MemTest86 Guide: How to Test Your RAM Properly (2026)

Desktop Troubleshooting

I’ve been diagnosing and repairing PCs for sixteen years, and MemTest86 has been in my toolkit from day one. It is, without question, the gold standard for testing RAM. Every professional repair tech I know uses it. If a machine is throwing random blue screens, freezing mid-task, or crashing during memory-intensive work, MemTest86 is the first tool I reach for—every single time.

Here’s the problem: most guides tell you to download it, boot from a USB, and wait for a pass or fail. That’s barely scratching the surface. Understanding how to run MemTest86 correctly, how to actually read the results, and what to do when errors show up is where the real value lies. If you’ve ever stared at a screen full of red error lines and had no idea what to do next, this MemTest86 tutorial is for you.

I’ll walk you through the complete workflow—from choosing the right tool, to creating your bootable USB, to isolating which RAM stick is faulty. Let’s get into it.

Section 1: MemTest86 vs MemTest86+ vs Windows Memory Diagnostic — Which One Should You Use?

Before you download anything, you need to know what you’re actually getting. There are three common options, and they are not equal.

MemTest86 (PassMark)

This is the version I use professionally, and the one this guide focuses on. MemTest86 is maintained by PassMark Software and is actively developed. The current version as of 2026 supports UEFI, runs up to 13 different test algorithms, generates HTML and JSON reports, and supports RAM speeds up to DDR5. It’s free for personal use and downloadable directly from memtest86.com. There is also a paid Pro version, but for most users and repair jobs, the free version covers everything you need.

MemTest86+ (Open Source)

MemTest86+ is a community-driven, open-source fork of the original MemTest86 codebase. It went dormant for years, then saw a major revival starting around 2021. It’s a legitimate tool—especially popular in Linux circles—and it works well. The primary difference is that MemTest86 (PassMark) tends to have faster updates, better UEFI compatibility, and more polished reporting. If you’re on an older BIOS-only machine that won’t cooperate with modern UEFI builds, MemTest86+ can be a solid fallback. For most modern systems, though, I stick with the PassMark version.

Windows Memory Diagnostic

This is the built-in Windows tool—you can launch it by pressing Win + R and typing mdsched.exe. It’s better than nothing. For a quick sanity check, it works. But it runs inside the Windows environment, which means it can’t access all of your system’s RAM (the OS itself is using some of it), and it runs far fewer tests than MemTest86. It also doesn’t stress-test memory the same way. I use it occasionally when I need a fast answer and MemTest86 isn’t an option, but it is not a substitute for a real RAM test. MemTest86 runs as a standalone OS from your USB drive, which means it can test every byte of memory without interference.

Bottom line: Use MemTest86 from PassMark. It’s free, actively maintained, and the most thorough option available.

Section 2: Creating a Bootable MemTest86 USB Drive

You’ll need a USB flash drive with at least 1GB of space. Anything on the drive will be overwritten, so use a spare. If you don’t have one handy, a cheap 16GB USB drive from Amazon works perfectly for this.

Step 1: Download MemTest86

Go to memtest86.com/download.htm and click the download button for the free version. You want the USB image option (not the ISO, unless you’re burning to a disc). The download is a ZIP file. Extract it to a folder on your desktop.

Step 2: Write the Image to Your USB Drive

Inside the extracted folder, you’ll find imageUSB.exe — PassMark’s own USB writing tool. Here’s how to use it:

1. Plug your USB drive into your PC.
2. Launch imageUSB.exe (run as Administrator if needed).
3. Step 1 in imageUSB: Select your USB drive from the list. Double-check it’s the right drive — this process will erase everything on it.
4. Step 2: Make sure “Write image to UFD” is selected.
5. Step 3: The memtest86-usb.img file should already be selected. If not, browse to it in the extracted folder.
6. Step 4: Click Write. Confirm the warning that your USB will be erased.

The write process takes about a minute. When it’s done, your USB drive is ready.

Alternatively: Use Rufus

If you prefer Rufus, download it, select the MemTest86 ISO file, choose your USB drive, and click Start. Rufus handles the rest. Either method works fine — imageUSB is just a convenient single-step option since it comes in the download package.

Section 3: Booting into MemTest86 from Your USB Drive

Getting your PC to boot from USB trips up a lot of people. Here’s the cleanest way to do it.

