the BIOS do the right things on resume to reset the hardware properly (and there is mention of this in an X58 chipset release note that I found that gives a BIOS instruction sequence necessary to properly do this - Gigabyte would have done this, wouldn't they?)

I'm currently running the system with ErP disabled in the BIOS (power management-related parameter that just might be relevant as it affects power to the motherboard in "S5" state - although that's not a very well-defined power state). So far, it's done a couple of days and maybe a half-dozen resumes without a problem, but sometimes it can go a week without showing any issues. It's also unclear whether there is any sensitivity to how long it has hibernated before resuming, so 7 hibernate/resume cycles in a day is not the same as one cycle per day for a week. I hate intermittent faults with a passion! My personal next-steps are going to be:

• replace graphics card with a different model (despite this being the second card that I have used but it's an easy one to try)
• reconfigure the disks by moving around data so I can use it normally with only one hard drive in use and the RAID controller and its disks disabled.

18 months this has been going on. I've just built a new system for my daughter-in-law with an i7 and Z68 chipset, so the mark 2 version of the CPU and newer chipset. Wonder if she'll notice if I sneak it out for testing here...?

[Aussie Allan]

   Man I'd say go for it ... Koo-koo-koo   What you have done so far, and what you have already been through ..... You'll kill your self if you don't follow this through to the end

  Here's another tit bit I can/will share ...... when I moved it this last house ....power problems .... big power problems! ....... quoted at 230V (UK) ....I would have been happy with even 240V ..... but the 262V peak down to 248V min ..... it was slowly roasting all my equipment

  Compounding this was a flaky region with blackouts and brownouts easily on a daily bases ..... we tested this out last year with an Oscilloscope covering the power going into the PSU during brown outs ..... the resulting bell negative curve during very short brownouts was expected ..... what was not expected was when well monitored the power after the PSU going into the board ..... what we saw was the expected negative bell curve immediately followed by a positive bell curve to the same magnitude ? ? ?

  Above all this , when the UPS was put into the loop .... this all stopped (all most,very small) .... mind you , the first cheap UPS did not have as big as an effect on getting rid of this problem as we both expected ...... we can only assume the more expensive unit had a superior,faster power Filter layout.

  Not a total fix, but a big step forward if you run any and all "S" states and or RAID.

 

  Aussie Allan