A game freezes mid-match, the screen tears into blocks of color, or Windows drops straight to a blue screen with a code like MEMORY_MANAGEMENT or IRQL_NOT_LESS_OR_EQUAL. Reinstalling the game does nothing. Rolling back GPU drivers does nothing. Nine times out of ten, gamers chase the GPU or the power supply first, because those are the parts they can see working hard. The actual culprit is often quieter: a RAM module that fails under sustained load, especially once XMP or EXPO profiles push it past its rated timings. This tutorial walks through testing memory the way hardware technicians do it, using MemTest86, MemTest86+, and Windows’ own diagnostic tool, so you can confirm or rule out RAM before you spend money on a new GPU that was never the problem.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
Why Random Crashes and Blue Screens During Gaming Often Trace Back to RAM
Games are one of the most demanding everyday workloads a consumer PC runs. A modern title streams textures, physics data, and AI state in and out of memory dozens of times per second while the CPU and GPU both hammer the same memory controller. Marginal RAM that passes light desktop use, browsing, and even video playback can fail the instant it’s asked to sustain that kind of throughput for an hour-long raid or an open-world session. That’s why RAM-related instability so often looks “random”: it isn’t tied to one game or one setting, it’s tied to how long and how hard the memory subsystem gets pushed.
The most common fingerprints of bad or misconfigured RAM during gaming are: crashes to desktop with no error dialog, Windows Stop codes referencing memory management, textures that flicker or corrupt for a frame and then vanish, save files that become unreadable, and system freezes that require a hard reboot. None of these are exclusive to RAM, which is exactly why testing instead of guessing matters. A failing SSD, an undervolted GPU curve that’s too aggressive, or a marginal power supply rail can produce nearly identical symptoms. MemTest86’s own troubleshooting documentation is blunt about this overlap: “not all errors reported by MemTest86 are due to bad memory,” since the test also exercises the CPU, its caches, and the memory controller on the motherboard, and a fault in any of those can surface as a reported memory error, according to MemTest86’s troubleshooting documentation.
That caveat matters for how you read your results later in this guide, but it shouldn’t discourage you from testing. The same documentation is equally direct that “in the vast majority of cases errors reported by the test are valid,” so a clean, repeatable error count is still the strongest single signal you’ll get without sending parts back to a manufacturer, according to MemTest86. This tutorial targets DDR4 and DDR5 gaming desktops, gaming laptops, and handhelds like Steam Deck and ROG Ally running Windows or SteamOS as of September 2026.
Prerequisites: What You Need Before You Start
You don’t need special hardware to test RAM, but a few things need to be in place before you boot into a memory test. Gather these first so you’re not interrupted halfway through a multi-hour pass.
- A spare USB flash drive, 1GB or larger, that you can fully erase (MemTest86 and MemTest86+ both overwrite the entire drive when creating a bootable image)
- A second, working computer to create the bootable USB drive if the machine you’re testing won’t reliably boot into Windows
- MemTest86 version 11.7 Build 1000 (Free edition) or MemTest86+ v8.10, both current as of their May 2026 releases
- Access to your system’s BIOS/UEFI setup screen (know the key: typically Del, F2, or F12 depending on motherboard brand)
- Windows 10 or Windows 11 if you plan to cross-check with the built-in Windows Memory Diagnostic tool
- At least 60–120 minutes of uninterrupted time per test pass, more for high-capacity kits (see the duration table in Step 7)
- Your RAM kit’s rated speed, timings, and voltage, printed on the module label or visible in your BIOS, so you can compare against what’s actually running
Before touching MemTest86 at all, pull a quick software-side inventory of what Windows currently sees installed. This confirms capacity, speed, and slot population, and gives you a baseline to compare against BIOS readings later.
Get-CimInstance Win32_PhysicalMemory | Select-Object BankLabel, Capacity, Speed, ConfiguredClockSpeed, Manufacturer, PartNumberRun that in an elevated PowerShell window. If ConfiguredClockSpeed reads lower than the rated Speed value, XMP or EXPO isn’t active and your RAM is running at a conservative default (JEDEC) speed, which is a useful clue when you get to Step 11.
