Back in November, I had installed the CORSAIR Dominator Platinum 128GB. It wasn’t until March that I booted the machine for the first time.
Normally, I would install the minimal components and run memtest before going too far. Problem was that everything was watercooled, so I needed everything installed before booting or I would risk burning up the gear.
Sometimes, sitting on the BIOS screen looking at the settings, it would hang. No longer responding to mouse or keyboard, you would have no choice but to hard powercycle it.
I tried running memtest originally, and it was flailing all over the place.
- If I ran with just one core, it would not see any memory problems, but it would hang.
- If I ran with more cores, it would see tons of memory problems. 30k+, especially on test number 10.
Everything was seated properly. I was really hoping the memory wasn’t bad.
A lot of people were talking about the difference between Samsung and Hynix chips. The general census was that Samsung chips did better with Ryzen. The unfortunate thing was that the only way to determine which you had was to remove the watercooling, remove the DDR4 and look at the physical label for a version number.
Trying various BIOS parameters, I found that changing the memory interleaving had the biggest impact on the error rates. On some tests (notably 3 and 4) it could bring tests down from thousands of errors to 5. Test 10 was still unruly.
I tried other settings as well. Really, everything I could get my hands on.
I tried different profiles, as well. Auto, DOCP, XMP.
I went through 52 complete OS reinstalls of Ubuntu in just over a month. To be fair, install itself usually failed as it would lock up while installing – so it was more like 52 attempts and only a dozen or so running systems.
Once running, things crashed all the time. Especially when compiling with Gradle or running Chrome. Everything would start SIGSEV’ing. The ext4 filesystem was getting corrupted regularly, causing fsck due to bad inodes. Since the system would kernel panic whenever I tried to shut down or reboot, the corruption got worse very quickly.
After running fsck a few times, the system would refuse to boot. This was due to broken symlinks, broken packages, etc.
Windows 10 Pro
I tried to install Windows 10 Pro as well. The installer would never complete because it would get to the memory diagnostic phase, and then blue screen. It didn’t matter whether I loaded any drivers or not.
I contacted Corsair to find out whether they had any particular BIOS settings I should try.
They sent me a replacement set of ram. You have two options:
- Take your system down, send the memory in, wait for new memory, put your system back together.
- Pay $2000, get the memory sent to you rather quickly, replace it in your system, ship it back (your cost), get reimbursed. This option requires end-to-end to be completed within 10 days if you want your money back.
I chose the second route. One thing to note was that UPS changed their shipping quote from $12 to $60 once I declared the value. This option, however, allowed me to continue working for a few days until I could get it swapped over.
- The old ram was v5.39.
- The new ram was v5.32. The fan on it was slightly different, as it came with a separate USB2-controller for the LEDs. iCUE, I believe.
Looking at this link, it states:
- Ver4.31 = Samsung 8GBit B-Die
- Ver5.32 = Hynix 8Gbit CFR (they are probably either AFR or MFR, though)
- Ver5.39 = Hynix 8GBit MFR
So it appears that my old and new ram are both Hynix 8GBit. The old was a newer(?) version. I am not sure what the CFR/MFR means.
No longer hangs
No longer SIGSEV’ing (though I am working on installing Arch now).
I was able to install it and get back to Elder Scrolls Online. In time for Chronicles of Elyria too!
I really should get my Rift and Omni setup.