GA-880GMA-UD2H: system freeze w/ graphical corruption

[asgallant]

I'm building a HTPC for a client, and I'm having an unusual issue that I can't seem to locate the cause of. Occasionally the system will freeze up and graphic artifacts show up on the monitor. This has occurred while making changes to BIOS settings, after POST when initializing AHCI drives, and within Windows. It sometimes happens on cold boots, when the system has been powered down for several hours, and sometimes when the system has been running for several hours. It happens at idle, full load, and in between. Temperatures are well within thermal limits (the max temp recorded by Sysfan was 45C). Power consumption is stable at around 55 watts (measured at the wall) at idle and 100 watts at full load. Prime95, SuperPi, and Memtest86 can run for hours without encountering any errors. When the freeze happens in Windows, there are no events logged in the management console to suggest what might have happened.

I popped a GeForce 6600GT I had lying around into the system to see if it would make any difference, and the freezes still happen, albeit without the graphics corruption (though I did have one freeze that replaced the Windows 7 desktop with a blank screen that was the greenish-blue Windows 98 desktop color). Stability through several passes of Memtest86 (run with the discrete card in and integrated graphics disabled to test the memory normally allocated to the integrated GPU), leads me to believe that the problem isn't with the RAM. With power consumption stable, within expected ranges, and well within the PSU's tolerance, I assume that the PSU isn't the culprit, but I'm going to borrow a power supply from a working system to confirm that. I also flashed the BIOS to the latest F5 revision available on Gigabyte's website just in case it was a BIOS issue, but the problem remains.

Eliminating all other culprits, it must be the MB or CPU, but I don't have another AM3/AM2+ MB to test the CPU on, and I don't have another AM3 CPU to test in this board. Given the graphical corruption, I suspect the MB, but things aren't always as they seem. Are there any tests I can do to determine what the likely culprit is? Is there some possibility that I didn't consider?

The system is running at stock speeds, and I haven't attempted to unlock the dormant CPU core. For what it's worth, all drivers are up to date. In BIOS, I selected "Load Optimized Defaults" then changed the SATA controllers to AHCI mode and enabled SMART mode for HD's, all other settings are stock.

System specs:
AMD Athlon II X3 415e (45 watt TDP)
Corsair 2x2GB DDR3 1333
Gigabyte GA-880GMA-UD2H
Western Digital Caviar Green 2TB
LG WH10LS30 Blu-Ray burner
Ceton InfiniTV 4 CableCARD tuner (waitlisted - Argh!)
SeaSonic SS-350ET 80+ Bronze
SilverStone GD05B HTPC case
Windows 7 x64

 

[damasvara]

Try reinstalling the video driver. Do you use cool n quiet? If you do, you might want to turn it off and see if the problem persists.

Hope this helps!

 

[bilbat]

I would post this in the Ceton thread at the Green Button, as well...

Also, check your SMART data to be sure the drive isn't doing deep recovery cycles - I had a crapped out Seagate that produced the same symptom - eventually failed completely, fixing problem!

 

[asgallant]

Damasvara, it's highly unlikely that it's a driver issue, seeing as how the problem occurs outside the Windows environment. I'll give it a shot, though.

Bilbat, sorry, that Ceton card is a misprint (the list was taken from my build parts list). I don't actually have the card in yet, and I'm given to understand the wait list ahead of me is absurdly long. I'll post over there anyway, and check the SMART status of the drive, that's one possibility that hadn't occurred to me.

 

[asgallant]

Update: I can now reliably force the freeze by running SuperPi calculations for 2M+ digits (sometimes 2M does it, otherwise 4M, 8M, or 16M will) or Prime95. SuperPi causes a freeze after a minute or two, Prime95 freezes the system almost instantly. This was not the case a few days ago, so I assume that whatever is wrong is getting worse.

 

[bilbat]

Hmm... I assumed this was in the class of 'stutters' - that the systems was working but 'busy'. The key to determining the difference is, well, a key! When it's in a freeze, press the CAPS LOCK key - if the light on the keyboard operates, it's just 'busy'; if not, it's locked up and not functioning. This works because the keyboard interrupt has the highest, or second highest, (never can remember [:bilbat:6] ) priority in the interrupt hardware - if the thing is working at all, the keyboard interrupt will get 'serviced'. Big difference in diagnosis - let me know which way it is!

