I recently rebuilt a Lenovo Legion tower around its existing Lenovo RTX 4090. The original system had random freezes and gaming crashes, especially during demanding visual scenes, so I replaced the motherboard and case and installed a Ryzen 9 9950X, while keeping the GPU, storage, and 64 GB of RAM. The crashes continued.
Overwatch reports that it lost connection to the GPU, while Diablo 4 produces a D3D error. Sometimes the display briefly goes black, the fans ramp up, and the game crashes while audio continues. Other times the entire system freezes and requires a hard reset. The failures are inconsistent but seem more likely after several hours of play or during major rendering spikes, such as loading into a new dungeon.
I have already replaced the power supply and checked the GPU power connection. FurMark can run successfully, but games still eventually crash—even after several hours at maximum settings. The system uses a Ryzen 9 9950X, 64 GB of RAM installed as four sticks, an NVIDIA RTX 4090, an ASRock X870 Riptide motherboard, and a 2 TB NVMe SSD.
When rebuilding the computer, I reset Windows on the existing drive instead of performing a completely fresh installation. Could leftover OEM drivers or software from the Lenovo system be contributing? Should I test with a clean Windows installation? Could four memory sticks or unstable memory settings be responsible, and should I disable the memory profile or reduce RAM speed? Would lowering the GPU clock be a useful diagnostic step, or does this point more strongly to a faulty graphics card, RAM, SSD, or another hardware problem?
2 Answers
A few different parts could still be involved. Four DIMMs can be harder for the memory controller to run reliably, especially if an overclocked memory profile is enabled. As a first test, load BIOS defaults, disable EXPO or any automatic memory overclock, and run the system with only two matching sticks in the recommended slots. If that helps, test the other pair and then try a lower memory speed.
Also test the RAM with a bootable memory diagnostic rather than relying only on game crashes. Check the SSD’s health and run a storage diagnostic as well. A clean Windows installation is worthwhile because resetting an OEM installation may leave vendor utilities and old drivers behind, but buying another Windows license should not be necessary. Create installation media and reinstall cleanly, then install only the motherboard chipset drivers and current NVIDIA driver.
FurMark passing does not completely clear the GPU. Games can stress the card, VRAM, drivers, CPU, and system memory in different ways. For diagnosis, try a modest GPU core and memory underclock or power-limit reduction. If that consistently prevents the crashes, it increases suspicion of marginal GPU or VRAM stability, but it is not a permanent fix.
The original processor was a Core i9-13900K, and the replacement is a Ryzen 9 9950X rather than an X3D model. Since the motherboard and CPU were changed but the symptoms remained, the GPU, RAM, SSD, drivers, or power delivery are still common points between the two builds.
I would change one thing at a time: run two RAM sticks at default settings, test memory overnight, check SSD health, perform a genuinely clean Windows/NVIDIA driver installation, and monitor GPU temperatures, hotspot temperature, VRAM temperature if available, and clock behavior while gaming. Also make sure the 4090’s power connector is fully seated and not sharply bent near the plug. If the crashes continue with known-good memory settings and a clean installation, a GPU fault becomes much more likely.
That makes sense. I had assumed the motherboard was the main suspect, but keeping the same GPU, RAM, and SSD means several likely causes were never eliminated. I’ll start with two DIMMs at BIOS defaults and then work through clean drivers and separate memory and storage tests.