Use the One-Time Boot Menu

You don’t need to change your permanent BIOS boot order. Most motherboards have a one-time boot menu key — press it during POST (right as the PC starts up) to select a temporary boot device. Common keys by manufacturer:

• Asus: F8
• MSI: F11
• Gigabyte: F12
• Dell: F12
• HP: F9
• Lenovo: F12 or Fn + F12
• ASRock: F11

If you’re unsure, try F12 first — it covers most systems. Spam the key repeatedly right after pressing the power button. When the boot menu appears, select your USB drive. If you see two entries for the USB (one labeled UEFI, one not), select the UEFI entry for MemTest86 v5 and later.

If MemTest86 Won’t Boot: Disable Secure Boot

Modern versions of MemTest86 are signed and should work with Secure Boot enabled. But I’ve run into machines — particularly certain Dell models and some HP consumer lines — where Secure Boot still blocks it. If you boot the USB and get kicked back to Windows or see a Secure Boot violation message, here’s what to do:

1. Enter your BIOS/UEFI settings (usually Del, F2, or F10 on startup).
2. Navigate to the Security or Boot tab (it varies by manufacturer).
3. Find Secure Boot and set it to Disabled.
4. Save and exit, then boot from the USB again.
5. After testing, go back in and re-enable Secure Boot.

On some Dell systems, the Secure Boot option is hidden under an “Advanced Setup” toggle at the top of the BIOS screen. Look for it there if you can’t find it at first glance.

Also check that Fast Boot is disabled — some boards skip USB scanning entirely when Fast Boot is on, which can make it look like the USB isn’t working.

MemTest86 Starts Automatically

Once booted, MemTest86 will display its interface and begin a countdown. After a few seconds, it automatically starts testing. You don’t need to configure anything for a standard test — just let it run.

Section 4: Running the Test Properly

This is where most guides let you down. “Just let it run” is technically correct but incomplete. Here’s how to approach the test based on your situation.

Default Configuration: 4 Passes

Out of the box, MemTest86 is configured to run 4 passes. Each pass runs the full suite of test algorithms against all of your RAM. For most casual diagnostic purposes — you’re getting crashes and want to know if RAM is the culprit — 4 passes is sufficient. If you get through 4 passes with zero errors, your RAM is almost certainly fine.

That said, some intermittent faults only show up under extended stress. For a thorough, professional-grade test, let it run overnight or longer. On a system with 16GB of RAM, expect roughly 2 hours per pass. A 32GB system will take proportionally longer.

What Each Test Is Actually Checking

MemTest86 runs up to 13 different test algorithms. Here’s a plain-English breakdown of the key ones:

• Test 0 — Address test, walking ones: Tests every address line in every memory bank using a “walking ones” bit pattern — a single 1-bit walks across all bit positions. This catches addressing faults where two memory locations get confused for each other.
• Test 1/2 — Own address test: Writes each memory location’s own address into itself, then reads it back. A mismatch means an addressing problem. Test 1 is sequential (single CPU); Test 2 runs in parallel across all CPU cores.
• Test 3/4 — Moving inversions, ones and zeros: Fills memory with all ones, then inverts to zeros and checks that the inversion propagated correctly. This is your fastest “hard fault” detector and catches the most obvious defects quickly.
• Test 5 — Moving inversions, 8-bit walking pattern: Same moving inversions approach, but uses an 8-bit wide walking pattern. Better at catching subtle errors in “wide” memory chips that use multiple bits per cell.
• Test 6 — Moving inversions, random pattern: Uses random numbers instead of fixed patterns. Particularly effective for data-sensitive errors that only appear with specific bit combinations. The random seed changes each pass, so more passes mean better coverage.
• Test 7 — Block move (movsl): Initializes memory with shifting patterns, then moves 4MB blocks around using CPU block move instructions. Stresses the memory bus and controller under real-world-like conditions.
• Test 9 — Modulo-20 random pattern: Uses the Modulo-X algorithm, which specifically bypasses CPU caching and buffering. This catches errors that moving inversions tests can miss because cache interference masked them.
• Test 11 — Bit fade test: Writes a pattern to all of memory, then pauses for several minutes before checking if any bits have changed on their own. This tests for capacitor charge retention issues — a hallmark of physically degrading RAM.

You don’t need to memorize all of this, but knowing the basics helps you understand what a test failure in a specific test actually implies about your hardware.

