X58A UD3R DMI Data pool

[Dark Mantis]

Yes, good luck. It has been known before to have more than one set of memory faulty and replaced for more.

[jdvoracek]

They are rated 7-7-7-20 though but I sure will try so in the morning to see what memtest does from coldboot.
8-8-8-24 1.66v QPI/VTT at 1.16v if I understand you correctly? Isn't that waaaay undervolted?

There's more to memory tuning than just tCL (CAS latency), tRCD, tRP, and tRAS (7-7-7-20) and QPI/Vtt. DIMM SPD/XMP has settings for tRRD, tWTP, tRFC, and tCMD. You can get at these with a program that reads the SPD profiles. My X58A-UD3R BIOS will automatically adjust these timings if BCLK or memory speed is increased over 1066 (unless I do that by enabling XMP).  If I ran my memory at 1600 by enabling the XMP profile, BIOS used the timings from the XMP profile in the SPD.  Otherwise, if I manually ran my memory at 1600, BIOS used other timings it apparantly calculated for tRRD, tWTP, tRFC, and tCMD.  In a nutshell, XMP used tRFC=72 but when I set the memory speed manually, BIOS used tRFC=60.  The other timings were slightly different, but tRFC was the killer.  Set at 60, memtest86+ failed only very intermittently, but I BSOD'ed every day or two, and Windows 7 always refused to resume from sleep.  With tRFC set at 72, life is GOOD all the time.  I varied QPI/Vtt all over the place, and it did not improve stability at all.  These timings may not matter most of the time, but they do sometimes.

[Trinitrotoluene]

I have set my tRFC to 88 right from the start command rate to 2, as recommended by OCZ. The other values are all defined by the bios. So that should also be OK. As far as I can see, all settings are correct...

[jdvoracek]

I have set my tRFC to 88 right from the start command rate to 2, as recommended by OCZ. The other values are all defined by the bios. So that should also be OK. As far as I can see, all settings are correct...

Well you are smarter than I because it took me almost 6 weeks to realize I needed to attend to tRFC.  I'd still get as many timings as I could from the closest SPD profile.  I got into trouble by letting BIOS do those settings, but I think you are likely okay since tRFC seems the most critical for me.  Again, good luck.

[Trinitrotoluene]

This morning I tested from coldboot, no errors (2 passes). However, first time I booted the system, it would shut itself off and turn itself on again right before or at the memory check post. But I have to say, I wasn't expecting anything like that to happen, so I may have missed the actual moment it happened. I do think this is something significant though. 1.66v for the ram, 1.180v for QPI/VTT set the timings to 8-8-8-24 88 2T, uncore still at 21x. I don't really feel confident about the no errors in memtest though. But maybe like you said, less aggressive specs might work better on a faulty module! Perhaps logic has just found its way in....

Anyway, I realize this post may not have any content to reply to, it's just a status update. The real discussion is still taking place at OCZ forum, hopefully me and them can both find a clear cause soon and take appropriate actions.

[Dark Mantis]

Please keep us updated as to your progress anyway.

[Trinitrotoluene]

Yes, if it's not for any of you guys. It sure might prove to be useful for someone being linked by google to this place!

I'm still at 1.66v for the ram, 1.180v for QPI/VTT set the timings to 8-8-8-24 88 2T. Did memtest coldboot again, errors:

http://img177.imageshack.us/img177/6384/20101107104044.jpg
http://img828.imageshack.us/img828/7614/20101107110906.jpg

So yesterday none, today, errors again. But remember, yesterday my PC did this vague auto-reboot when posting. Might have something to do with it. Also, when I just posted on the OCZ forum again, first my firefox crashed. Then when continuing my post, I had a memory_management BSOD. So that sure looks like an in-your-face confirmation of the memtest errors upon coldboot. And your story about bad performing cold chip on the sticks.
Also the errors, MB-wise, seem to be occurring in the same places most of the time. Don't think I posted that much screenshots here, those are on the OCZ forum.

I've tried various QPI/VTT voltages: 1.360, 1.340, 1.320, 1.300, 1.200, 1.180 - all of them give memtest coldboot errors. Those errors seem to go hand in hand with system instability as well. I didn't get system crashes on all voltages, but that was only a matter of time: I've ran each voltage for just one day, before moving on to the next memtest coldboot. It's not that I get a crash every hour or minute.
I even slacked off the timings a bit. Cold boot errors still persist and system instability is still there. Not as bad as with the previous kit, that just had one stick that was bad at all times. This one seems to have at least one stick with a chip that performs bad when cold and is still unstable when warmed up. That is my best guess right now when I trust memtest. But maybe it's still down to something else, though the options become more and more limited.

