Top RAM evaluators help you assess performance accurately.
Top RAM evaluators help you assess performance accurately.
I just wanted to get opinions on the currently available methods for testing RAM. Most forum posts here don't really talk about the different options and Reddit posts tend to be contaminated by those who don't know what bootable utilities are. Passmark's classic memtest86 is still regularly updated: https://www.memtest86.com/download.htm There is an open source offshoot too called Memtest86+: https://memtest.org/ And Reddit seems very enthusiastic about TestMem5: https://github.com/CoolCmd/TestMem5 but that is in OS software and can therefore miss things. I've always used Passmark but am generally a big fan of Open Source projects.
haven't really looked into bootable memtest, but from what I've seen, it's just as dependable as any other tool. Unlike the Windows version which doesn't perform well for other options, there are alternatives like Prime95 and tools such as ycruncher or Karhu.
I've found my luck has turned out differently. TestMem5 spots mistakes that MemTest86 misses, especially when the setup is correct, and it usually does so promptly. There was an instance where it ran for over two hours of MemTest86, yet it would consistently fail within ten minutes using the 1usmus_v3 configuration. It's my preferred choice. I still run MemTest86 occasionally for my home lab setups just for ease, but realistically if the system is unstable enough to justify a memory stress test, it will fail no matter what I use. The main distinction appears in the random errors that seem like software flaws rather than hardware problems, where the memory is only slightly off—not a major defect.
TM5 performs well and I also appreciate Karhu, which has consistently shown reliability. Just remember to test both overnight on 12-hour runs before finalizing your memory setup.
A subtle distinction is ram stability and IMC stability. For practical purposes it might not matter so much which it is from a user perspective, but it could lead to this type of situation. I've had a situation in the past that will run memtest overnight without a single error, but give errors within seconds of running Prime95 that was easily proven not due to core stability. I think the access pattern is more adverse to IMC stability. TestMem might be doing something similar. I hear some people like y-cruncher to test ram too, because again it is a very ram access intensive workload with mixed reads and writes. So like all stability testing, there is no "one thing" that does everything. Run a mix of intensive workloads. I'd do at least one pass of memtest to ensure coverage and to catch any bigger problems. Then throw a selection of others at it to increase confidence.
Yes, at least one person here has noticed the issue is likely memory instability. MemTest86 didn't detect it, but TestMem5 did. The system would crash with XMP enabled, yet running MemTest86 overnight wouldn't trigger an error. Testing TM5 would cause problems within about half an hour, while individual sticks would either fail consistently or run forever. Swapping out that stick fixed XMP functionality. I agree multiple tests are helpful, but in my experience, errors tend to appear in TM5.
Thanks for the different perspectives. It seems RAM is quicker than ever yet more unstable, making testing extremely challenging. It feels as though modern RAM now has manufacturer-specific compatibility problems, similar to what GPUs have faced before.