I've been troubleshooting random RX 9070 crashes for several months and can't determine whether the problem is the graphics card, motherboard, CPU, or software. My system has a Ryzen 7 5700X, MSI MAG B550 Tomahawk with the latest BIOS, 32 GB of DDR4-3200 memory, a new Cooler Master 850W Gold power supply, Windows 11, and three monitors. I've tested two different RX 9070 cards, and both behaved the same way. My older RX 6600 works reliably in the same system.
The crashes happen during games, video playback, or while idle. The PC usually freezes with looping audio, unresponsive input, and lost display output. Sometimes the driver recovers after a few minutes, but I generally have to restart. Device Manager may then show the card normally, or report Code 22 or Code 31 until another reboot.
Windows reports LiveKernelEvent codes including 141, 193, 1b0, a1000001, and a2000002, along with AMD watchdog and report dump files. Event Viewer also records WHEA-Logger Event 18 errors identifying the processor core, a Machine Check Exception, and a Cache Hierarchy Error on APIC ID 3. These WHEA errors generally appear when the RX 9070 crashes.
I've already tried different driver versions and clean installations, updated Windows, chipset drivers, and motherboard firmware, tested another PCIe slot, forced PCIe Gen 3, changed Resizable BAR and Above 4G Decoding, tested memory with and without XMP, used one monitor, replaced the power supply, and separated the computer from other appliances on the electrical circuit. None of this helped. The GPU is currently removed, and I'm considering replacing the B550 board while keeping the CPU, memory, power supply, storage, Windows installation, and graphics card unchanged. Could this still realistically be a motherboard or AM4 compatibility issue, or should I focus more on the CPU, memory, or processor stability?
3 Answers
It would be worth testing memory and individual cores separately with tools such as MemTest, OCCT, Prime95 small FFTs, CoreCycler, or Linpack Xtreme. Try one memory stick at a time in the recommended slots as well. A bad DIMM, memory channel, or marginal CPU core can produce cache-hierarchy WHEA errors even when normal gaming appears stable.
Since two RX 9070 cards show the same behavior while the RX 6600 is stable, the common platform components deserve more attention. A motherboard or CPU/PCIe compatibility problem is still possible, especially if the WHEA event is only appearing with the newer card. Swapping the motherboard while leaving everything else unchanged is a reasonable isolation test, provided important data is backed up first. Testing a separate operating system or live Linux environment could also help distinguish a Windows or driver issue from a hardware stability problem.
The WHEA cache-hierarchy error points more toward CPU or platform stability than a straightforward graphics-driver problem. Check whether PBO, Core Performance Boost, Curve Optimizer, or any other automatic overclocking is enabled, and monitor CPU voltage. Even default-looking settings can expose a marginal Ryzen core. A CPU stability issue could be triggered more often when the newer GPU changes system power or PCIe behavior.
I’ve already run the available memory tests and they passed, but I haven’t completely ruled out a weak CPU core. The small-FFT and per-core tests sound like the next useful step.