For anyone that wants more info, here's the link to the thread on the OCZ forum: http://www.ocztechnologyforum.com/forum/showthread.php?78922-Possible-faulty-module&p=572861#post572861

If the problem is indeed bad performing cold chips etc. etc., it has proven to be hard to nail down! It's not easy to test, in this case, whether the problem is a bad memory slot or the stick. Because one time the stick might give errors, next time nothing. But remember to DO trust memtest upon coldboot if you're suspecting anything and when things happen to go hand in hand with firefox crashes and BSOD's with memory_management, amongst others. Now this is all IF IF IF. I don't want to distrust OCZ or anyone more knowledgeable than myself by going ahead of things!


Awaiting advice from OCZ now!

[Trinitrotoluene]

Received first set of RMA instructions from OCZ, waiting for final instructions with RMA number.

Fingers crossed 

[Trinitrotoluene]

So let me copy paste this one right here as well:

So today I received the new kit from Taiwan, the RMA'd kit was indeed broken. Or at least I assume it was, since it was replaced. Thing is, booting up memtest straight after I corrected bios settings (7-7-7-20, 88, 2T - 1.34v QPI/VTT, 1.64v ram) Pass 0 Test 6, 5000,0MB range I had a total of 1018 errors. I think it was only at test 6, I guess I could scroll down in some way but I didn't know how.

If I understand it correctly, 5000,0MB is at "50 percent" of module 3 in slot 5. I unlocked all the brackets on the modules, reseated them again, Did another 1 pass run, no errors. But now I'm kind of clueless, especially considering I already RMA'd 2 previous kits that were in fact faulty. But sure, reseating problems happen to the best of us.

Will use the system as much as possible, though I might not have a lot of time for that now. Might monitor it with remote desktop I suppose! But please give me your view on all this! Reseating, bend CPU pins, faulty again, anything else - whatever you suspect, please share.

[Dark Mantis]

As you said it could have been a badly seated module that caused the faults but it is impossible to prove now. All I can suggest is to give it a good burn in test and put some pressure on the memory and subsystem and see if anything gives. Try running Prime95 with blend tests.

[Trinitrotoluene]

I ran a couple of quick blend tests, not for full testing purpose just yet. A minute ago I started an other blend test and my PC rebooted, no BSOD or error. Just when I got back into windows it reported CCC: monitor program has sstopped working. Probably nothing useful or any kind of cause, just a report.


Problem signature:
  Problem Event Name:   APPCRASH
  Application Name:   MOM.exe
  Application Version:   2.0.0.0
  Application Timestamp:   49ef8e68
  Fault Module Name:   ntdll.dll
  Fault Module Version:   6.1.7600.16385
  Fault Module Timestamp:   4a5be02b
  Exception Code:   c0000005
  Exception Offset:   00000000000159ad
  OS Version:   6.1.7600.2.0.0.256.1
  Locale ID:   1033

[Dark Mantis]

Probably nothing of importance as it is only relating to the graphics driver. But worth keeping a note of.

[Trinitrotoluene]

I am now on my 3rd ram kit as you know. Still have freezes while using the system, not a BSOD so far but that might require more usage. The coil whine is fixed with setting the voltage to a fixed value. However, in Prime I was able to create/duplicate an error in the blend test.


The coil whine thing is RMA'able so to say, but there might also be something else wrong with the board, assuming ram, cpu , psu and gpu are now OK. I take it I need to RMA directly through Gigabyte? I'm located in The Netherlands but I suppose I need to RMA through rma.gigabyte.us?

Anything else I should keep in mind or need to know? I noticed bend/damaged pins are not covered for RMA. Lets hope I don't find anything weird going on with the socket...

[Dark Mantis]

Hi
If i remember rightly you purchased the board in the USA didn't you? If so it will have to be RMAed back there. I would try contacting Gigabyte's Netherlands office and see if they would be happy to do it there but I think you are going to be stuck with sending it back.

[Trinitrotoluene]

I've purchased it in a Dutch webshop, I guess if I RMA it to them, they would still RMA it to the Gigabyte center? Or perhaps not, depending on the problem. So I'm not sure what the best solution is for RMA.