Step 1: Confirm Your Symptoms Actually Point to Memory
Before spending an hour or more on a memory test, check Windows’ own crash logs. Open Event Viewer, expand Windows Logs, and click System. Filter for Event ID 1001 with source “BugCheck,” which records the Stop code from your last blue screen. Codes worth treating as memory-suspicious include 0x0000001A (MEMORY_MANAGEMENT), 0x0000000A (IRQL_NOT_LESS_OR_EQUAL), 0x0000003B (SYSTEM_SERVICE_EXCEPTION), and 0x00000124 (WHEA_UNCORRECTABLE_ERROR). None of these guarantee bad RAM on their own, since a corrupted driver or an unstable CPU overclock can throw the same codes, but seeing two or three of them recur specifically during gaming sessions is a strong reason to move forward with a full memory test.
Also check Reliability Monitor (search “View reliability history” in the Start menu). It plots crashes on a timeline against driver and Windows updates, which helps you rule out a specific bad driver update as the actual trigger before you blame the memory sticks.
Check Motherboard Debug LEDs Before You Even Open MemTest86
If the crashes are severe enough that the PC sometimes won’t POST at all, rather than just crashing mid-game, check your motherboard’s onboard debug indicator before doing anything else. Most gaming motherboards from ASUS (Q-LED), MSI (EZ Debug LED), Gigabyte, and ASRock include four small LEDs or a two-digit debug display labeled CPU, DRAM, VGA, and BOOT near the 24-pin power connector. A memory-related boot failure lights the DRAM indicator specifically, which narrows the problem down before you’ve even created a USB drive. If the DRAM LED stays lit and the system won’t reach the OS at all, that’s a seating or compatibility issue to resolve first: reseat every module, try one stick at a time in the primary slot, and confirm the kit is on the motherboard’s Qualified Vendor List (QVL) before assuming the modules themselves are defective. Only once the system reliably boots into Windows or a USB drive does a MemTest86 pass become useful, since the tool needs a working boot environment to run its test suite at all.
Step 2: Reset BIOS to Default and Disable XMP/EXPO Before Testing
This step gets skipped constantly, and it’s the single biggest source of confusing, contradictory test results. Reboot into your BIOS/UEFI setup and load optimized or fail-safe defaults (usually F5, F9, or a menu option literally labeled “Load Defaults”). This clears any manual overclocks, disables XMP/EXPO memory profiles, and resets memory voltage to JEDEC standard. Save and exit.
Run your first MemTest86 pass at these default, non-XMP settings. The goal of this first pass is to answer one question: is the memory itself faulty, or is it only unstable at the overclocked speed the XMP/EXPO profile requests? If the modules pass cleanly at JEDEC defaults but throw errors once you re-enable XMP, the sticks aren’t defective, the specific timing/voltage combination the profile is asking for is simply too aggressive for that particular kit-motherboard-CPU combination. That’s a BIOS tuning problem, not a warranty claim, and Step 11 covers how to fix it.
Step 3: Choose the Right Memory-Testing Tool for Your Situation
There isn’t one universal answer here. Windows’ built-in tool is fastest to launch and good for a first pass; MemTest86 and MemTest86+ are more thorough and are what most hardware reviewers and system builders reach for when diagnosing a genuinely unstable build.
| Tool | Current Version (Sept. 2026) | Cost | Secure Boot Support | Best For |
|---|---|---|---|---|
| MemTest86 (Free) | 11.7 Build 1000 | Free | Yes, Microsoft-signed | Most gaming desktops and laptops, up to 16 CPU cores used |
| MemTest86 Pro | 11.7 Build 1000 | $116 one-time license | Yes | Enthusiasts wanting automated reports, up to 512 cores |
| MemTest86 Pro USB | 11.7 Build 1000 | $144 (pre-loaded drive) | Yes | Users without a spare blank USB drive |
| MemTest86+ (open source) | v8.10 | Free, GPLv2 | Partial, varies by firmware | x86, x86-64, ARM AArch64 and LoongArch64 systems, community-maintained |
| Windows Memory Diagnostic | Built into Win 10/11 | Free, no download | N/A, runs inside Windows boot | A fast first pass without creating a USB drive |
For this tutorial, MemTest86 Free is the primary tool because it covers the widest range of gaming hardware, is signed for Secure Boot, and needs no license to run a full diagnostic pass. Steps 9 and beyond also cover the Windows-native option for a quick sanity check. Independent walkthroughs of the same core process, including How-To Geek’s comparison of Windows Memory Diagnostic and MemTest86, reach a similar conclusion: the built-in Windows tool is a reasonable first pass, but a dedicated boot-time tester catches more.
