A stress test runs your PC at full load for an extended period to see whether anything goes wrong. It's the best way to check a new build, a new graphics card or an overclock before you trust the PC with an important game session or a long render.
Forge PC has not tested these tools for this guide. It is based on the developers' own documentation: OCBASE for OCCT, the Great Internet Mersenne Prime Search (GIMPS) for Prime95, UL for 3DMark and the HWiNFO website, plus Intel's and AMD's processor specifications.
Why stress test?
UL sums it up well: stress testing places a high load on your system for an extended period to expose stability and reliability problems, and it can also reveal faulty hardware or the need for better cooling. UL suggests running one after buying or building a new PC, upgrading your graphics card or overclocking your GPU.
Common reasons to stress test:
a new build, to catch faulty parts while they're still easy to return
an overclock or undervolt, to check it's actually stable
crashes, freezes or random restarts, to find out which part is struggling
a new cooler or case fans, to see how temperatures change
Before you start
Save your work and close other apps. A failed stress test can crash the PC.
Install a monitoring tool. HWiNFO is free and shows CPU and GPU temperatures, clock speeds, power draw and much more. Our HWiNFO and MSI Afterburner guide explains how to set it up.
Check your CPU's temperature limit. Look it up on the manufacturer's spec page. For example, Intel lists a maximum operating temperature of 105°C for the Core Ultra 9 285K, while AMD lists a Tjmax of 95°C for the Ryzen 9 9950X3D. Modern CPUs slow themselves down near these limits, so running right up against them during a stress test means your cooling has little headroom.
Laptops: plug in the charger and put the laptop on a hard surface.
Step 1: Test the CPU with Prime95 or OCCT
Prime95 has been a popular CPU stress test with overclockers and system builders for many years, according to its developers. To use it:
Download Prime95 from mersenne.org and unzip it.
Run the program and choose Just Stress Testing when asked.
Start the torture test. The developers say the default options are enough for a well-balanced stress test.
Watch temperatures in HWiNFO, and check Prime95's worker windows for errors.
OCCT is an alternative that its developer, OCBASE, describes as an all-in-one stability, stress test, benchmark and monitoring tool. OCCT Personal is free for personal use but can't be used in a commercial or business environment, and OCBASE says paying Patreon members get unlimited test duration and no wait between tests, so the free version has limits. The latest stable version at the time of writing was 17.1.11, released on 13 November 2026. As well as stressing the CPU, OCCT has memory, 3D Adaptive and Power tests.
If the PC crashes, reboots or reports errors, the CPU or its settings aren't stable. That often points to an overclock or undervolt that's too aggressive, or a cooling problem.
Step 2: Test the graphics card with 3DMark
3DMark's stress tests loop a graphics benchmark continuously. UL says that in 3DMark Advanced Edition the test loops 20 times and runs for about 10 minutes, which is usually enough to uncover stability or cooling problems.
At the end, 3DMark gives you a Frame Rate Stability score: the average frame rate of the worst loop divided by that of the best loop. UL says your system passes if it completes all loops with a Frame Rate Stability of at least 97%.
A crash or visual glitches (artifacts) may point to a GPU stability or reliability problem, according to UL.
The PC overheating and shutting down means you may need better cooling.
A low stability score usually means performance is dropping as the card heats up.
If you already use Cinebench or 3DMark for benchmarks, our Cinebench and 3DMark guide covers the basics.
Step 3: Test the memory
Memory errors can cause random crashes that are hard to trace. OCCT has a memory test, and for a thorough check from outside Windows, use MemTest86. Our MemTest86 guide walks you through it. Test your memory again after enabling an XMP or EXPO profile.
Step 4: Test the whole system and the power supply
Loading the CPU and graphics card at the same time is the hardest test for your power supply and case cooling. OCCT has a Power test for this. Run it only after the CPU and GPU have each passed on their own, and keep a close eye on temperatures.
If the PC shuts off instantly under combined load but passes the individual tests, the power supply is a likely suspect. Use our PSU calculator to check whether yours has enough headroom.
What to do if something fails
Temperatures too high: check that the CPU cooler is seated properly and fans are spinning, and read our guide on how to stop a PC overheating.
Crashes during an overclock: go back to stock settings. If the crashes stop, your overclock wasn't stable.
Crashes at stock settings: update the BIOS and drivers, test the memory, and try parts one at a time.
Errors or artifacts only on the GPU: do a clean driver install with DDU before assuming the card is faulty.
Our verdict
A stress test is the quickest way to find out whether a new build, upgrade or overclock is stable. Use Prime95 or OCCT for the CPU, 3DMark's stress test for the graphics card (a pass means at least 97% frame-rate stability across all loops), MemTest86 or OCCT for memory, and OCCT's Power test for the whole system, with HWiNFO open throughout. Stop any test if temperatures approach your CPU's or GPU's limits. Forge PC has not tested these tools for this guide; the steps come from their developers' own documentation.