Increasing memory voltage can mask bad memory

[nray]

Case study: 4 Super Talent T6UB1GC5 (1.8V CAS 5 667Mhz) sticks in an MSI P6N SLI Platinum motherboard

Problem: 1 bad stick, all errors masked when run at 1.85V with three other good sticks

Symptom: entire system runs slowly with 1 bad stick and 3 good ones at 1.85V even though there are no errors of any kind

Solution: test each stick individually at stock voltage (1.8V) to find the bad stick with Memtest 86+ and remove that bad stick of memory

 

[ziddey]

hmm that's interesting. when I overclock my d9gmh 7xxxx to 1200 5-5-5, it'll pass memtest forever but will feel incredibly sluggish compared to 400 4-4-4. bandwidth and latency tests are all good

 

[dopee123]

Yeah, they need to make a more sensitive memtest that can detect this phenomenon.