If you’d rather stick with a fully open-source stack end to end, the Memtest86+ project site documents the same GPLv2-licensed v8.10 release referenced in the table above, including its Linux ISO builds for anyone testing from a non-Windows rig.
Step 4: Download and Verify MemTest86
Go to the official MemTest86 site and download the Free edition’s USB image creation tool for Windows. Avoid third-party mirror sites; MemTest86 has been bundled with adware on unofficial download portals in the past. The download is a small installer, not the bootable image itself, since the tool builds the image directly onto your USB drive during the next step.
If you prefer the fully open-source route, MemTest86+ v8.10 is available from its official project site and ships as a Windows USB installer as well as Linux ISO images, useful if you’re setting this up from a Steam Deck or Linux-based rig rather than Windows.
Step 5: Create a Bootable MemTest86 USB Drive
Plug in the spare USB drive and back up anything on it first, this process erases it completely. Run the MemTest86 USB image creation tool as Administrator, select your USB drive from the dropdown, choose the “Auto (UEFI/Legacy)” boot mode unless you know your board requires strictly one or the other, and click “Create.” The tool formats the drive, writes the bootable image, and typically finishes in under two minutes on a USB 3.0 drive.
Common mistake here: using a USB drive that’s already partitioned with multiple volumes, or one formatted with an unusual file system. If the creation tool throws an error or hangs, reformat the drive to a single FAT32 or exFAT partition in Windows’ Disk Management first, then retry.
Step 6: Boot From USB and Launch the Test
Restart the PC with the USB drive inserted and tap your motherboard’s boot menu key during POST, before Windows starts loading. This differs by brand, so use the table below if you’re not sure which key applies to your board.
| Motherboard / OEM Brand | Boot Menu Key |
|---|---|
| ASUS | F8 or Esc |
| MSI | F11 |
| Gigabyte | F12 |
| ASRock | F11 |
| Dell / HP / Lenovo laptops | F12 (varies by model) |
Select the USB drive from the boot menu. MemTest86 loads its own lightweight environment, independent of Windows, and after a few seconds automatically begins Test 1 of its default test suite. You’ll see live counters for pass number, elapsed time, and any errors found, updating in real time on screen.
Step 7: Run a Full Pass (or Two) and Know How Long to Wait
MemTest86’s default configuration runs a sequence of 13 individual test algorithms, covering everything from simple address-line checks to complex random-pattern stress tests designed to catch marginal cells that only fail under specific bit patterns. A single full pass through all 13 tests, on all installed memory, is what most guides mean by “one pass.” Duration scales directly with capacity and, to a lesser extent, memory speed.
| Installed RAM Capacity | Approximate Time for One Full Pass |
|---|---|
| 8GB | 20–35 minutes |
| 16GB | 40–70 minutes |
| 32GB | 1.5–2.5 hours |
| 64GB | 3–5 hours |
| 128GB | 6–10 hours |
For a quick confidence check, one full pass with zero errors is a reasonable minimum before assuming stability. For a genuine intermittent-crash diagnosis, most experienced builders run it overnight for two full passes, since some marginal cells only fail after the memory has been under continuous load for an extended period and won’t show up in the first 20 minutes.
Step 8: Read and Interpret MemTest86 Results
If MemTest86 finds a problem, it logs it immediately on screen with the failing test number, the memory address involved, and the expected value versus the value actually read back. A logged error looks roughly like this in the on-screen table:
Test: 7 [Block Move, 64-bit]
Pass: 1
Error confirmed at: 0x0000000412A8F110
Expected: FFFFFFFFFFFFFFFF
Actual: FFFFFFFFFFFFFFFE
CPU Core: 3 Slot: DIMM_A2Even a single confirmed error, reported once and reproduced on a repeat pass at the same or nearby address, is enough to treat the module as suspect. Zero errors across two full passes at default (non-XMP) settings is the bar most technicians use to call a kit clean. If MemTest86 reports errors but the system also locks up or reboots unexpectedly during the test itself, rather than just logging clean error rows, that points toward a broader hardware issue rather than memory alone. MemTest86’s own documentation notes that “if the system also freezes when booting into the operating system, the problem is likely hardware related,” which includes the possibility of a failing CPU, motherboard, or power delivery component rather than the memory sticks specifically, according to MemTest86’s freezing and lockups guide.
