Opening
A recent XDA piece made the case for the motherboard speaker header, and the POST beep code is still the fastest way to diagnose a build that will not even reach the boot logo. The problem is that most modern PC failures happen after Windows has finished loading. A random reboot at 3 a.m., a thermal shutdown during a game, a bugcheck on wake from sleep, a drive that quietly stops responding under load. The speaker header stays silent for all of it. This is where the best apps for PC hardware troubleshooting on desktop earn their place: they read sensors the BIOS never exposes, stress specific subsystems until the weak one falls over, and parse the crash dumps Windows writes when it panics.
What to look for
We picked tools that cover the categories a real PC failure lands in, not just the pretty ones. The list favours apps that read every common sensor rather than a curated subset, so a rogue voltage rail or a fan reporting zero RPM shows up on the first pass. Log export matters: a live view is fine for a quick check, but reproducing a crash means correlating temperatures and clocks against a Windows event around the same second. Stress-test presets matter because building a realistic worst case by hand is slow. Safe defaults matter more than they sound; a badly tuned Furmark run has cooked GPUs. Where a tool is open source we call it out, because the community forks tend to outlive the originals.
Quick comparison
| Tool | Best for | OS | Log export | Cost |
|---|---|---|---|---|
| HWiNFO | Full sensor map plus system report | Windows | CSV | Free (Pro paid) |
| CPU-Z | CPU, chipset, and memory identification | Windows | TXT and HTML | Free |
| HWMonitor | Lightweight live sensor check | Windows | TXT | Free (Pro paid) |
| MemTest86 | RAM fault testing from a USB stick | UEFI (Windows, macOS, Linux hosts) | HTML report | Free (Pro paid) |
| CrystalDiskInfo | SATA and NVMe SMART health | Windows | TXT | Free |
| FurMark 2 | Sustained GPU stress and thermal check | Windows, Linux | Log file | Free |
| OCCT | CPU, memory, GPU, and PSU stability tests | Windows | CSV plus certificate | Free (Personal), paid tiers |
| BlueScreenView | Reads Windows minidump crash files | Windows | HTML, XML, CSV | Free |
The 8 best apps for PC hardware troubleshooting on desktop
1. HWiNFO for the full sensor map and system report
HWiNFO by Martin Malik is where most Windows troubleshooting sessions start. Version 8.4x reads every common motherboard sensor, per-core CPU temperatures and voltages, per-DIMM thermal sensors, GPU memory junction temperatures on modern cards, and per-NVMe drive metrics, then logs the lot to CSV for after-the-fact analysis. The System Summary and Report Writer produce a single file that captures firmware versions, driver versions, and detected components, which is what most vendors ask for before they will even open a ticket.
Where it falls short: The default sensor panel is dense, and the initial polling interval is aggressive enough to warm a laptop chassis on its own. The Pro tier gates a few remote and automation features.
Pricing:
- Free for personal use
- HWiNFO Pro around 30 dollars per year for commercial or advanced use
Platforms: Windows (installer and portable, 64-bit only in recent releases)
Download: hwinfo.com
Bottom line: The default install for anyone diagnosing a Windows PC. Reduce the polling interval, enable logging, then reproduce the fault.
2. CPU-Z for CPU, chipset, and memory identification
CPU-Z by CPUID is the small utility that answers the boring questions first: which stepping is the CPU, which chipset is the board, which XMP or EXPO profile is active, and are all four DIMMs running at the rated speed. Version 2.20 covers current Intel Core Ultra and AMD Ryzen 9000 parts, exports a text or HTML report of the identification and validation data, and includes the built-in stress test that catches an unstable overclock quickly.
Where it falls short: Not a monitoring tool. The sensor page is limited to a handful of values, so pair it with HWiNFO or HWMonitor for live readings.
Pricing:
- Free
Platforms: Windows (32-bit and 64-bit installer, portable ZIP)
Download: cpuid.com/softwares/cpu-z.html
Bottom line: Run this first when a build boots but behaves oddly. Nine times out of ten the answer is XMP silently dropping to the JEDEC default.
3. HWMonitor for a fast live sensor snapshot
HWMonitor by CPUID, in the 1.6x line, is the lightweight sibling to HWiNFO. It reads per-core CPU temperatures and package power, per-GPU sensors for NVIDIA and AMD, LPCIO voltages and fan RPMs, DIMM thermal sensors where present, and SMART attributes on connected drives. Launch time is quick, the UI shows minimum, current, and maximum for every reading, and the TXT export is easy to paste into a support forum.
