Weird problem that I *think* is bad RAM, but hard to tell so far, help?

D

dewbak75

Limp Gawd
Joined
Nov 12, 2012
Messages
252
Lately I've been having some weird problems with one of my machines at home. The core specs are:
Gigabyte Z68MX-UD2H-B3
intel Core i3-2100(not overclocked, using the on-die VGA)
16GB(4x4)GB Corsair Vengeance DDR3-1600, not sure of exact model, pretty sure it's CMZ16GX3M4A1600C9B

I started getting some intermittent errors in a problem which the author says *could* be symptomatic of bad memory, but I also knew the hard drive in that machine had seen a LOT of use. The machine would also lock up from time to time. Then more recently(the last few days), the system began responding VERY slowly. Like, I would right-click on My Computer and it would take nearly 10 seconds for the menu to pop up, and other forms of sluggishness. Eventually I decided to replace the system drive and just reinstall Windows 7, start fresh. On that machine it's pretty easy to do, there's only a few programs running on it. But I seem to be having the same issues. The Windows re-install went SLOOOOOW, and the inevitable download of 300MB worth of updates as soon as the install finished took FOREVER, the installation of those updates took even longer. I STILL haven't been able to get Service Pack 1 to finish installing successfully. And this was after installing the various drivers for the motherboard devices. I began to think it might be the memory after all. I'd tried running it using the XMP at 1600MHz, and at 1333MHz, same result. So I burned Memtest86 to a disc and ran it with all four sticks in. ERRORS GALORE, like hundreds of them(sorry, I'm writing this from work and don't have the output in front of me to refer to). So, time for process of elimination. I tried each stick by itself in slot 1, ran a full pass until Memtest said it was clean and I could stop(I know many say to run it for a lot longer, but I just didn't have the time). All 4 sticks ran with no errors. So I tried the last stick in each of the other memory slots, and slots 2 & 3 also came back clean. The test on slot 4 was still running when I had to leave for work. Assuming it comes back clean, my next test will be a pair of sticks, set up for dual-channel mode, giving the machine 8GB to work from, and let that test run for several hours. If it turns out that machine just isn't stable with all four memory slots filled, it can get by alright with 8GB instead of 16GB. I may pick up a 2x8GB pack at some point for it.

Few questions:

*could it be that all four sticks are good, but the board just isn't stable with all four slots filled?
*would nudging the memory voltage up a bit help with stability? It's currently running the default 1.5v, should I try going higher, like 1.65v?
*Is there another likely cause for this type of behavior that I'm just missing?

Thanks for ANY help you guys can provide
 
Which memtest? Try MemTest86/86+ and Gold Memory, each overnight. I've had some memory pass for 4-5 hours, and one person had to switch to Gold Memory to find a bad bit, which it did in under an hour but took another 10 hours to find it again. I assume that Corsair 1600 MHz uses 1333 MHz chips, but it's possible it has slower chips or even no-name chips of unknown speed ratings. XMP ratings aren't nearly as strict as chip factory ratings.

Test the hard disk with the factory diagnostic, HDAT2, or MHDD. MHDD's surface scan indicates slow sectors, not just outright bad ones, but it may not work with HDs bigger than 1 TB.

Measure the power supply voltages with a digital meter, not through the motherboard's hardware.
 
I'm using Memtest86, version 4.1, running the default test(the one it starts if you don't choose an option in the menu and just let it begin after booting from the CD)

Is it odd though that having all four sticks in would cause Memtest86 to show errors on the very first pass, but not when each stick was tested individually?


edit:
Also, before I began the Windows 7 reinstall, I swapped out the hard drive for one that had seen much less wear. The system behavior was pretty much the same as it had been with the previous hard drive. The memory errors I got when all four sticks were running definitely makes me think it's a memory issue, I'm just not sure if it's a single stick and for some reason the errors that show up right away when all sticks are in use will need more exhaustive testing to reveal when just testing that stick, or if the board just can't handle running with all four slots filled.
 
