Memtest86 always reports bad ram on a specific motherboard
Hi all, I've inherited an old Dell PowerEdge desktop, with 3GB DDR2 RAM in modules of 1GB. Naturally, I ran Memtest86 from one of the Ubuntu CDs. With all 3 RAM modules, memtest fails at approximately 50% of the whole test, at approximate offset of 128MB (40,000++ errors). I then tested each of those RAM modules in isolation, same result. I ran the test of 1 module in a different slot, same result. I then took one of my old DDR2 modules, and tested that in isolation, same result. I have no reason to suspect that my old DDR2 module is faulty. I ran memtest with that module some time ago in an old motherboard and had no issues. The ram was stored in an anti-static bag, but maybe it's possible for that old RAM module to get zapped in storage. Could it be possible that the RAM controller on the motherboard is partially damaged? I have ubuntu running on this machine, and so far no detectable problems. Could it be that memtest86 itself has a bug or cannot execute the tests correctly?
From what I remember, all the BIOS settings relating to RAM are set to auto, but even so the BIOS beeps at me on bootup saying that the memory configuration is not optimal.
I am trying to decide if it's worth keeping it for an experimental box or just salvaging workable parts and recycling the rest. Thanks! Srdjan
On Fri, May 09, 2014 at 10:43:59PM +0100, Srdjan Todorovic wrote:
Could it be that memtest86 itself has a bug or cannot execute the tests correctly?
Possibly, one thing to check for might be any bios updates, also try setting the RAM speeds manually in the BIOS (try underclocking the RAM) if those can't resolve it then I'd start looking at a hardware fault but this could be caused by cpu or pci cards so I'd strip the system down to a bare minimum. One other option is that if it's the same memory area then you can always try specifying the memory range linux can use to exclude the faulty area. Also, trying to run some user space memory testing tools. Adam
You also need to look at the data in the error events memtest86+ has returned, I forget the exact terminology they use but you will have a column of data stored and a column of results returned, if there is an error then these should be different, but how are they different ? is it a specific bit or group of bits which are always high or low. If multiple sticks are causing errors in any socket and you have eliminated what Adam said (always worth a bios reset to factory defaults as well) then depending on your specific CPU architecture you may be looking at a Northbridge or CPU fault, an overheating northbridge or maybe just a CPU which needs reseating or checking for bent pins. Another thing to remember is that you can get confusion in where the error is in terms of offset compared to the hardware if bank interleave is switched on. Finally the Memtest86+ Author has been very helpful in the past when I have had results back which were confusing. But I think false positives with memtest86+ are very rare. Check the offset between the expected and error and the start and finish addresses of where you get error data to work out if it is just one or a few address lines. if it is you most likely have a hardware fault.
On 09/05/14 22:43, Srdjan Todorovic wrote:
Hi all,
I've inherited an old Dell PowerEdge desktop, with 3GB DDR2 RAM in modules of 1GB. Naturally, I ran Memtest86 from one of the Ubuntu CDs. With all 3 RAM modules, memtest fails at approximately 50% of the whole test, at approximate offset of 128MB (40,000++ errors).
I then tested each of those RAM modules in isolation, same result. I ran the test of 1 module in a different slot, same result.
Try disabling USB in the BIOS. May sound unconnected but there was a bug on some Poweredge motherboards where the BIOS reserved some memory for legacy USB and it caused Memtest to fail.
participants (4)
-
Adam Bower -
Srdjan Todorovic -
Tony -
Wayne Stallwood