Where it falls short: No graphing in the free build, no threshold alerts, and less breadth than HWiNFO on unusual boards. HWMonitor Pro adds logging and remote monitoring for a one-time fee.
Pricing:
- Free
- HWMonitor Pro around 25 dollars, one-time
Platforms: Windows (32-bit, 64-bit, and ARM64 portable builds)
Download: cpuid.com/softwares/hwmonitor.html
Bottom line: Reach for HWMonitor when the question is “what is the temperature right now” rather than “what happened at 3.14 a.m.”.
4. MemTest86 for RAM fault testing from a USB stick
MemTest86 by PassMark is still the reference tool for RAM diagnostics. Version 11.7 boots from a UEFI USB stick, so it runs before any operating system and can test the DIMMs without Windows caching, page files, or drivers in the way. The suite covers pattern tests, row hammer, and address wraparound, and it is Microsoft-signed for Secure Boot so it works on a stock Windows 11 or 12 machine without disabling anything in firmware. Errors are written to an HTML report on the same USB.
Where it falls short: Ties up the machine for hours; a full pass on 64 GB of DDR5 takes most of a working day. The free edition covers what a home user needs; the Pro edition adds ECC error injection and network deployment.
Pricing:
- Free
- MemTest86 Pro from 44 dollars, one-time
Platforms: Boots on any UEFI x86 or ARM PC. Windows, macOS, and Linux hosts can prepare the USB.
Download: memtest86.com
Bottom line: Run a full pass overnight when you suspect memory. On Linux desktops the userspace memtester is a fine sanity check; Memtest86 on bare metal is what actually settles the question.
5. CrystalDiskInfo for SATA and NVMe SMART health
CrystalDiskInfo by Crystal Dew World, current 9.9 release, is the drive-side counterpart to HWiNFO. It reads SMART attributes on SATA drives and the NVMe Health Information Log on M.2 drives in a single window, colours the overall status caution or bad when a threshold trips, and surfaces the two numbers people actually chase: temperature and total bytes written. A companion CrystalDiskMark handles the throughput side when you want to check whether the drive is behaving under real IO.
Where it falls short: Windows only. Alerts are basic. USB-to-SATA bridges are hit or miss depending on the enclosure chipset.
Pricing:
- Free (donationware)
Platforms: Windows. On Linux, smartctl from smartmontools reads the same data; on macOS, DriveDx wraps it in a UI.
Download: crystalmark.info/en/software/crystaldiskinfo
Bottom line: Install on any Windows box that stores something you cannot afford to lose. Check monthly, act when the status is not Good.
6. FurMark 2 for sustained GPU stress and thermal checks
FurMark 2 by Geeks3D, current 2.10 release, is the deliberately unrealistic GPU torture test that surfaces cooling problems most games will not. The rewrite added Vulkan support, Linux builds for x86_64 and ARM64, and a set of presets that stress at fixed resolutions rather than the old free-for-all. Pair it with HWiNFO logging and watch the memory junction temperature on an RTX 40 or RX 7000 card; if it climbs past 100 C, the thermal pads on the memory chips are usually the issue rather than the core cooler.
Where it falls short: Aggressive by design. Run it on stock clocks, not on an untested overclock, and stop the moment the temperature climbs faster than expected. Some vendor drivers flag Furmark and throttle to protect themselves, which distorts the results.
Pricing:
- Free
Platforms: Windows (32-bit and 64-bit), Linux x86_64 and ARM64
Download: geeks3d.com/furmark/downloads
Bottom line: Use for reproducing thermal shutdowns and validating a new cooler. Not for benchmarking gaming performance.
7. OCCT for stability, memory, and PSU tests in one place
OCCT by OCBase, current 2026 build, is the closest thing the free-tier market has to a full stability suite. The CPU, memory, and GPU tests each target a specific fault domain, and the power supply test drives CPU and GPU simultaneously to expose an undersized or aging PSU that a single-load test would miss. OCCT also reads sensors during a test, so voltages, temperatures, and clocks are logged against the error timeline without a second tool. The free Personal edition covers stability, monitoring, and benchmarking; paid tiers add extended runs, remote control, and certificates.
Where it falls short: The interface has grown busy as more tests landed. The stability standard for a modern CPU is at least an hour under a Large data-set pass; a five-minute check does not prove anything.