Last edited:
You said your memory was running at 1.5v. Did you manually set this voltage in the bios? Or are you assuming that auto=1.5v? If the former, I would try bumping it up a notch because maybe the board is having trouble in that area. If the latter, try to manually set the voltage and main timings to see how that goes.

It definitely sounds like a memory problem to me too.
 
coffaholic said:
You said your memory was running at 1.5v. Did you manually set this voltage in the bios? Or are you assuming that auto=1.5v? If the former, I would try bumping it up a notch because maybe the board is having trouble in that area. If the latter, try to manually set the voltage and main timings to see how that goes.

It definitely sounds like a memory problem to me too.
Click to expand...

The memory voltage is set at [auto] in the BIOS, I'll try manually tweaking the voltage, maybe bumping it up as high as 1.65v.
 
Ok, looks like I probably have a bad DIMM after all, and I think I've narrowed it down to two. Here's what I've done since the morning. I ran through one pass of Memtest86 v4.1 with each DIMM in slot 1, and with the last DIMM in all four slots, which is what led me to believe that the problem wasn't a specific DIMM or a specific memory slot. With this board, if you want dual-channel operation, you put a pair of DIMMs in either slots 1 and 3(the white ones) or slots 2 and 4(the blue ones). I rant Memtest with a pair in the blue slots, no errors. I ran the other pair of DIMMs in the white slots, ERRORS GALORE. So I took that pair out, and tried the pair that had tested fine in the blue slots, only this time in the white slots, they appear to be testing fine. So am I wrong in guessing that either one of that other pair is bad and the problem only shows itself when it works in a pair, or that pair is just matched poorly for some reason? Either way, I'm gonna let this pair run Memtest for a while in the white slots, since the manual recommends using the white slots if you only use 2 DIMMs, and if I don't get any errors, I'll just resume normal operations only with 8GB instead of 16GB. I can live with that. :)

note - I also tried running with all four slots filled and the memory & VTT voltage upped a little, that didn't help.
 
One thing that always has let me know I should run a memtest is if my browser starts closing frequently. Browsers left open for any length of time chew on an *enormous* amount of physical and virtual memory, and if any of it is going bad, they'll crash frequently.

BSODs in games is another reason to run a memtest, but that can be GPU-related as much as anything as well.
 
dewbak75 said:
I rant Memtest with a pair in the blue slots, no errors. I ran the other pair of DIMMs in the white slots, ERRORS GALORE. So I took that pair out, and tried the pair that had tested fine in the blue slots, only this time in the white slots, they appear to be testing fine. So am I wrong in guessing that either one of that other pair is bad and the problem only shows itself when it works in a pair, or that pair is just matched poorly for some reason? Either way, I'm gonna let this pair run Memtest for a while in the white slots, since the manual recommends using the white slots if you only use 2 DIMMs, and if I don't get any errors, I'll just resume normal operations only with 8GB instead of 16GB. I can live with that. :)

note - I also tried running with all four slots filled and the memory & VTT voltage upped a little, that didn't help.
Click to expand...

It's Corsair. It has heatsinks. It has overclocked or mystery chips under those heatsinks. Those factors don't add up to great memory that always works.

More modules means more capacitance load on the memory bus, and that means slower signals and less time for the memory to respond. I think most BIOSes automatically compensate for this by choosing slower settings when more modules are in the motherboard. Also each DIMM slot varies slightly in timing, so it's not unusual for marginal memory to work in some slots but not others.

I'd look in each slot with a magnifier and strong light for crud and bent contacts, and although it rarely helps, I'd try cleaning the slot contacts with a piece of cardboard the thickness of a DIMM (stack layers if needed) and dampened with contact cleaner or alcohol. Insert and remove it several times to clean all the contacts.
 
You must log in or register to reply here.
Back
Top