Acceptable number of acceptable memtest errors is two.
Acceptable number of acceptable memtest errors is two.
I encountered issues with a group of G skill RAM modules. After running memtest, I observed 10,000 errors in just 10 minutes. I returned the unit and received replacements yesterday. The units appear to be new, packaged properly, and labeled as refurbished, though I still wanted to verify. I tested them for seven hours and found only 22 errors. Online reports suggest any count above zero indicates faulty RAM. I plan to RMA this batch if necessary, but I want absolute certainty before labeling the next set as defective. Could it be that they’re misaligned or a BIOS setting is triggering false readings?
Above zero is problematic, right? It might indicate dirty or poor contact on the RAM or CPU. Use stock clocks without XMP settings. Overclocking could cause instability unless adjusted manually.
The fun thing about this is any change I make will take 8 hours of testing to see the results. I think XMP is on but I haven't tweaked any settings beyond that for sure. I ran memtest on a test stick of corsair ram for 8 hours while waiting for the RMA and saw zero errors so I would think that means the cpu isn't an issue. I was concerned there might be dust in the slot so I got some bright lights and a magnifying glass and really eyeballed it but it looked spotless. So I guess I'll try disabling XMP and run it again. Edit: It's running so I guess I'll post again when I wake up. I also saved the error report from the first one so if this one shows a pattern that might help narrow it down. Edit 2: it's already showing 2 errors 2 minutes in but I'm gonna let it finish in case it gives any helpful info. Edited April 18, 2023 by Araphen
@Kilrah gave a solid answer, but I’d like to include a few more points. Maybe your motherboard is worth considering too. If you’re using an older BIOS version, upgrading to the newest one would be beneficial. Recent updates often address memory issues and support newer hardware.
From the screenshots, the failing tests are 7 and 13. You can run just those two for faster testing. If they pass, proceed with the full test as a verification. The issue might be related to bad RAM or compatibility problems with your CPU/mobo. Updating the motherboard BIOS could help, but only if you're certain your system already has the latest version to avoid corruption during the update.
Thanks for your support on this matter. A few details that could have helped earlier were that only one of the returned sticks had memory errors during testing. The other, when XMP was enabled, showed no issues at all. This suggests the XMP profile and BIOS version worked well with this RAM model.
The first test with both sticks and XMP active took 7 minutes and 9 seconds, recording 23 errors (see attached memtest report #1). The second test without XMP finished in 7 minutes and 11 seconds with 883 errors (report #2).
I’m curious if the RAM was damaged by an incorrect XMP profile or if it’s just a coincidence that the second test had more errors. I ran three additional tests using known good sticks in different slots, all showing zero errors.
The motherboard is ASRock B450M Pro4 with BIOS version 3.60. I updated it to version 5.3, which claims better memory compatibility. After restarting and installing both RMA G. and Skill sticks, the system now has 231 errors after 33 minutes. I’m considering stopping this test since it’s already failed and might be best to check each stick separately.
Based on your results, I believe your motherboard and CPU are functioning properly, and the BIOS appears to be working as expected. The lower error count during the XMP test seems unexpected, so there’s no clear reason for it. Once you reach just one error, it’s best to stop testing immediately. Unless you’re interested in seeing exactly how many errors appear, I recommend returning the entire G.Skill RAM set and purchasing a new one. Mixing RAM sticks isn’t advised; opt for a compatible replacement instead.
I believe there was a minor confusion. The initial test with XMP recorded 23 errors. The next test without XMP showed 883 errors. After running two controlled tests using a reliable stick, the results remained consistent. The third test involving the uncertain RAM was also conducted without XMP and continues to run. I’m sharing the latest progress at this moment—around 6:37:26 with 2974 errors. It appears the number of faults is increasing with each test. Running memtest on such faulty hardware could potentially harm otherwise healthy RAM. My detailed approach stems from the fact that this specific configuration was sent to replace the previous failed G skill RAM. This means two sets of defective RAM in a row. What are the odds? The only viable option I can consider is testing the faulty RAM on another machine, though I worry it might be time-consuming.
It seems like two separate kits are failing in the same way, suggesting an issue with the CPU or motherboard. To verify, you might need to try a different CPU or motherboard. A local IT service could assist if you lack access to another system.