Does memtest not alway pick up errors on faulty memory?

loafer87gt

Limp Gawd
Joined
Feb 2, 2005
Messages
440
I have been pulling my hair out for about the past month trying to figure out the cause of my random BSOD on my new Kentsfield build. 90% of the time I was getting a NV4disp.dll error leading me to believe that the Nvidia Forceware drivers were at fault. Anyways, in the past weeks the errors had become more frequent, and Tuesday morning I had an IRQ not less than equal error which led me to fire up memtest to once again test my modules. Immediately, I noticed a series of errors come up. Testing each of my four 1GB sticks at a time, I found one stick that generated a series of errors with the program. Pulling it out, and running just two of the modules, things appear rock solid, and I haven't had a program crash yet.

Could this module have been the source of my problems all along? When I had my first crash about a month ago, I ran memtest to see if it was a memory related problem and it ran through all night without a hitch. Is there any chance that Windows was stressing the module more than memtest and thats why I was having the random crashes?

Anywhoo, its nice to have things nice and stable again. :)
 

mavalpha

[H]F Junkie
Joined
Jan 3, 2005
Messages
10,451
Windows (at least 3D programs) ALWAYS stresses the RAM more than MemTest. Yes, it's quite possible.
 

(cf)Eclipse

Freelance Overclocker
Joined
Feb 18, 2003
Messages
30,027
Windows (at least 3D programs) ALWAYS stresses the RAM more than MemTest. Yes, it's quite possible.

yes and no. sometimes windows memtest picks up errors normal memtest doesn't, and vice versa. i'm a proponent of using both, along with superPI.

but in short, yes, it's quite possible for ram to be unstable and memtest to not show any errors.
 

Redbeard

Official Corsair Rep.
Joined
Nov 3, 2005
Messages
1,859
memtest stable and 3d stable can be 2 very different things

While this is true, it sounds more to me like one of the sticks of RAM went bad. Memtest and Orthos are two of the based possible ways to test memory stability. Memtest 86+ is one of the best programs specifically to test for bad memory, and it's amazingly accurate the vast majority of the time. You can almost be sure that if memtest fails, you've got bad memory, and if memtest passes, whatever problem you're having USUALLY isn't related to the memory.

There have been a few cases where this wasn't the case (DDR1 Infineon revC at high frequency, for example) but the majority of the time, that's the way it rolls.
 

wdn

Limp Gawd
Joined
Dec 6, 2006
Messages
504
It is possible for Memtest to not detect and errors, but the memory is still bad anyway. But it is unlikely.

However if Memtest detects any errors then the memory is definitely bad and you can take that to the bank.
 

brachy33

Gawd
Joined
Oct 14, 2004
Messages
556
It is possible for Memtest to not detect and errors, but the memory is still bad anyway. But it is unlikely.
However if Memtest detects any errors then the memory is definitely bad and you can take that to the bank.

Don't you think you should reword that last sentence? If memtest86 detects errors during testing then it's possible that the timings and/or voltage is insufficient for the current modules and they aren't 'necessarily' bad. Case in point: My old OCZ Platinum Rev2 PC3200 sticks in a DFI Lanparty nF4 SLI-D board. It wasn't until I cranked the voltage to 2.9V (if I remember right) and manually set the timings to 2-3-2-5 1T did things work right and did Memtest pass all testing.
 

wdn

Limp Gawd
Joined
Dec 6, 2006
Messages
504
Don't you think you should reword that last sentence? If memtest86 detects errors during testing then it's possible that the timings and/or voltage is insufficient for the current modules and they aren't 'necessarily' bad. Case in point: My old OCZ Platinum Rev2 PC3200 sticks in a DFI Lanparty nF4 SLI-D board. It wasn't until I cranked the voltage to 2.9V (if I remember right) and manually set the timings to 2-3-2-5 1T did things work right and did Memtest pass all testing.

Well sure, I kind of assumed it goes without saying that if you are running memtest to diagnose defective memory sticks then you will be running running it at the manufacturers rated speeds, voltages and latencies. So yes, what you are saying is indeed correct if you run at some other settings or wrong settings you might get a bad result. I am running it to test whether the manufacturer has kept his performance promise which is what you paid good money for, or whether you have been gypped by the manufacturer. So let's say if my memory sticks that I paid a whopping $276 for are rated PC6400 4-4-4-12 @ 2.1V lets say to pick some numbers out of thin air, and the memory fails memtest-86 at that rated speed, then those sticks are DEFECTIVE by definition.
 

Damuman

Limp Gawd
Joined
Dec 1, 2006
Messages
230
So let's say if my memory sticks that I paid a whopping $276 for are rated PC6400 4-4-4-12 @ 2.1V lets say to pick some numbers out of thin air, and the memory fails memtest-86 at that rated speed, then those sticks are DEFECTIVE by definition.

I almost pissed myself laughing after that.
 

(cf)Eclipse

Freelance Overclocker
Joined
Feb 18, 2003
Messages
30,027
So let's say if my memory sticks that I paid a whopping $276 for are rated PC6400 4-4-4-12 @ 2.1V lets say to pick some numbers out of thin air, and the memory fails memtest-86 at that rated speed, then those sticks are DEFECTIVE by definition.

only if you can prove that it's not the memory controller or motherboard ;)
 

Redbeard

Official Corsair Rep.
Joined
Nov 3, 2005
Messages
1,859
only if you can prove that it's not the memory controller or motherboard ;)

Or ghosts. Don't forget ghosts.

Ghosts are the #4 cause of memory failures. I can't prove this with fact, but my memory failures only happen late at night when I feel a cold gust of air and the hair stands up on the back of my head, so it must be ghosts.

Or bees. But I don't think it's bees. There's no honey in my computer. I tried, but I killed the water pump because I guess it's not rated for something as viscous as honey.
 
Top