Take a photo of any error screen with your phone before rebooting. The DIMM slot label and address range matter for Step 10, and MemTest86 doesn’t save this to a persistent log file by default in the Free edition.
Step 9: Cross-Check With Windows Memory Diagnostic
Windows ships its own memory tester, and it’s worth running as a second opinion since it uses a different test pattern set than MemTest86 and requires no USB drive. Press Win+R, type the command below, and press Enter.
mdsched.exeChoose “Restart now and check for problems.” The PC reboots into a blue diagnostic screen and runs automatically; press F1 during the test if you want to switch to the more thorough “Extended” test mode instead of the default “Standard” set. When it finishes, Windows reboots normally and you won’t see a results dialog immediately, results are written to the Event Log instead. Pull them with this PowerShell query:
Get-WinEvent -FilterHashtable @{LogName='System'; ProviderName='Microsoft-Windows-MemoryDiagnostics-Results'} -MaxEvents 1 | Format-ListThis isn’t a replacement for MemTest86, its algorithms are less exhaustive, but agreement between the two tools (both clean, or both flagging errors) is a useful sanity check before you start pulling modules out of the case.
Step 10: Isolate Which Stick Is Bad (Single-DIMM Testing)
A full-kit test tells you something is wrong; it doesn’t tell you which physical module. With the system powered down and unplugged, remove all but one memory stick, leaving it in the slot the motherboard manual marks as the primary single-DIMM slot (usually A2 or DIMM_2, check your manual, as running single-channel from the wrong slot is a common source of confusing boot failures). Boot MemTest86 again with just that one module installed and run at least one full pass. Repeat with each remaining stick, one at a time, noting pass or fail for each.
If one stick consistently fails alone while the others pass cleanly, you’ve found your faulty module and can pursue a warranty RMA with confidence, most memory kits carry a limited lifetime warranty. If every stick passes individually but the full kit fails together, the problem is more likely a slot-population, motherboard QVL compatibility, or memory controller issue rather than a single bad chip, and Step 11 is where to look next.
Step 11: Rule Out BIOS, Voltage and Overclock Settings Before Blaming the RAM
If your modules passed at JEDEC defaults in Step 2 but failed once XMP or EXPO was re-enabled, the sticks themselves are very likely fine. What’s happening is that the automatic profile is requesting a frequency, timing, and voltage combination that this specific motherboard and CPU memory controller can’t sustain reliably, a mismatch that gets more common as DDR5 kits push past 6000 MT/s.
Re-enable XMP/EXPO, then manually back off one speed tier (for example, from DDR5-6400 down to DDR5-6000) and re-test. If that passes cleanly, you have a stable, if slightly slower, configuration. Alternatively, keep the rated frequency but manually loosen the secondary timings by one or two steps, or add a small amount of extra DIMM voltage (in 0.01–0.02V increments, staying within the module’s rated maximum), then re-run MemTest86. Both approaches are standard tuning steps rather than signs of defective hardware, and this is the same territory covered in more depth in a dedicated RAM overclocking walkthrough and the guide on enabling XMP and EXPO profiles correctly in the first place.
Also confirm your BIOS is current. Motherboard vendors frequently ship AGESA (AMD) or microcode (Intel) updates specifically to improve memory training and compatibility with newer, faster kits, meaning a board that couldn’t run your RAM stable six months ago may handle it fine after a BIOS update, without touching a single manual setting.
Step 12: Decide — Reseat, Downclock, RMA, or Replace
By this point you have enough data to make a real decision instead of guessing. Power down, unplug the PC, and physically reseat every module: remove each stick, inspect the gold contacts for dust or oxidation, and reinstall firmly until both slot clips click. A surprising share of “bad RAM” cases are actually a stick that never fully seated during the original build or a previous upgrade.
If reseating doesn’t resolve a repeatable error on one specific module, that module is the one to RMA or replace, not the whole kit, unless the manufacturer requires matched-kit replacement (many do, to preserve dual-channel matching). If every stick tests clean individually and only fails at rated XMP/EXPO speed, running at the slightly downclocked, verified-stable setting from Step 11 is a completely reasonable permanent fix rather than a compromise. MemTest86’s own guidance on persistent freezing issues is straightforward on this point too: when a system won’t even reliably boot into an operating system, “replacing the RAM is a first option” worth trying before deeper motherboard or CPU diagnosis, according to MemTest86.