How Much RAM Time to Budget

• 8GB RAM: ~1 hour per pass, ~4 hours for 4 passes
• 16GB RAM: ~2–2.5 hours per pass, ~8–10 hours for 4 passes
• 32GB RAM: ~4+ hours per pass — overnight minimum
• 64GB+ RAM: May require more than one overnight session

Start the test before you go to bed. Check the results in the morning.

Section 5: Reading the Results

When MemTest86 finishes (or you check mid-run), the results screen tells you everything you need to know. Here’s how to interpret it.

The PASS Screen

If all tests complete with zero errors, MemTest86 displays a green PASS result with “0 Errors” at the bottom of the screen. That’s your all-clear. Your RAM is passing every pattern, address, and data integrity check the tool knows how to run. You can now rule out RAM as the cause of your problems and look elsewhere — GPU, storage, software, or potentially the CPU’s integrated memory controller (which MemTest86 doesn’t test directly).

When Errors Appear

Any error is a red flag. MemTest86 reports errors in a table that includes:

• Test # — which test algorithm triggered the error (e.g., Test 5: Moving inversions, 8-bit pattern)
• Failing Address — the specific memory address where the error occurred (displayed in hexadecimal)
• Expected value — what MemTest86 wrote to that address
• Actual value — what it read back
• Error count — how many errors have accumulated

A single error in the first few minutes is enough to suspect bad RAM. You don’t need to wait for thousands of errors to take action — even one is significant.

What the Failing Test Tells You

• Errors in Test 0 (walking ones address test): Likely an addressing fault — could indicate a problem with the memory controller or a physical issue with the DIMM’s address pins.
• Errors in Tests 3–6 (moving inversions): Classic memory cell defects. The RAM itself is most likely the culprit.
• Errors only in Test 9 (Modulo-20): Subtle defects that only surface when caching is bypassed. Often intermittent and temperature-sensitive.
• Errors in Test 11 (bit fade): RAM cells are losing charge over time. This points to physical degradation of the memory module.

Keep in mind: not every error is definitively bad RAM. MemTest86 is implicitly testing your CPU, L1/L2 cache, and memory controller as well. That’s why isolation is the next step.

Section 6: What to Do When MemTest86 Reports Errors

If your RAM test comes back with errors, here’s the systematic process I use to isolate exactly what’s failing. This is the part most guides skip, and it’s the most important part of the whole workflow.

Step 1: Reseat All RAM Sticks First

Before assuming the worst, power down the machine, unplug it, open the case, and reseat every RAM stick. Pull each one out completely, then firmly press it back in until both retention clips snap into place. I’ve resolved surprisingly many “RAM errors” this way. A stick that wasn’t fully seated will fail consistently but isn’t actually defective.

Step 2: Test One Stick at a Time

If errors persist after reseating, remove all but one RAM stick and run MemTest86 again. Test each stick individually, in the same slot (I use Slot A1 or the first slot by the CPU). Keep notes on which stick was installed when which test was run. This is the fastest way to pinpoint a bad module.

For systems with 4 sticks, a binary search speeds things up: test 2 sticks at once, and if errors appear, split those 2 and test each one alone.

If you’ve identified a bad RAM stick and need a replacement, check the compatibility of your system before ordering — you’ll want to match the DDR generation, speed, and ideally get a kit. DDR4 replacement kits on Amazon cover most modern systems; newer builds may need DDR5.

Step 3: Test Different Memory Slots

If a stick that fails in one slot passes in another, the problem isn’t the RAM — it’s the slot. Motherboard memory slots can go bad, particularly on older boards or systems that have been moved around. Test the same stick in every available slot. If one slot consistently produces errors and others don’t, your motherboard has a failed slot. In that case, use the remaining good slots (check your motherboard manual for the recommended slot configuration when running fewer DIMMs).

Step 4: Rule Out the Motherboard

If you have access to a known-good stick of RAM (borrow one from another system of the same DDR generation), install it and run MemTest86. If it passes, your board is fine and your original RAM is the issue. If the known-good stick also fails, you’re likely looking at a motherboard or CPU memory controller problem rather than defective RAM sticks.

A Note on RAM That’s Intermittently Bad

Some RAM defects are temperature-sensitive. A stick might pass at room temperature and fail after the system has been running for an hour and the case has warmed up. If your system only crashes after extended use, run MemTest86 after the PC has been on for a while (boot from USB after it’s already been running Windows for 30–60 minutes). Also consider that in multi-DIMM setups, middle sticks run hotter than outer ones due to reduced airflow — swapping stick positions can sometimes expose or mask these temperature-related faults.

