Re: [gentoo-user] Question about flakey RAM
On Thursday 29 Jan 2015 22:13:28 Volker Armin Hemmann wrote: > Am 28.01.2015 um 00:28 schrieb walt: > > Yesterday I installed 4GB more of RAM in this machine for a total of 8GB, > > and the machine soon began random segfaulting and even a kernel crash or > > two, so obviously I suspected the new RAM was faulty. > > > > I let memtest86+ run overnight and it found zero memory errors. Today I > > exchanged the new RAM anyway and got a different brand this time, and > > that fixed the problem. > > > > My question is why didn't memtest86+ find any errors? Could it be that > > the first RAM I bought was actually okay but this machine didn't like it > > for some reason? Both were DDR3/1333MHz, just from different > > manufacturers. > > Since this was not mentioned yet: > > Maybe because the ram was not faulty at all. > > Maybe it really operated in the range of allowed tolerances - and those > were never crossed with memtest as a very light system load. > > But with an OS booted, the CPU, graphics solution, harddisks all sucking > power like mad, your mainboard or PSU might not be able to deliver as > stable currents as the specifications demand. Some memory is more > tolerant than other. Yes, I've witnessed this too after adding 2 new memory modules of a different size to the originals and from a different manufacturer, in a box with a suspect PSU. Memtest+86 was not erroring out, but the system was crashing when put under pressure. Typically I would get errors when more than the size of the old memory started being used. This got worse over time, as the PSU components were ageing. Eventually I replaced a capacitor in the PSU and the memory problems disappeared. It has been already mentioned, but it is worth noting that some BIOS/MoBos are more sensitive to different brands of memory. In those cases I found that using the same make and size modules resolves the problems. -- Regards, Mick signature.asc Description: This is a digitally signed message part.
Re: [gentoo-user] Question about flakey RAM
Am 28.01.2015 um 00:28 schrieb walt: > Yesterday I installed 4GB more of RAM in this machine for a total of 8GB, and > the machine soon began random segfaulting and even a kernel crash or two, so > obviously I suspected the new RAM was faulty. > > I let memtest86+ run overnight and it found zero memory errors. Today I > exchanged the new RAM anyway and got a different brand this time, and > that fixed the problem. > > My question is why didn't memtest86+ find any errors? Could it be that the > first RAM I bought was actually okay but this machine didn't like it for some > reason? Both were DDR3/1333MHz, just from different manufacturers. > > > Since this was not mentioned yet: Maybe because the ram was not faulty at all. Maybe it really operated in the range of allowed tolerances - and those were never crossed with memtest as a very light system load. But with an OS booted, the CPU, graphics solution, harddisks all sucking power like mad, your mainboard or PSU might not be able to deliver as stable currents as the specifications demand. Some memory is more tolerant than other.
Re: [gentoo-user] Question about flakey RAM
On Tue, 27 Jan 2015 15:28:11 -0800 walt wrote: > Yesterday I installed 4GB more of RAM in this machine for a total of 8GB, and > the machine soon began random segfaulting and even a kernel crash or two, so > obviously I suspected the new RAM was faulty. > > I let memtest86+ run overnight and it found zero memory errors. Today I > exchanged the new RAM anyway and got a different brand this time, and > that fixed the problem. > > My question is why didn't memtest86+ find any errors? Could it be that the > first RAM I bought was actually okay but this machine didn't like it for some > reason? Both were DDR3/1333MHz, just from different manufacturers. As an addition to earlier posted comments: 1) memtest86+ has a bit fade test which is not enabled by default (at least for 4.x branch which is the latest in tree now), so you have to enable and run it manually. IIRC it is enabled by default in 5.x branch (bug pending in bugzilla). By the way 5.x have some additional tests which may find faults unknown to 4.x 2) The same frequency is not enough to guarantee memory banks compatibility. They may require different timings or, less probably, voltage. Some BIOS tuning may help here. 3) Memory may be (un)buffered, (un)registered, ecc/non-ecc. Many of these combinations are not compatible with each other. 4) In some rare cases even banks with the same parameters from different manufacturers are not compatible due to technological differences (this goes down to how logical circuits are implemented). Best regards, Andrew Savchenko pgpiXoHXB_nSL.pgp Description: PGP signature
Re: [gentoo-user] Question about flakey RAM
On 01/27/2015 03:28 PM, walt wrote: > My question is why didn't memtest86+ find any errors? Could it be that the > first RAM I bought was actually okay but this machine didn't like it for some > reason? Both were DDR3/1333MHz, just from different manufacturers. > If the timing/voltage is set wrong in the BIOS this can happen. I had "bad" memory sticks where the BIOS assumed certain timings and voltage, but when I set them to the manufacturers recommendations (manually changing voltage and timings, and no, I was not overclocking...) they were fine. I ran the memory I had in its "bad" state and memtest checked out okay after leaving it for three days straight testing. Weirdest thing I'd ever seen. Dan
Re: [gentoo-user] Question about flakey RAM
On 28/01/2015 01:28, walt wrote: > Yesterday I installed 4GB more of RAM in this machine for a total of 8GB, and > the machine soon began random segfaulting and even a kernel crash or two, so > obviously I suspected the new RAM was faulty. > > I let memtest86+ run overnight and it found zero memory errors. Today I > exchanged the new RAM anyway and got a different brand this time, and > that fixed the problem. > > My question is why didn't memtest86+ find any errors? Could it be that the > first RAM I bought was actually okay but this machine didn't like it for some > reason? Both were DDR3/1333MHz, just from different manufacturers. RAM, like everything else made in a factory, is built to tolerances. So is your CPU and motherboard. A positive result from memtest+ (something failed) is definitive - there really is a problem and it is likely the RAM. Or maybe your RAM just doesn't like your motherboard but this is rare. A negative result from memtest (nothing failed) is not definitive - it doesn't mean the RAM and your system is not faulty, it just means memtest+ didn't find anything. Sometimes you have to run memtest+ for 48 hours to trip over the problem whereas your running OS does it immediately (it's a computer, go figure) Keep in mind memtest+ is an artifical testbed - it tries it's best to find issues but it's not the same thing as your running system. And there's lots of variables: Have you overclocked? Over or under volted? Is your PSU OK, could the running system stress it out? Or maybe timing tolerances between that RAM stick your motherboard are close to the edge. I think your machine just didn;t like that RAM and it would work fine in 999 other machines. It happens sometimes, manufacturing and test rigs are not 100 perfect. They are close, but never 100% -- Alan McKinnon alan.mckin...@gmail.com