7 Common Pitfalls When Testing RAM for Gaming Crashes
- Testing with XMP/EXPO still enabled on the first pass. This conflates “the RAM is defective” with “the RAM can’t hit its advertised speed on this board,” which need completely different fixes.
- Stopping after five or ten minutes with zero errors. Many real-world failures only appear after 30+ minutes of sustained thermal and electrical load, exactly the conditions a long gaming session creates.
- Ignoring a single error as a “fluke.” One confirmed, reproducible error at the same address on a repeat pass is meaningful data, not noise, even if the rest of the pass is clean.
- Testing all sticks together and never isolating them individually. A full-kit failure without single-DIMM testing tells you almost nothing about which physical module, or whether it’s even the RAM, is at fault.
- Using an old, worn, or already-corrupted USB drive to create the bootable image. A flaky USB drive can itself cause boot failures that look like a hardware problem but have nothing to do with your memory.
- Skipping the BIOS default reset before the first test. Testing an already-overclocked or already-XMP-enabled system muddies whether you’re diagnosing the RAM or diagnosing your own tuning.
- Assuming a passing MemTest86 result rules out every crash cause. As MemTest86’s own documentation puts it, the tool “cannot diagnose many types of PC failures,” since GPU instability, PSU sag under load, storage corruption, and software bugs all produce crash symptoms that look similar but sit entirely outside what a memory test can catch, according to MemTest86.
Troubleshooting: 9 Common RAM Testing Problems
| Problem | Likely Cause | Fix |
|---|---|---|
| USB drive doesn’t appear in boot menu | Boot mode mismatch (UEFI vs Legacy) or Secure Boot conflict | Recreate the USB using “Auto (UEFI/Legacy)” mode; temporarily disable Secure Boot if the drive still won’t show |
| MemTest86 hangs at 0% or won’t start a pass | Corrupted USB write or unstable overclock still active | Reset BIOS to defaults (Step 2), recreate the USB drive on a different port/drive |
| System reboots itself before any errors are logged | Power delivery, PSU, or CPU instability rather than RAM | Test with one stick at a time; check PSU cables and consider a GPU/CPU stress test to isolate the component |
| Errors only appear in Test 7 (Block Move) or Test 13 | These are the most sensitive tests for marginal cells and controller instability | Treat as a real result; retest that module alone to confirm |
| Windows Memory Diagnostic shows no results in Event Viewer | Log query ran before the results event was written, or logging was disabled | Re-run the PowerShell query a minute after boot completes; check “Applications and Services Logs” as a fallback location |
| System won’t boot at all with new RAM installed | Module not on motherboard’s QVL, needs a BIOS update, or not fully seated | Reseat firmly, check the motherboard QVL list, update BIOS to latest version |
| MemTest86 passes clean but crashes still happen in-game | Non-memory cause: GPU, storage, thermal throttling, or a specific game/driver bug | Check Reliability Monitor for the exact driver at time of crash; test GPU stability separately |
| One stick fails only in specific slots, not others | A damaged motherboard slot rather than a bad module | Test the “good” stick in the suspect slot; if it also fails there, the slot itself is the problem |
| Test takes far longer than the estimate table suggests | Running the Pro edition’s extended test list, or CPU throttling during the test | Check ambient case temperature; confirm which test list (Standard vs Custom) is selected |
Advanced Tips: Testing RAM on Steam Deck, ROG Ally and Handhelds
Gaming handhelds use soldered, non-removable LPDDR5/LPDDR5X memory, so you can’t isolate individual sticks the way you can on a desktop, and a hardware failure almost always means a motherboard-level repair rather than a swap. Still, confirming whether instability is memory-related is useful before sending a unit in for service. Boot the handheld into desktop mode (SteamOS’s KDE desktop on Steam Deck, or Windows desktop on ROG Ally), and run a Linux-based memory tester from a terminal if the device is running SteamOS or another Linux-based OS:
sudo pacman -S memtester
sudo memtester 2G 1That command tests 2GB of memory for one full pass without needing to reboot into a separate bootable environment, useful on a handheld where creating and booting from an external USB drive is far more cumbersome than on a desktop tower. On Windows-based handhelds like ROG Ally, mdsched.exe from Step 9 still works normally since it’s a standard Windows feature. If a handheld shows repeatable errors, that’s a strong data point for a warranty claim rather than a DIY repair, given the soldered memory. If you’re troubleshooting a handheld that won’t power on at all rather than one that’s simply crashing, that’s a different diagnostic path covered separately in the Steam Deck power troubleshooting guide.
