Oh. Well, I ran memtest again last night at 2-3-3-6 and got 1 error in like 6500 cycles. By the way, I am using "memtest", the little MFC application, not "memtest 86" the DOS app. Does this matter?
Ok, I put it back to the default CAS latencies, and I got 1 error in like 300 cycles. Is this really bad enough to make it worth it to go through the RMA an be without a computer for week(s)?
RMAing the dimm might not solve the problem. It could be that the brand and/or model is not 100% compatible with your mobo.
I had a MSI board recently that really didn't like the GEIL PC3200 I put in it. The ram worked great in an ASUS board, and some Kingston ram worked fine in the MSI board.
Any memetest errors are bad. period. They will likely result in corrupt data, crashing applications, etc.
Well I put memtest86 onto a floppy disk and booted from it, but the program didn't load, just a bunch of jibberish went down the left side of the screen. What did I do wrong?
I had the same problem when I started using MEMTEST86. Apparently the install program compiles the program as it installs it on floppy. I think you need to find a WIN 98SE machine to create the install floppy. I had the same symptoms you did after creating the bootable program disk under Win 2000 Pro. I created the floppy under Win 98 and it worked fine after that.
The Gurus at the MEMTEST86 web site say that running the 7 (or more) tests three times should prove your memory stable, but others insist you should run it for an entire day.
I figure if you get through 20 or 30 iterations without a single error, it is unlikely you will get an error after 24 hours.
Also -- and anyone can correct me -- If you get memory errors consistently at the same memory locations -- it's more likely you have a defective module.
If you get memory errors at random memory locations, (it would seem) that the problem is a BIOS voltage setting -- not the integrity of manufacture.
You know what, I get 1 error after an overnight run of memtest. I have no problems with the PC, no crashes or anything, so I'm not gonna even worry about it.
That sounds like what a statistician would call an "outlier".
The reason they produce ECC (error correction and control??) memory at a higher price (72-bits instead of 64) is to generate a hashing code in the extra bits that can detect the actual bits that were "in error" and correct them. If you were to run a nuclear reactor or do hospital monitoring with a PC, you'd get ECC modules.
As I understand it, a stray Alpha particle and other kinds of cosmic detritus can cause the occasional resetting of a bit somewhere.
No certainty, but -- one error in a 24-hour time-frame? The odds are greater that it was a stray cosmic ray. . . .
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.