For more symptoms that point to bad RAM before you even run a test, check out our guide on common bad RAM symptoms and signs.

Section 7: Advanced — Testing RAM at XMP/DOCP Speeds

This section is for users running RAM at XMP (Intel) or DOCP/EXPO (AMD) profiles — that is, speeds above the base JEDEC specification. If your RAM is rated at 3200 MHz, 3600 MHz, 4800 MHz, or higher and you’ve enabled an XMP profile in your BIOS, the testing process has an important nuance.

Test at JEDEC Speeds First

Always start by testing at stock JEDEC speeds. Go into your BIOS, disable the XMP/DOCP profile, and let the RAM run at its baseline speed (usually 2133 MHz for DDR4, or 4800 MHz for DDR5 at JEDEC spec). Run MemTest86 at this speed first.

If you get errors at stock JEDEC speeds, the problem is the RAM itself — it’s defective regardless of overclocking. Send it back.

If it passes at JEDEC speeds, re-enable your XMP/DOCP profile in the BIOS and run MemTest86 again.

Errors at XMP Don’t Always Mean Defective RAM

This is one of the most misunderstood aspects of RAM testing. XMP is technically an overclocking profile. It pushes your memory controller, CPU, and BIOS memory training beyond JEDEC specs. Errors at XMP can mean:

• The XMP profile is too aggressive for your specific CPU/motherboard combination
• Your memory controller doesn’t fully support the rated XMP speed
• The BIOS needs an update to handle the XMP profile correctly
• The RAM is genuinely unstable at the rated XMP speed on your platform
• The RAM has a defect that only manifests under higher clock stress

Before concluding the RAM is defective, try updating your motherboard BIOS. Manufacturers frequently release updates that improve memory compatibility and XMP training. Also try relaxing the timings slightly — if your XMP profile is CL16 at 3600 MHz, try bumping to CL18 and see if errors disappear. If slightly looser timings stabilize the system, your RAM isn’t bad; your profile just needs tuning for your specific platform.

Testing Protocol for XMP

1. Test at stock JEDEC → Pass? Good, RAM is functional.
2. Enable XMP/DOCP → Test again → Pass? XMP is stable, you’re done.
3. XMP test shows errors → Update BIOS → Retest.
4. Still errors → Try relaxing timings by 2 CL steps → Retest.
5. Errors persist even at relaxed XMP → RAM may not be compatible with your platform at rated speed, or there’s a genuine defect. Contact the manufacturer — most RAM vendors have a warranty process for this.

For reference, Corsair’s own support documentation recommends testing at stock CMOS settings first before enabling XMP, which aligns with this protocol.

If you’re upgrading RAM on a laptop and ran into compatibility issues, see our walkthrough on the Dell laptop RAM upgrade process — including slot compatibility and slot count considerations.

Conclusion: Your MemTest86 Quick Reference Checklist

MemTest86 is an indispensable tool — but only when used correctly. Running it for five minutes and calling it a clean pass is not a RAM test. Running it overnight on individual sticks, interpreting which test failed, and systematically isolating the cause — that’s a RAM test. After sixteen years of doing this professionally, the methodology above is what I follow on every single machine that comes through the door with suspected memory issues.

Here’s your quick reference checklist to take away:

• ☐ Download MemTest86 (free, PassMark) from memtest86.com
• ☐ Write to USB using the included imageUSB.exe tool
• ☐ Boot from USB using your one-time boot menu key (often F11 or F12)
• ☐ Disable Secure Boot if MemTest86 fails to launch
• ☐ Let it run at least 4 passes (overnight for thorough testing)
• ☐ Zero errors = RAM is clear; any errors = investigate further
• ☐ Reseat all sticks first before assuming anything is broken
• ☐ Test one stick at a time to isolate a failing module
• ☐ Test each stick in multiple slots to rule out a bad motherboard slot
• ☐ For XMP systems: test at JEDEC first, then re-enable XMP and retest
• ☐ XMP errors ≠ automatically bad RAM — check BIOS updates and timings first

If you’ve worked through all of this and concluded a RAM stick is bad, don’t sit on it. Defective RAM under load can corrupt files and cause data loss. Replace it promptly — replacement RAM kits are available on Amazon with same-day or next-day delivery in most areas, and the installation takes about five minutes.