To run an AIDA64 system stability test, download the AIDA64 Extreme Edition trial from aida64.com, open the app, click the System Stability Test icon, tick the components you want to stress (CPU, FPU, cache, memory, local disks, GPU), and click Start. Let the test run for at least one hour — if your PC freezes, shuts down, or shows errors, one of the selected components is unstable.
Applies to: Windows 10 (22H2) and Windows 11 (23H2, 24H2, 25H2) | Last updated: April 15, 2026
Key Takeaways
- AIDA64 Extreme Edition’s System Stability Test stresses the CPU, FPU, cache, RAM, local disks, and GPU simultaneously — one tool covers almost every component a PC can fail at.
- The software is paid, but the 30-day trial is perfectly fine for a one-off diagnostic run on a problem machine.
- Run the test for at least one hour — short runs miss thermal failures, memory errors that only surface under sustained load, and power-delivery problems.
- A freeze, BSOD, or sudden shutdown during the test means hardware failure. Re-run the test with one component at a time to isolate which part is unstable.
- AIDA64 is a stress test, not a diagnostic test. For error detection, use MemTest86 for RAM, CrystalDiskInfo and SeaTools for drives, and dedicated GPU tools for graphics card faults.
Quick Steps
- Download AIDA64 Extreme Edition (portable ZIP or installer) and extract or install it.
- Launch
aida64.exefrom the extracted folder. - Click the System Stability Test icon in the top toolbar (the green graph icon).
- Tick the components to stress: CPU, FPU, Cache, System Memory, Local Disks, GPU.
- Under Preferences, enable Prevent computer from going to sleep.
- Click Start and leave the test running for at least one hour while watching temperatures and error counts.
- Click Stop, then Save to export the graph as a PNG for your records.
When to Run a Stability Test
I reach for AIDA64 when a customer reports something vague — “the PC freezes sometimes,” “it crashes during gaming,” “random BSODs.” When the fault doesn’t reproduce on demand, you need a tool that loads every major component hard enough to provoke the failure. AIDA64’s stability test does exactly that: it runs the CPU, memory, and GPU at 100% for as long as you want, and if the system is unstable, it will show up within an hour.
You do not need to run this on a healthy PC — stress testing a stable machine for no reason is pointless wear. Use AIDA64 when something is already wrong and you need to narrow down whether the cause is hardware or software.
Step 1: Download and Install AIDA64
Go to the AIDA64 downloads page and pick AIDA64 Extreme — this is the desktop edition with the System Stability Test. Engineer, Business, and Network Audit are overkill for a home diagnostic.
You’ll see two download choices for Extreme: a portable ZIP and a full executable installer. I always pick portable for one-off diagnostic work — no installer to run, no leftover registry keys or Start menu entries to clean up when you’re done. Extract the ZIP and launch aida64.exe directly.
Note: AIDA64 is paid software with a 30-day trial. The trial is fully functional for stability testing — that’s all you need for troubleshooting a client machine. If you’re in IT and use this regularly, buy a licence.
Step 2: Start the System Stability Test
Open AIDA64. Close any trial popups. The main window shows hardware information trees (motherboard, CPU, storage, etc.) — useful for inventory, but not what you came for. On the top toolbar, find the green graph icon labelled System Stability Test and click it.
The stability test window has a panel of checkboxes at the top-left:
- Stress CPU — integer workloads
- Stress FPU — floating-point workloads (this one pushes the hottest thermals)
- Stress cache — L1/L2/L3 cache subsystem
- Stress system memory — RAM read/write patterns
- Stress local disks — reads from every drive attached to the system
- Stress GPU(s) — GPU compute load
For a first pass, tick all of them. If you already suspect a specific component, run that one alone — it makes failures faster to diagnose.
Tip: If you tick Stress GPU(s), AIDA64 will prompt to change the Windows Timeout Detection and Recovery (TDR) value. Accept the change and restart Windows — without it, Windows will kill the GPU test after 2 seconds, thinking the driver has hung.
Step 3: Configure Preferences and Monitor the Test
Before clicking Start, open the Preferences tab at the bottom of the window. On the General page, tick Prevent computer from going to sleep. If you skip this, Windows will send the PC to sleep after your configured timeout and cut the test short.
Back on the main stability test window, click Start. CPU usage spikes to 100% immediately. The tabs at the top of the graph panel let you watch different metrics live:
- Temperatures — CPU package temp, individual cores, GPU, drives. Watch for anything over 95 °C on the CPU or 85 °C on the GPU — that’s your cooler failing, or dried thermal paste.
- Cooling fans — RPM for each fan header. If a fan drops to 0 RPM mid-test, you’ve found a dead fan.
