: The primary tool for testing the GPU chip itself and general hardware stability. MATS (Memory Automated Test System)
Several robust, supported diagnostic tools are available for diverse use cases:
The most famous sub-program bundled within MODS. MATS is specifically designed to test the dedicated video memory (VRAM) of a graphics card. It identifies the exact memory module (or chip) that is throwing read/write errors. Key Features and Capabilities download nvidia modular diagnostic software
Before downloading third-party software, use the native tools built directly into your NVIDIA graphics driver. Right-click your desktop and open the . Click System Information in the bottom left corner.
Modified files can contain trojans or keyloggers targeting enthusiasts. : The primary tool for testing the GPU
Open-source utilities that use CUDA or OpenCL to test the reliability of GPU memory.
| Error Code | Meaning | Likely Fix | |------------|-------------------|-------------------------------------------------| | MEM 0x10 | Single-bit ECC error (uncorrected) | Card is failing – RMA if under warranty. | | MEM 0x20 | Multi-bit error – data corruption imminent | Immediate replacement required. | | BUS 0x01 | PCIe link width error (x16 downgraded to x8/x4) | Reseat card, clean PCIe slot, check motherboard. | | FB 0xFF | Framebuffer corruption | VRAM overheating – replace thermal pads. | | THM 0x80 | Hotspot delta >20°C from edge temp | Poor die contact – re-paste GPU. | It identifies the exact memory module (or chip)
: To test a discrete GPU, you often need an Integrated GPU (iGPU) or a secondary card to view the output, as the test card may not display a signal during the process.
For most users, the modular diagnostic software you need is DCGMI . It is the command-line diagnostic toolkit for NVIDIA data center GPUs.
The technician executes a command to test the VRAM. A typical command looks like this: ./mats -e 20 Use code with caution.