Z68MA-D2H-B3 + Corsair 2x4GB Consistent BSOD

[supernes]

Hello,

I'm having a weird issue with what I presume is either the memory or the motherboard of a PC I recently built.

First off, here's the configuration:

GIGABYTE GA-Z68MA-D2H-B3 (BIOS F9)
Corsair VX550W PSU
Intel Core i5 2500 (non-K, Default Clock/Voltage)
Corsair Vengeance 8GB (CMZ8GX3M2A1600C9 ; Default 1.5V 9-9-9-24)
SAPPHIRE VAPOR-X HD5770 1GB GDDR5 (Default Clock)
4 SATA Hard Drives (3 SATA2 + 1 SATA3)
2 120mm Fans
Windows 7 Professional x64

As mentioned, the RAM is a 2x4GB DDR3 set. If I run the system with either one of the memory sticks, it works fine. But whenever I use them both, I get a BSOD in about 5 to 10 minutes. Usually it's a MEMORY_MANAGEMENT stop, sometimes a PFN_LIST_CORRUPT. The system crashes both with a previous install of the OS and during the installation procedure itself. This leads me to believe that it's not a driver issue. Happens on a fresh install as well as after installing the chipset drivers that came with the motherboard.

Things I've tried to no avail:

• Ran Memtest86+ 4.20 for a good few passes with both sticks, and with each stick individually - no errors were detected
• Tried running the system in a single-channel configuration - seemed to run fine for a while, but the BSODs returned eventually
• Tried relaxing the timings, turning off XMP, entering timings manually - no effect whatsoever.
• Ran the Intel Processor Diagnostic Tool with the default config and a memory stress config, everything passes (with a single ram stick in, obviously)

One thing to note is that the system runs seemingly fine when I boot up Mac OS X 10.7, however under certain conditions it would start filling up memory as "unused" and run into the swap, something that I certainly can't classify as "acceptable" for an 8gig system under the loads I put it through.

What am I missing? I've never ran into a problem like this before, and would certainly appreciate any suggestions you can throw at me.

Thanks in advance!

[MiiCK]

Just one things DM is going to suggest is that you run 10 full cycles of MemTest on each stick to be sure, 1 or 2 cycles isn't enough.

Can you try the ram in another machine? Try swapping the ram slots around? Few easy things to try would be clearing CMOS, loading optimized defaults etc.

[Aussie Allan]

  Welcome Supernes ... It's one where you'll be getting a lot of advice ..... some worthy of noting ..... others might lead to "The Fix" ...

  For starters under "worthy" ..... your Vid Card! ....... peak load is around the 235/240W ..... this is near on half of the PSU out put and running things on a knife edge from the get/go!

  Look towards raising the QPI/VTT in BIOS/memory management as a possible "The Fix" ....often raising this value slightly when increasing memory population stops BSOD's .... remember to keep memory voltage and QPI/VTT with in 0.5V of each other ..... QPI/VTT value can be higher but not lower otherwise the on-board controller (CPU)  can be strained or damaged.

  Aussie Allan

 

[supernes]

Thanks for the quick response, guys.

@MiiCK - Trying another pair of RAM sticks is the next thing I'm going to do, I'll ask a colleague with a similar setup to lend me his for a day or so, I'll be sure to report back once I try this. Regarding Memtest, I once let it run overnight (6+ hours) with both sticks in, that's about 8 or so runs I think, and it didn't report any problems. Maybe I'll repeat the procedure again (on each stick individually) just be absolutely sure.

@Aussie Allan - Yeah, I'm worried about the card's drain too, even more so after taking into account that the PSU's a couple of years old at this point. But can using a single RAM bank instead of two really make that much of a difference in my case? I've stressed the card a lot with the single stick setup, and I haven't experienced any problems that would indicate that the power wasn't enough.

I took another look at the voltage settings in the BIOS just after posting, and noticed that the "XMP Profile Recommended" VTT/QPI voltage was 1.2V, but was set to 1.050V on Auto on a separate screen (Advanced Voltage Settings).

In summary, the voltages right now are set to 1.5V for DRAM, 1.188V Vcore (according to CPU-Z, didn't mess with that in the BIOS), and 1.2V manually for VTT/QPI. What voltage that you're suggesting I keep close to VTT? AFAIK, 1.5V is the standard for DDR3, and the sticks I've got are rated for that.

Anyway, I'm starting to eliminate variables. I unplugged all hard drives besides the one with Windows installed, and got halfway through typing this response before it died on me again (after bumping VTT and hanging the remaining disk outside the ridiculous dock it sits in usually). I removed the GPU and am running on the IGP right now, we'll see how far it goes.

Also, for posterity, here are the voltages and temps that HWMonitor reports. Some of them are way off, is that a reporting problem or is it a miracle I'm able to boot at all?

Edit: It seems stable for now, but when I try to run TouchBIOS I get a C1 stop BSOD (SPECIAL_POOL_DETECTED_MEMORY_CORRUPTION).. could be unrelated, but it happens every time.

Thanks again.

[Lsdmeasap]

If you are using slots 1+3 try 2+4 instead, sometimes this helps, reverse that if you are currently using 2+4.

