I've been dealing with severe stability problems on my AM5 system for about a year, including random AMD driver timeouts, black screens, GPU fans suddenly running at full speed, complete system freezes, and occasional BIOS messages saying the system started in safe mode after a cold boot. This can happen even when I haven't shut down the power strip, although cutting power overnight seems to make the boot problem more likely.
The system uses a Ryzen 7 9800X3D, ASUS TUF GAMING B650-PLUS WIFI motherboard, ASRock Radeon RX 7900 XTX, 1000W Be Quiet! Pure Power 12 M PSU, and 64GB of DDR5-6000 running EXPO. The memory consists of two 16GB GoodRAM modules rated at 1.35V and two 16GB Kingston modules rated at 1.40V, with all four motherboard slots populated. The GPU uses three separate PCIe power cables and is seated correctly.
I've already updated the BIOS, chipset drivers, and graphics drivers, including a clean graphics-driver installation. Disabling EXPO did not solve the crashes. Could the four mixed memory modules, their different voltage ratings, or the AM5 memory controller be responsible? Would replacing them with one matched 2x32GB DDR5 kit likely improve stability, or should I be investigating the GPU or motherboard instead? Strict reliability is more important to me than maximum memory speed.
4 Answers
The POST recovery message strongly points toward memory training or another low-level initialization problem, so I’d solve that before blaming the graphics card. Avoid disconnecting power at the strip while testing, since a full loss of standby power can trigger another memory-training cycle. Make sure the BIOS is current, use the board’s recommended DIMM slots, and test with conservative memory settings.
For troubleshooting, don’t rely on EXPO or mixed memory settings. Load BIOS defaults, run only one matched pair in the recommended slots, and test each module independently if necessary. A memory test from a bootable utility is worthwhile too. Different rated voltages do not automatically mean the system is faulty, but the motherboard has to choose one setting for all four sticks, and four-DIMM DDR5 operation at 6000 is not guaranteed.
The memory configuration is the first thing I’d isolate. Four DDR5 modules are much harder for the AM5 memory controller to run, especially at 6000 MT/s, and mixing kits can make training and long-term stability worse even when the individual kits have similar timings. Try removing one complete kit and test with only two matching modules. If that helps, replace everything with a single matched 2x32GB kit rather than combining separate kits. You can also temporarily test around 4400 MT/s or lower to see whether the boot failures disappear.
I’m increasingly suspicious of the memory. Removing one kit is awkward because of the large CPU cooler, but I’ll test two matching modules before replacing anything.
The graphics symptoms could still be driver-related. Since every driver has not behaved well for you, try a known-stable older Adrenalin release rather than assuming the newest one is best. Completely remove the current package first, then install the older version. Also verify that the GPU power connectors are each on separate PSU cables, which you’ve already done. If crashes continue with a minimal two-DIMM setup and several driver versions, then testing another GPU or motherboard would help separate the hardware causes.
I haven’t tried rolling back to an older graphics driver yet. I’ll test that, but I’ll also simplify the memory configuration so I’m not chasing two variables at once.
Disabling EXPO did not change the behavior in my case, so I wouldn’t assume that alone rules out the RAM. Reducing the number of modules is a more useful test.