![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
So I recently had my rig melt down and seriously stop working. After some troubleshooting help the determination was that it crashed hard and corrupted the bios. It seems that the crash was due to memory issues despite having run the memory in the rig at xmp timings for the better part of year. About half of that was with a 3700x cpu and about half witha 5900x cpu. That thread is here: https://hardforum.com/threads/diagn...and-cant-get-it-to-boot-solved.2014411/page-2
Condensed version is that I got a bunch of new parts and got everything running and still had memory issues. my 32gb kit of corsair vengeance LPX would run at spd speeds of 2133, but not at the xmp speed of 3200. I got new ram and proceeded to RMA the LPX stuff. (it is naniy chips, CL16 BTW). Errors were on the block move 64 moves test.
For the RMA corsair wants memtest86 results despite having been provided with screen shots where I have literally thousands of errors. But memtest86 takes forever to run, and I got work to do. Since I was able to revive the old board I thought was hosed as after a few attempts I was able to use flashback ot get functioning bios on it, and I had a spare PSU, video card, cooler, and CPU, I figured I'd just do a cardboard box build and run my memtest86 there. I did. One stick at a time passed. In both slots. I checked I was in XMP mode, and the bios say I am and trying the faster timings at 3200. So then I put both back in and ran it again. And it passed. I'm running both sticks a second time through the full 4 cycles and it looks like they will pass again.
My understanding is that the memory controller is on the CPU with the zen architecture. Is there a reason it might be stable on a 3700x vs a 5900x? looking back on my troubleshooting, I don't believe I ever ran the tests on anything but the 5900x before.
Is it happier in open air because of temps?
Why the heck would it suddenly be stable in memetest86? This makes me very confused and I trust it not all.
My plan was to buy another fast-ish nvme drive and RMA the ram and use the new replacement to put together a spare 3700x box. If it is passing the test should I just skip RMA and do that, or am I justified in giving the ram all the squinty eyed looks?
Condensed version is that I got a bunch of new parts and got everything running and still had memory issues. my 32gb kit of corsair vengeance LPX would run at spd speeds of 2133, but not at the xmp speed of 3200. I got new ram and proceeded to RMA the LPX stuff. (it is naniy chips, CL16 BTW). Errors were on the block move 64 moves test.
For the RMA corsair wants memtest86 results despite having been provided with screen shots where I have literally thousands of errors. But memtest86 takes forever to run, and I got work to do. Since I was able to revive the old board I thought was hosed as after a few attempts I was able to use flashback ot get functioning bios on it, and I had a spare PSU, video card, cooler, and CPU, I figured I'd just do a cardboard box build and run my memtest86 there. I did. One stick at a time passed. In both slots. I checked I was in XMP mode, and the bios say I am and trying the faster timings at 3200. So then I put both back in and ran it again. And it passed. I'm running both sticks a second time through the full 4 cycles and it looks like they will pass again.
My understanding is that the memory controller is on the CPU with the zen architecture. Is there a reason it might be stable on a 3700x vs a 5900x? looking back on my troubleshooting, I don't believe I ever ran the tests on anything but the 5900x before.
Is it happier in open air because of temps?
Why the heck would it suddenly be stable in memetest86? This makes me very confused and I trust it not all.
My plan was to buy another fast-ish nvme drive and RMA the ram and use the new replacement to put together a spare 3700x box. If it is passing the test should I just skip RMA and do that, or am I justified in giving the ram all the squinty eyed looks?