How to Stress Test a New PC Build

Why You Should Always Stress Test a New Build

When you boot Windows for the first time and everything seems fine, there is a natural temptation to call it done. The problem is that a PC can appear perfectly stable under light load and then crash the first time you throw a demanding game or render at it. A stress test forces the system to run at maximum load for a sustained period, which surfaces problems that normal browsing or even light gaming might never trigger.

The kinds of problems that stress testing catches include: thermal throttling from a poorly mounted cooler, RAM instability from a loose stick or bad XMP profile, a power supply that can't maintain stable voltage under load, and GPU issues that only appear when the card is pushed hard for more than a few minutes.

Finding these problems during a test on day one means you can fix them before they cause data loss, corrupt a game save, or cause a crash mid-render. The time investment is a couple of hours. The alternative is finding out something is wrong three weeks later in the worst possible situation.

What a Stress Test Actually Does

A stress testing tool doesn't run actual games or applications. It runs calculations specifically designed to use as much of the CPU, GPU, or RAM as possible, as continuously as possible. The goal is to push the hardware to a state it wouldn't normally reach under everyday use, and hold it there long enough to expose any weaknesses.

For the CPU, stress tests typically max out every core simultaneously for extended periods. Real workloads rarely do this, so the temperatures and power draw you see during a CPU stress test will be higher than you'd see in most games. That's intentional. If your cooler can handle a CPU stress test without thermal throttling, it will handle anything you actually throw at it.

For the GPU, stress tests generate demanding 3D scenes in a loop to push the card's temperature, power draw, and VRAM to their limits. A GPU that crashes after five minutes of stress testing but seems fine in the desktop would likely crash in demanding games too, just less predictably.

Running both at the same time also stresses the power supply, which is useful for catching PSU instability that only shows up under combined full-system load.

How to Stress Test the CPU

Before starting a CPU stress test, open a hardware monitoring tool so you can watch temperatures and clock speeds in real time. Most monitoring software shows per-core temperatures, overall package temperature, and whether the CPU is maintaining its rated boost clocks or dropping speed due to heat.

Two widely used free tools for CPU stress testing are Prime95 and OCCT. Prime95's "Small FFTs" test is the most aggressive and runs the CPU at maximum heat output. The "Blend" test is more balanced and also puts some pressure on RAM, which makes it a good all-rounder for a first test.

Run the CPU stress test for at least 20 to 30 minutes. If the system crashes, freezes, or you see errors in the stress test output window, something is wrong. If temperatures stabilise and the test runs clean, the CPU and its cooling are working correctly. A 45-minute to one-hour run gives even more confidence, especially if you want to rule out borderline stability issues.

What to watch: CPU temperature should stabilise within the first five to ten minutes. If it keeps climbing past that point, the cooler isn't keeping up. If clock speeds drop significantly mid-test, the CPU is throttling due to heat. Both are problems to investigate before calling the build done.

How to Stress Test the GPU

GPU stress testing follows the same principle as CPU testing: push the card hard and sustained, watch the numbers, and see if anything breaks. FurMark is the most commonly used GPU stress tool. It is deliberately brutal and will push GPU temperatures higher than most games. Some people call it unrealistically demanding, which is fair, but that's also what makes it useful for stability verification.

Alternatively, running a demanding game benchmark in a loop is a less aggressive but more realistic test. Many games include built-in benchmarks you can loop. This tests the GPU under conditions close to what you'll actually use it for.

During the GPU stress test, monitor the GPU temperature, the GPU clock speed, and the fan speed. On most modern GPUs, the fan curve ramps up as temperature rises, then the clock speed starts to drop if the card can't stay cool enough. This process is called thermal throttling and it means the card is hitting its temperature limit and protecting itself by reducing performance.

What to watch: Temperature stabilising below the card's rated maximum (usually in the 80-90 degree range depending on the card's design). Fan speed increasing as expected. Clock speeds holding steady rather than dropping. Any crash, driver reset, or black screen during the test is a flag that needs investigating before the build is considered stable.

Temperature Numbers and What They Mean

Temperature tolerances vary between components, but there are general ranges that tell you whether something is worth worrying about.

