Hello community. I have noticed strange thing when I wanted to test new memory kit I ordered
Hello there. I am solving weird issue related to my PC.
I used to have rock stable machine, which passed with ease both memory and System test OCCT Platinum certificate. Memory test from 07/2024: https://www.ocbase.com/stabilityCertificate/669a36dd19452e20799c7255 (With TX-1000 Prime PSU)
System test from 04/2024: https://www.ocbase.com/stabilityCertificate/662d80b2b51b702c91966887 (With Focus+ 850W PSU)
Hardware inside
MB: ASUS ROG STRIX X670E-A Gaming Wi-Fi
CPU: Ryzen 9 7950X
CPU cooler: Noctua NH-D15S chromax.black (later added NA-HC3 decorative covers which are visible on photo with blueish RGB fired up)
Memory: See paragraph below
SSD_1: Seagate FireCuda 530 – 2TB (Windows , programs and non-Steam games)
SSD_2: Solidigm P44 Pro – 2TB (Steam library storage)
SSD_3: Kingston SSD FURY Renegade – 2TB (Local data storage)
PSU: Seasonic Prime TX-1000
Dedicated sound card: CREATIVE Sound Blaster Z SE
VGA: AMD Radeon RX 7900 XTX – ASUS TUF OC edition.
UEFI version is most recent non-beta version for my mobo: 2403 from 27th September 2024.
I have recently swapped RAM out for A-DATA 2×32 GB 6400 MT/s CL32 kit and I dialed down the clock to 6000 MT/s until I get 9800X3D. Basically I got this kit https://www.amazon.com/XPG-6400MHz-288-Pins-Heatsink-AX5U6400C3232G-DCLARBK/dp/B0C6HW2G8C but with ROG Se7en themed heatsink. I have tried EXPO II profile downclocked from 6400 MT/s down to 6000 MT/s, MemTest86 has passed two runs with optional DMA test with ease. The same with G-Skill kit as mentioned above. I used to have G Skill Trident Z5 2×32 GB 6000 MT/s CL30 from local retailer. I still have the kit available (I cancelled its second hand selling) https://www.alza.cz/EN/g-skill-64gb-kit-ddr5-6000mhz-cl30-trident-z5-rgb-black-d7372082.htm
However the tricky part comes to when I launch CPU+RAM test in OCCT. Usually it will crash in 60 – 120 minutes but there is no error caught in OCCT or EventViewer in Windows prior to the crash (only generic Kernel Power issue that PC shut down unexpectedly). It goes straight to black screen and PC reboots.
So in order to rule out RAM as culprit I put back G Skill kit. Dialed in XMP II profile which has out of box 6000 MT/s profile and loaded in setting in UEFI with which I was able to and again it will crash after about 60 – 120 minutes straight to black screen with no error logged. Power supply is only 5 months old, handles with ease VGAs 7900 XTX power demands and Power test in OCCT will not casue the crash. G-Skill Trident Z5 is listed in my mobo's QVL list. Any advices what I should focus on during troubleshooting?
I have also noticed, my motherboard used to auto-tune ProcDqDs to 34.3 Ohms in the past and I could pass all necessary tests (OCCT, Prime95, Karhu). Now it trains memory to High Z impedance at speeds when I dial in clocks larger than 5200 MT/S which is quite strange, because 6000 MT/s in DIV1 mode and quite loose timings should not be big deal for IMC. And I was able to pass stress tests including Prime95 in Large FFTs mode and Karhu.
I have already tried many things like reseting CMOS, reseating CPU + RAM sticks (I have also cleaned LGA pads and DIMM golden contacts using pure ethanol), putting back G Skill (old DIMMs), checking cables if they are firmly plugged in, basically all basic troubleshooting advices. I also tried single channel (only one DIMM) also with no results. I even have manually overriden ProcDqDs back to 34.3 Ω, but this also did not restore stability. Also previously would UEFI dial in 34.3 Ω automatically
Typical crash, neither OCCT nor Karhu detect errors. PC crashes straight into black screen
I was also greeted by this error in UEFI
Current values as of Oct-2024, not stable right now
Values from Jul-2024, it was stable and I could pass anything from OCCT tests