 

[asgallant]

The Caps Lock key light does not turn on. Incidentally, the SMART status of the HDD is good.

I'm pulling the power supply from this PC to test in the new build shortly. I'll update when I get results.

 

[asgallant]

I have confirmed that the power supply is not the problem. That leaves either the CPU or MB, is there any way to tell which one? I'd rather not RMA one and have the problem be with the other (or RMA both and have to pay a restocking fee for the one that isn't borked).

 

[jaquith]

Since 'you' believed memory -
Q1 - What DIMM slots are the sticks in installed in?
Q2 - Please post a {Link} for the exact RAM?
Q3 - Ganged or Unganged? {try both}

Otherwise I would pull the MOBO and breadboard; oddball 'stuff' is often a short.

 

[asgallant]

RAM is in slots 1 and 3 (channels 0 and 1 on this MB), in unganged mode. It is Corsair XMS3 4GB (2 x 2GB) DDR3 1333 TW3X4G1333C9A (http://www.newegg.com/Product/Product.aspx?Item=N82E16820145251). I switched over to ganged mode and the system froze instantly upon starting Prime95. Switching to slots 2 and 4 has no effect. Given that the freeze happened in BIOS setup and while initializing AHCI drives, and the system was stable through Memtest86 for over 8 hours last night, it seems unlikely that this is a memory issue.

 

[jaquith]

Yeah, well that 'instantly' told me that you choose incompatible RAM.

Corsair Tested - http://www.corsair.com/configurator/product_results.aspx?id=1529925

GA Certified {QVL} - http://www.gigabyte.us/products/product-page.aspx?pid=3632#memory%20support%20list

RMA/Exchange the RAM:
CORSAIR XMS3 DHX 4GB (2 x 2GB) DDR3 1333 (PC3 10666) Dual Channel Kit TW3X4G1333C9DHX http://www.newegg.com/Product/Product.aspx?Item=N82E16820145198

CORSAIR DOMINATOR 4GB (2 x 2GB) DDR3 1600 (PC3 12800) CMP4GX3M2A1600C9 http://www.newegg.com/Product/Product.aspx?Item=N82E16820145296

{CPU} | 1 | 2 | 3 | 4 | ; 1/3 or 2/4

 

[asgallant]

I've never had a problem with memory that wasn't on a compatibility list, but it's worth looking into. I'll check with the folks at Corsair to make sure it's a compatibility issue before I RMA it though.

 

[jaquith]

If your CPU was a Phenom II then I would 'know' it is compatible, but the CPU listed is Athlon II.

Phenom II Corsair http://www.corsair.com/products/phenomii/default.aspx

I never buy RAM unless: 1. Certified {QVL}, 2. Tested RAM Mfg, 3. Ticket RAM Mfg. Otherwise I would breadboard as I've stated.

The RAM 'should' work in either Ganged or Unganged mode, and if it does not then either the RAM is 'bad'/'incompatible' or the CPU is 'bad.'

 

[asgallant]

I ran the system through the OCCT tests, and it was rock stable throughout them all. I spent all night watching Blu-Ray movies without a hitch. I fired up Prime95 and the system froze. I rebooted, started SuperPi, and it froze. Does OCCT not test any memory? If so, that would seem to support the theory that it is a RAM compatibility issue.

 

[jaquith]

ref - http://www.ocbase.com/perestroika_en/index.php?Faq

OCCT is not a test that 'I' use, but I can read - the RAM 'test' appears to be an option; I would choose 'Mix.'

You've see what I posted, and the other possibility is the CPU itself - but I would "3. Ticket RAM Mfg" from above - for the reasons stated above.

 

[asgallant]

The current release of OCCT doesn't have a RAM test in the list, but the Linpack test does check RAM, and that was stable as of two days ago. I'll run an extensive test again today.

 

[jaquith]

Running obscure or older Test Apps results, particularly negative, might be more of an issue of the App itself than pass/fail of those tests.

Worst, you can have a 'pass' memtest and still have problems with your RAM. Therefore, that's why it is SO important to use ONLY compatible RAM. I would be 'crazy' not to verify ALL components prior slapping them together. Also, if you changed the AHCI and enabled SMART after the OS was installed; I would carefully look for corruption. I've come to the conclusion that AHCI causes more problems than it does good. -- if you did this then a simple fix is simply moving the Port {unplug Port_0->Port_1}.

