And even then it was literally at the 9 hour mark. I don't want any errors but how seriously do I take a single error in 9 hours of runtime?
You take knowing that any time anything crashes or a strange behavior occurs it just might be your RAM.
i haven't used stress testing for memory or cpu oc in 6 years or so. i check in game performance to avoid any possible unstability related degradation and then I just leave it.
I've built 3 different Rigs. Only on one of them i used like half hour Ram test. All three of them are working great, no crashes and stuff.
@Danny_G13 If you tested with OCCT lower than 8.2.5 maybe that one error is nothing, look at the changelog for OCCT 8.2.5: OCCT 8.2.5 Cpu Fixed a random false positive that could ( not always) appear during a steady load test lasting for more than 6 hours. It resulted in one error detected per thread in the log window Power Fixed the same issue
I would recommend to use this for testing RAM https://www.overclock.net/threads/memory-testing-with-testmem5-tm5-with-custom-configs.1751608 Alternatively Karhu RAM test (but it's not free software)
It would bug me enough to go try out new settings. About 2 weeks ago I had a mem test going overnight for about 8 hours with about 600% coverage. Woke up to no errors, but left it running anyway while I made some breakfast, and came back about an hour later to a single error. That caused me to just default everything back to 2133Mhz for a while. Did some more settings, and managed to do over 1000% coverage with no errors, which I'm pretty confident in being stable. I used HCI MemTest. Any kind of RAM instability would easily crash Vulkan apps in my case.
I agree completely. I lived with a host of memory errors in OCCT, and I did indeed encounter memory-based crashes in games. I have now swapped my hardware and all crashes are gone. One error is indeed too many. Everything is now stock and I'm going nowhere near overclocking. Not that I did last time mind you.
Honestly I don't trust OCCT all that much but that's just me. I would download Aida64 and run a stress test. Select Stress CPU , FPU, cache, memory, and GPU. If your system can pass that for 3 or 4 hours I would say you're golden. If this is a workstation and you need absolute stability then test longer but I really don't think its necessary
Everyone has their favourite stress tests, but OCCT proved its use by reporting errors so stringently on what ended up being a flaky system. Might give Aida a bash as well for the hell of it.
Running Aida on AORUS Z590 and 10850K, no overclocking anywhere and the full stress test as you said, temps went up to 100C quickly without fans spinning at full. This test is absolutely brutal for thermals, and I'm running a full waterblock AIO on my GPU and an NZXT Kraken X72 on the CPU. With all fans now max and the thing sounding like it'll take off, it's about 80C. Reminds me of Prime 95 and small FTTs and the crazy heat that generates.
I firmly agree with this now. Yes, I was using a workable PC with only minimal issues, but minimal is still issues. I've made big changes - same RAM, but a Z590 mobo and a 11900 and now it's completely rock solid, with low temperatures, no noise and great performance. You just feel happier when you know the hardware is stable.