MSI Kombustor is a free, dedicated GPU stress testing tool built on the FurMark engine. It’s designed to push your graphics card to 100% utilization across all shader processors, testing thermal performance, power delivery stability, and memory reliability. This guide walks through proper usage for both stability validation and overclock testing.

What MSI Kombustor Is (and Isn’t)

Kombustor is a stability and thermal testing tool, not a performance benchmark. It generates an extremely demanding OpenGL/Vulkan workload — far more intensive than typical gaming. This makes it effective for:

Verifying GPU cooling is adequate after a new install
Validating GPU overclocks and memory overclocks
Detecting VRAM errors (with the VRAM burn-in test)
Identifying early signs of GPU failure (artifacting, driver crashes)

It is not a good indicator of gaming frame rates. A GPU that passes Kombustor but fails in games usually indicates a driver issue, not hardware.

Installation

Download from msikombustor.com or through MSI’s website.
The free version is fully functional for stress testing — the paid “Pro” version adds some additional presets.
Install and run as Administrator for accurate sensor access.

The Kombustor Interface

On launch, you’ll see:

Resolution: Select your test resolution (720p for lower VRAM usage, 1080p/1440p/4K for more stress)
Preset: The rendering test to run
Options: Anti-aliasing, fullscreen mode, benchmark duration
Render API: OpenGL, OpenCL, Vulkan, DirectX 12
Score: Points-per-second metric (higher is better, but not useful for gaming comparisons)
GPU Temperature: Live readout
Power/Utilization: GPU load percentage

Choosing the Right Test Preset

Preset	What It Tests	Recommended For
Furry (default)	Shader performance	General GPU stress
Xtreme burn-in	Maximum GPU stress	Thermal/power limit validation
GPU Burn-in (Vulkan)	Vulkan API stress	Modern GPU validation
VRAM Burn-in	Memory stress with error checking	Post-overclock VRAM validation
PhysX	NVIDIA PhysX simulation	Older NVIDIA cards

For most users: start with Furry or GPU Burn-in (Vulkan) at 1080p or 1440p.

Running a Standard GPU Stress Test

Preparation

Close all applications including browsers and background processes.
Open HWiNFO64 alongside Kombustor to monitor sensors independently.
In HWiNFO64, watch:
- GPU Temperature
- GPU Memory Temperature (if available)
- GPU Power
- GPU Clock (to detect thermal throttling via clock drops)

The Test

Select Furry preset, 1080p resolution, fullscreen mode.
Click Burn-in to start.
Run for 30 minutes minimum — 60 minutes for a thorough stability check.
Watch for:
- Artifacts: Flashing pixels, color corruption, geometry errors
- Driver crash: Screen goes black, returns to desktop with “Display driver stopped responding”
- Temperature plateau: Should reach a stable max within 10–15 minutes and hold flat
- Clock drops: GPU core clock dropping below rated boost = thermal throttling

Reading the Results

After a clean 30-minute run:

Note the maximum GPU temperature — compare against your GPU’s rated TJ Max:
- NVIDIA RTX 4000 series: TJ Max = 83°C
- AMD RX 7000 series: TJ Max = 110°C (the GPU die throttles at 110°C but is designed for it)
- NVIDIA RTX 3000 series: TJ Max = 93°C
Check GPU Memory Temperature if reported:
- GDDR6: Should stay under 95°C
- GDDR6X: Can reach 110°C under load (normal for RTX 3080/3090)
- GDDR7: Should stay under 85°C under typical load
Verify GPU Clock remained stable — no significant drops from your card’s rated boost clock

VRAM Stress Test (Memory Error Detection)

The VRAM Burn-in test writes patterns to GPU memory and checks for errors. This is critical after a VRAM overclock.

Select VRAM Burn-in from the preset dropdown.
Run for 20–30 minutes.
Kombustor reports memory errors at the bottom of the window.
Zero errors = pass. Any errors indicate your memory overclock (or stock settings if no OC) is unstable.

Note: Some GPUs with degraded VRAM (especially older RTX 3000-series cards with GDDR6X) may show errors at stock speeds, indicating hardware degradation.

Stress Testing After a GPU Overclock

If you’ve overclocked your GPU with MSI Afterburner or AMD Software:

Validation Protocol

Core clock overclock only: Run Furry for 30 minutes. Look for artifacts or driver crashes.
Memory overclock only: Run VRAM Burn-in for 20 minutes. Zero errors = stable.
Both combined: Run GPU Burn-in (Vulkan) for 60 minutes as the final validation.

If you get artifacts during core OC testing:

Reduce core clock offset by 25 MHz in Afterburner and retest
Alternatively, increase GPU voltage if your card supports it (increases temps)

If you get VRAM errors:

Reduce memory clock offset by 50–100 MHz and retest
Note: GDDR6X memory errors are less reliable as error indicators — use VRAM errors + actual gameplay artifacts together

Temperature Limits: Safe Ranges

Component	Safe Operating Range	Throttle/Limit Point
GPU Core (NVIDIA)	Under 80°C ideal	83°C (throttle begins)
GPU Core (AMD RX 7000)	Under 90°C	110°C
VRAM (GDDR6)	Under 85°C	~95–100°C
VRAM (GDDR6X)	Under 105°C	~110°C
GPU Hotspot (AMD)	Under 100°C	~110°C

If your GPU hits throttle limits during Kombustor, consider:

Reseating and repasting the GPU cooler (if aftermarket heatsink)
Increasing fan curve aggressiveness in Afterburner/AMD Software
Improving case airflow
Undervolting the GPU to reduce heat at equivalent performance

Kombustor vs Other GPU Stress Tools

Tool	GPU Load Level	Memory Stress	Artifacts Visible
MSI Kombustor	Very High (100%)	Optional	Yes
3DMark Firestrike Stress	High	Moderate	Yes
Unigine Superposition	High	Moderate	Yes
Heaven Benchmark	Moderate	Low	Yes
Gaming (actual games)	Variable	High	Game crashes

For initial validation: use Kombustor. For realistic gaming stability: use Superposition or actual game playtesting after Kombustor passes.

Common Issues and Fixes

Kombustor crashes immediately

Update GPU drivers (use DDU for a clean install first)
Try a different render API (OpenGL instead of Vulkan)
Run as Administrator

Screen goes black and returns to desktop

GPU driver crash — typically indicates instability under the stress load
If at stock clocks, try DDU + fresh driver reinstall
If overclocked, reduce core or memory OC

Artifacts visible during test

Visual corruption during Kombustor almost always points to an unstable GPU state
If at stock: could indicate GPU degradation or cooling issue
If overclocked: reduce OC values

Conclusion

MSI Kombustor is an essential tool for validating GPU health after installation, driver updates, or overclocking. A 30-minute clean run with stable temperatures and zero artifacts gives you confidence that your GPU can handle any real-world workload. Pair it with HWiNFO64 for complete sensor coverage and use the VRAM burn-in test whenever you push memory frequencies.