For CPUs: under 70 degrees during a stress test is excellent. 70 to 85 degrees is normal for higher-end chips under full load. Above 90 degrees consistently is too hot and suggests the cooler isn't doing its job, the thermal paste isn't applied correctly, or there's inadequate airflow in the case.

For GPUs: most modern cards are designed to run up to around 83 to 90 degrees under load, and they manage this automatically through their own fan curves. If you're seeing temperatures close to 95 degrees consistently, it's worth checking your case airflow. A GPU that isn't getting fresh air to pull from will thermal throttle regardless of how fast its fans spin.

The other number to watch is the VRM temperature if your monitoring software shows it. VRMs are the components on the motherboard that regulate voltage to the CPU. They can run quite hot (up to 90-100 degrees on some boards under stress), but sustained temperatures above 100 degrees on VRMs indicate either poor board design for your CPU or inadequate airflow past that part of the motherboard.

Case airflow matters a lot here. A tightly packed case with no front intake will make everything run hotter than the same components in a well-ventilated setup. If temperatures are higher than expected, check that all your case fans are installed in the correct direction before assuming a component is defective.

What Crashes and Errors Actually Tell You

A crash during a stress test is useful information, not just bad news. Different types of failures point to different components.

A blue screen of death (BSOD) during a CPU stress test with an error related to memory management often points to RAM instability, not the CPU. It's worth noting what the exact error code says. Common RAM-related BSOD codes include MEMORY_MANAGEMENT and PAGE_FAULT_IN_NONPAGED_AREA.

A system that simply powers off during a stress test, without a blue screen, usually points to the power supply or thermal protection. If the CPU or GPU hits its maximum temperature limit, it will shut the system down to protect itself. Check temperatures first. If they were fine, the PSU cutting out under load is the other likely cause.

A GPU driver crash (where the screen goes black briefly and Windows shows "display driver stopped responding and has recovered") during a GPU stress test suggests the graphics card is unstable. This can be due to insufficient PCIe power, a slot that isn't fully seated, overheating, or in some cases a defective card.

Any errors showing up in the Prime95 or OCCT output window, labeled as "HARDWARE FAILURE" or similar, mean the test has detected incorrect calculations, which is a strong indicator of RAM instability or CPU instability under load.

Memory Testing Is a Separate Step

Running Prime95 with the Blend mode does put some pressure on RAM, but it isn't the same as a dedicated memory test. For a thorough check of your RAM, MemTest86 is the standard tool. It runs from a USB drive before the operating system loads, tests every address in your RAM for errors, and takes a few hours for a complete pass.

A single pass of MemTest86 with zero errors is a good baseline. Some people run two passes for extra confidence. If you get any errors on MemTest86, your RAM is faulty, incorrectly installed, or being pushed beyond what it can handle at the current speed settings.

This is also where XMP or EXPO settings become relevant. Running RAM at its rated speed using XMP is technically an overclock, and not every RAM kit runs that profile perfectly on every motherboard. If you're getting MemTest errors with XMP enabled, try dropping the RAM speed back to the default JEDEC speed in BIOS to see if the errors go away. If they do, the XMP profile may be too aggressive for that specific board-and-RAM combination.

How Long Each Test Should Run

A 30-minute CPU stress test is a reasonable minimum. If you want high confidence, run it for an hour. Stability issues that don't show up in the first 10 minutes often appear between the 20 and 45-minute mark as temperatures reach their steady-state equilibrium and the system settles into sustained operation.

A 20 to 30-minute GPU stress test is enough for most purposes. Running a demanding game benchmark in a loop for 30 minutes on top of that gives you a real-world validation alongside the synthetic test.

MemTest86 takes the most time. One full pass across a large RAM kit can take two to four hours. Running it overnight is the easiest approach. Set it going before you sleep and check the results in the morning.

The combined combined sequence to clear a new build: CPU stress test for 30-60 minutes, GPU stress test for 20-30 minutes, MemTest86 for one or two full passes. If everything passes without errors, crashes, or unexpected temperatures, the build is stable. You can close the case, plug in your peripherals, and use the thing. That's when it actually becomes done.