What tRFC is auto setting?  Check with AIDA64 in the motherboard > chipset section > Select NB at the top, not what you see on the left in the BIOS as this is not what's currently used.

1.2 QPI/Vtt may be a bit high, especially if you are not overclocking anything, try 1.1-1.15 instead because higher than needed can cause issues too.

As for HWMonitor, those are just incorrect, the app needs updated to properly detect your board, or you need to update to the latest version (if he's already updated it)  AIDA64 will correctly show your voltages, as will the latest version of Easytune.

[supernes]

@Lsdmeasap - thanks for your suggestions.

I've tried all combinations of slots, including 1+2 and 3+4 (single channel).

I've read that 1.2V VTT is indeed high, maybe it's an outdated recommendation.

AIDA64 shows the proper voltages (12.172V on +12V), Bumping VTT to 1.1V didn't help, still BSODs in 45 seconds flat.

tRFC should be 128 (@800), but leaving it on Auto + XMP shows it as 138 in MIT Current Status. @666 it's 108, but still unstable with both sticks in.

I tried turning off XMP and entering the timings manually for both 800MHz and 666MHz, but the BSODs persist.
Lately I've been getting SYSTEM_SERVICE_EXCEPTION (3b) in various built-in drivers, with Windbg reporting the probable cause as: memory_corruption ( nt!MiGetTopLevelPfn+65 ).

Tried updating the chipset drivers, which included the DRAM controller, but that obviously didn't do anything useful.

The OS install is on a rather old Samsung 160GB drive, but I ran an integrity scan using some Samsung tool and it didn't find anything. Will do another run just to be sure. Is there any chance at all it could fail so consistently if it was a HDD problem?

Running the system on the Integrated Graphics Processor (64+2MB reserved) is still stable with both sticks, as is running with the Radeon GPU with any one of the two modules.

 I just can't figure this out.

[Aussie Allan]

   Sorry supernes ....I should have been clearer re the 0.5V limitations .......Example:..... if you have 1.55V on the memory .... it's not recommended to go below 1.05v QPI/VTT ..... well that's my understanding ..... can you clarify this Lsdmeasap

  Aussie Allan

[Rolo42]

Running the system on the Integrated Graphics Processor (64+2MB reserved) is still stable with both sticks, as is running with the Radeon GPU with any one of the two modules.

It's rock-solid in either of these two configurations?

I wonder if the card's assigned memory range is conflicting with the 4-8GB RAM memory range.

See if you can view it in Device Manager and take note.  Clear DMI data (with both sticks installed) and see if it changes.

[supernes]

@Rolo42 - That's an interesting point, hadn't occurred to me...

I tried something else, moved the GPU to the PCIe x8 slot, and 40+ minutes in, the system hasn't crashed (running at 1333MHz ATM). Also, after reloading the Optimized/Failsafe defaults in the BIOS, the system left the IGP on, and subsequently reserved the 66MB framebuffer for it.

There are a couple of ranges (000A0000-000BFFFF, E0000000-EFFFFFFF) assigned to the HD5770 that coincide with another device (2nd Generation Intel(R) Core(TM) Processor Family PCI Express Controller), but judging by the name, that's to be expected. The third range (FBEC0000-FBEDFFFF) partially overlaps the PCIe Controller (FBE00000-FBEFFFFF), and that's it. Again, that's after a BIOS reset to defaults and the GPU in a different slot, going strong.

I might try to move the GPU back to the x16 slot and clear DMI data as you suggested. Thanks for that idea, sounds very much plausible.

Also, I ran Memtest86+ for 10 passes last night (~10.5 hrs), not a single error was detected, so even if the RAM is faulty, I can't prove it.

[Lsdmeasap]

   Sorry supernes ....I should have been clearer re the 0.5V limitations .......Example:..... if you have 1.55V on the memory .... it's not recommended to go below 1.05v QPI/VTT ..... well that's my understanding ..... can you clarify this Lsdmeasap

  Aussie Allan

Yes, that is correct, which is why default is 1.5V memory and 1.05V QPI/Vtt.  You can use any amount of either, of course 1155 CPU's don't usually like more than 1.2-1.24 QPI/Vtt from what I've seen, but you can use any amount you like as long as you keep that 0.5 difference.   Like if you are using 1.65V ram, then you want at least 1.15V QPI/Vtt.  You can go higher/closer to the memory voltage or above, just not below that 0.5 amount.

Supernes, hope you can figure this out eventually, sounds like you are on the right path!   Keep an eye on I/O Latency at the bottom of the advanced memory settings page, sometimes I see this jump to 1 instead of 2 with auto, and generally 1 is unstable without doing it on purpose with a lot of voltage.

[supernes]

Thanks for your suggestions guys, it seems everything's stable now!

I moved the GPU back to the x16 slot, memory's running at 1600MHz, everything's peachy.. as long as I leave the IGP enabled... go figure.

Anyway, I'm kind of hesitant to do any more snooping, so I'm just going to leave it at that.

Thanks again!

[Lsdmeasap]

Great to hear everything is running smoothly for you now!

Post back anytime if you need more help with anything!