- Voltages — rail stability under load. Significant sag (e.g., 12V dropping below 11.6V) points at a failing PSU.
- CPU clock — confirms the CPU holds boost clocks and doesn’t thermal-throttle.
- CPU usage — should pin at 100% on every core. Cores stuck below that suggest parking or a pinned thread.
Step 4: Let the Test Run Long Enough
Leave the test running for at least one hour. In my experience, most thermal failures show up within 20 minutes — the cooler can’t keep up, temperatures climb past safe limits, and the CPU either throttles hard or the system shuts down. Memory errors often need longer. If the machine makes it to the one-hour mark without issue, every component you ticked is stable under full load.
A failure looks like one of these:
- Hard shutdown — the PC powers off. Almost always thermal (CPU or GPU) or power delivery (PSU, VRM).
- Freeze — the screen stops updating, input dies. Usually memory instability, CPU fault, or a failing drive under load.
- BSOD — the stop code tells you what to investigate.
WHEA_UNCORRECTABLE_ERRORis a hardware exception;MEMORY_MANAGEMENTandPAGE_FAULT_IN_NONPAGED_AREAoften point at RAM. - AIDA64 reports an error — the stability test has its own error counter. Any number above zero means something failed internally.
Step 5: Isolate the Faulty Component
If the combined test fails, re-run it one component at a time. Stop the current test, uncheck everything except one component, and start again. Work through CPU, FPU, cache, memory, disks, and GPU individually. The first one that crashes the system is your culprit.
Once you’ve identified the suspect component, follow up with a dedicated diagnostic test — AIDA64 is a stress test, so it tells you whether a component can hold up under load but doesn’t always explain why it fails. For targeted error detection, use:
- MemTest86 for RAM — detects individual memory cell errors AIDA64’s stress pass would miss.
- CrystalDiskInfo + SeaTools for drives — reads SMART data and runs surface tests.
- FurMark, OCCT, or 3DMark for GPUs — pushes the GPU harder than AIDA64 and checks for artefacts.
- Full hardware diagnostic workflow — my complete checklist for diagnosing a failing PC.
Step 6: Save the Test Results
Once you’ve decided the test has run long enough, click Stop. The graph keeps its full history. Click Save to export the session as a PNG image with the temperature, clock speed, and utilisation curves. I save every test run to a customer’s job folder — if a component fails within warranty, that graph is useful evidence, and if a machine later develops different symptoms, you can compare against the earlier baseline.
Frequently Asked Questions
Is AIDA64 free?
No. AIDA64 is paid software, but every edition ships with a 30-day trial that’s fully functional for stability testing. If you only need to diagnose one PC, the trial is enough. Regular IT work warrants buying an Extreme or Engineer licence.
How long should I run the AIDA64 stability test?
At least one hour for a meaningful result. Thermal failures usually appear within 20 minutes, but memory or power-delivery issues can take longer to manifest. Anything less than 30 minutes is not a real test — the system will look stable simply because the failure hasn’t had time to occur yet.
What does it mean if my computer shuts down during the test?
A sudden shutdown under full load is almost always thermal or power-related. Start by checking CPU and GPU temperatures against the graph — anything over 95 °C on a CPU means thermal shutdown. If temperatures are fine, suspect the PSU: a failing or undersized power supply can trip under sustained high draw. Re-test with the GPU unchecked; if it stops crashing, the PSU can’t handle CPU + GPU together.
Can AIDA64 damage my computer?
A healthy PC with working cooling will handle AIDA64 without issue — the test is designed to load components hard, but within safe operating limits. Where people get burned is running the test on a machine with a failing cooler or dried thermal paste; in that case, the CPU will thermal-throttle or the board will cut power before any real damage occurs. Modern CPUs and GPUs self-protect. Watch the temperatures, and stop the test if you see sustained readings above 95 °C.
Can I use AIDA64 to test just one component?
Yes. The stability test’s component checkboxes let you stress any combination — CPU only, GPU only, memory only, and so on. For isolating a fault after a combined test has crashed, running one component at a time is the correct workflow. Just remember AIDA64 stresses; it doesn’t diagnose errors the way MemTest86 or CrystalDiskInfo’s surface tests do.
Does AIDA64 replace Prime95 or OCCT?
Not entirely. Prime95 pushes the FPU harder than AIDA64 on Intel chips, and OCCT’s power test is better at catching VRM issues. AIDA64’s advantage is that it stresses everything in one pass — CPU, memory, storage, and GPU — which is ideal for catching “something is wrong, I don’t know what” faults. I use AIDA64 as the first pass and fall back to Prime95 / OCCT when the issue is clearly CPU-specific.