 

[asgallant]

The results of the tests themselves aren't what's important, because the freeze does happen at other times (like when changing settings in BIOS). Those two programs have simply become reliable ways to recreate the freeze.

AHCI was enabled on the HD from the start, it was only the Blu-Ray player that gave me problems with it.

I'm still waiting to hear from Corsair about the RAM (I suspect their American staff are on Thanksgiving break this weekend, so I probably won't hear anything from them until Monday at earliest). I would be quite happy to find that it really is the cause of all the trouble, because it is a simple fix.

 

[jaquith]

Q - If you disconnect the Blu-Ray does the problem go away? I see the AHCI cause those problems as well with several DVD, and it's another reason not to use it. I only use AHCI with 'real' hot-swap and eSATA drives.

Yep, Thanksgiving is here in the US - I gave my staff off starting Wednesday afternoon. You'll probably hear back Monday - Tuesday.

 

[asgallant]

Happy belated Thanksgiving! It took me most of yesterday to climb out of my turkey coma >;o)

I'll give it a try with the Blu-Ray drive disconnected. I wanted AHCI enabled because installing Windows from a SATA drive in IDE mode takes an eternity (and was, unfortunately, the only time I couldn't run the drive in AHCI mode). Someday I'm going to have to figure out how to do an install using the OEM kit.

[edit]
Disconnecting the Blu-Ray drive doesn't change anything.

 

[phxbl8]

I think I have the same problem installing GA 880GA-UD3H with:
AMD Athlon II X4 630 2.8gb
4gb * 2 Mushkin Enhanced PC3-10666
2 Samsung Spinpoint F3 HD502HJ 500Gb sata drives
Cooler Master 600W Silent Pro power supply
and an ASUS 24x Dvd burner

I RMA'd my first MoBo and am seeing the same symptoms with this one. Initially I was in the BIOS screens for about 45 minutes when the screen locked up and the image was somewhat jumbled (graphically corrupt). The first MoBo ran for 2 hours and I got Windows installed before that happened. With the first board I did finally get a beep code of ONE long beep (5 seconds long). No beep code yet with this one it still posts.
Im using a boot cd with memtest86+ and it sees no problems with my RAM but I admit my exact RAM doesn't show up on the MoBo list (I thought I saw it in a review). Today is day 30 (newegg) so I will try to get an RMA approved for the RAM. Wish me luck. With my MoBos and RAM I mostly get a black screen but I have seen the corrupt graphics described in this discussion. I can't even install Windows with this one and Ive seen it lock up in the BIOS with graphics corruption. If you agree that just because RAM passes memtest86+ doesn't mean its compatible that would help because I feel a bit lost on this one.

 

[asgallant]

From what I've been able to glean so far from the folks at Gigabyte is that if the RAM is Phenom II compatible (and Athlon II is essentially Phenom II minus the fat stacks of cache), then it should be compatible with this combo. Corsair still hasn't gotten back to me yet (which I find irritating, since they had an entire business week - even slow as heck Gigabyte tech support got back to me already), so I have no idea what their position is.

What is the exact model of your RAM? Did you try locating your it on Mushkin's compatibility list (here: http://www.mushkin.com/)? It's currently down for maintenance, check back in a few hours to see if it's there (they may be updating with new compatibility info now, so try anyway even if you've already looked).

[edit]

There is apparently an issue with the Phenom II/Athlon II memory controllers: http://forum.giga-byte.co.uk/index.php?PHPSESSID=3e56d07bec5f15d442d5fb0921a76438&/topic,2515.0.html. This seems to apply only to RAM running faster than 1333MHz, which neither your nor my RAM kits do at stock, but the evidence seems to point that way. Did you overclock your CPU or RAM?

In light of this news, I've downclocked the RAM in this system to 1066 to see if that fixes the problem.

 

[asgallant]

Since underclocking the RAM to 1066, the system has been rock stable through several runs of SuperPi and 17 hours or so on Prime95. Stability in Memtest86 prior to underclocking seems to mesh with reports from users who have blown out their memory controllers by running RAM faster than 1333MHz, so I'm leaning towards the CPU as the culprit on this one, despite running the RAM at stock. What do you guys think?

 

[asgallant]

After speaking with AMD, they recommend RMA'ing all three (CPU, MB, RAM). Seems a little extreme to me, but better safe than sorry, I suppose.