Pricing:
- Free for personal use
- Paid tiers from around 8 dollars per month for extended features and commercial use
Platforms: Windows
Download: ocbase.com
Bottom line: Run the Power test when a PC reboots under load and HWiNFO shows the CPU staying inside spec. The PSU is often the answer.
8. BlueScreenView for reading Windows crash dumps
BlueScreenView by NirSoft is the small utility that opens the minidump files Windows writes when it bugchecks and turns them into a readable table: bug check code, faulting driver, module addresses, and time of crash. It runs portable from a folder, exports HTML, XML, or CSV, and reads dumps written by every Windows release from XP through 11. Windows 11 does not always enable minidump creation by default; the Startup and Recovery panel is where to switch it back on so the tool has something to read.
Where it falls short: The faulting driver is a suggestion, not a verdict. A recent NVIDIA or storage driver often shows up because it happened to be executing when memory or PSU noise corrupted the stack. WhoCrashed by Resplendence is a reasonable alternative if you prefer plain-English summaries over raw stack information.
Pricing:
- Free
Platforms: Windows (32-bit and 64-bit portable)
Download: nirsoft.net/utils/blue_screen_view.html
Bottom line: Keep it on a USB stick with the rest of the recovery toolkit. Cross-reference the bug check code with reproductions from HWiNFO logging before you replace anything.
How to pick the right one
Start with the symptom, not the tool. Random reboots under gaming load point to power delivery: run OCCT’s Power test with HWiNFO logging, and check the 12 V rail for droop during the drop. Thermal shutdowns and clock throttling point to cooling: FurMark 2 on the GPU, an OCCT CPU pass for the processor, and HWiNFO to spot the temperature that trips the limit. Blue screens on wake, on print, or on a specific game point to a driver or a memory bit flip: BlueScreenView first for the bug check code, then MemTest86 overnight if the code is 0x0000001A, 0x0000001E, or 0x00000050. Slow file operations, stalled installs, and games that hitch on level load point to storage: CrystalDiskInfo for SMART, then a benchmark run to see whether the drive is throttling.
On Linux, lm-sensors and smartmontools cover what HWiNFO and CrystalDiskInfo do on Windows; run sensors-detect once after install to enumerate the board’s chips. On macOS, Apple Diagnostics remains the first pass (Command-D on Apple Silicon at power-on), with iStat Menus as the third-party equivalent of a lightweight sensor panel.
FAQ
Which app do we install first on a Windows PC that will not stay stable?
HWiNFO. Enable the sensor logging, reduce the polling interval to one second, and reproduce the fault while it records. Almost every other tool on this list is easier to interpret with an HWiNFO log open alongside.
Do we still need MemTest86 if Windows already includes a memory diagnostic?
Yes. The Windows Memory Diagnostic runs a shorter test suite and shares the machine with UEFI runtime services during the pass. MemTest86 boots from bare metal, runs longer patterns, and produces an HTML report at the end. When we suspect memory, we run both.
Is Furmark still safe to run on modern GPUs?
At stock clocks and with sensor logging open, yes. NVIDIA and AMD have both hardened power limits and thermal throttling since the original Furmark scare, so a healthy card will throttle rather than fail. Stop the run the moment memory junction temperature climbs past 100 C or the fans hit maximum without holding a stable temperature.
What is the desktop equivalent of these tools on Linux?
lm-sensors for temperatures, voltages, and fan speeds. smartmontools (smartctl) for SATA and NVMe SMART. memtester for a userspace RAM check, or MemTest86 on a USB stick for bare-metal testing. stress-ng for CPU and memory stability. Nvidia-smi and radeontop for GPU inspection.
How do we decide whether the problem is the CPU, the memory, or the power supply?
Run OCCT’s CPU test first; if it errors quickly and HWiNFO shows normal voltages, the CPU is suspect. If it errors with the CPU voltage or the 12 V rail sagging, the PSU is more likely. If CPU tests pass but a MemTest86 pass reports errors, the memory or the memory controller is suspect. Isolating one variable at a time is slower than swapping parts, and it is the only way to actually know.
Are these apps portable enough to keep on a USB stick?
HWiNFO, CPU-Z, HWMonitor, CrystalDiskInfo, BlueScreenView, and MemTest86 all ship in portable form. A single 8 GB USB drive is enough to carry all six plus a bootable MemTest86 partition, which is the loadout we take to any repair job.