A Complete Worked Example: Diagnosing a Crashing Gaming PC From Start to Finish
Here’s how the full process looks end to end on a realistic case: a desktop with a 32GB DDR5-6000 kit that crashes to desktop roughly once every two to three hours of gaming, with no consistent game or setting tied to the crash.
- Event Viewer showed two prior Stop codes: 0x0000001A (MEMORY_MANAGEMENT) and one unlogged hard crash with no BugCheck entry, both during gaming sessions longer than 90 minutes.
- BIOS was reset to defaults, which disabled the XMP profile and dropped the kit to its JEDEC default of DDR5-4800.
- A full MemTest86 pass (32GB, roughly 2 hours) at these default settings completed with zero errors across all 13 tests.
- XMP was re-enabled, restoring DDR5-6000, and a second full pass was run. This time, Test 7 (Block Move) logged three confirmed errors, all at addresses mapping to the same physical DIMM slot.
- Single-DIMM testing isolated the fault: the stick in slot A2 failed alone at XMP speed; the stick in slot B2 passed cleanly alone at the same XMP speed.
- As a temporary fix, XMP was re-enabled but manually stepped down one speed bin to DDR5-5600. A follow-up two-pass MemTest86 run came back completely clean on the full kit.
- The system ran three weeks of normal gaming at DDR5-5600 with zero crashes while the flagged module went through RMA under the kit’s lifetime warranty.
That sequence, reset defaults, test clean, re-enable XMP, retest, isolate by slot, apply a verified-stable interim setting, is the same pattern that applies whether you’re troubleshooting a budget build or a high-end rig with a graphics card upgrade on the way; it’s also exactly the kind of stability check worth running before any GPU overclocking session, since unstable RAM can masquerade as a bad GPU curve when both are being tuned at once.
When RAM Isn’t the Problem: What Else Causes Similar Crashes
If two full MemTest86 passes come back completely clean at both default and XMP/EXPO settings, resist the urge to run a third pass hoping for a different answer, move on to the next likely component instead. A GPU running an overly aggressive undervolt or overclock curve is one of the most common look-alikes, since it produces the same crash-to-desktop and driver-timeout symptoms under sustained load. A power supply that’s aging or undersized for a recent GPU upgrade can sag under transient spikes during demanding scenes, again mimicking memory instability. Storage corruption on the drive holding your game and page file, and outright software bugs in a specific game engine or driver version, round out the usual suspects. If you’re not sure where to look next, a full GPU and CPU stress test is the logical next diagnostic step, and if the system won’t even reliably power on to reach that point, start with the basics in the PC won’t turn on troubleshooting guide instead.
It’s also worth remembering that a memory error doesn’t automatically mean the DIMM itself is defective. As MemTest86’s documentation notes, “not all errors reported by MemTest86 are due to bad memory,” and can instead trace back to a motherboard’s memory controller, socket, or trace routing, an important distinction if you’re deciding between an RMA on the RAM kit versus the motherboard, according to MemTest86. For a first-time build or a fresh RAM installation, it’s worth reviewing the basics in a RAM installation guide to rule out a seating issue before assuming anything is actually broken.
ECC vs Non-ECC Memory: Does It Matter for Gaming Stability?
Server and workstation memory often uses ECC (Error-Correcting Code), which can detect and silently correct single-bit errors on the fly rather than crashing or corrupting data. Consumer gaming DDR4 and DDR5 kits are almost universally non-ECC, prioritizing speed and cost over error correction, which is exactly why the manual testing process in this guide matters: without ECC, a marginal bit flip has no automatic safety net and will eventually surface as a crash, a save-file corruption, or a rendering glitch instead of being silently fixed in hardware. Some workstation and a handful of newer high-end desktop platforms do support ECC UDIMMs, but for the overwhelming majority of gaming rigs, MemTest86 and the process above are the practical substitute for the error-correction gaming RAM doesn’t have built in, as explained in background material on how ECC memory works.
Frequently Asked Questions
Does MemTest86 work with Secure Boot enabled?
Yes. MemTest86’s bootable image is signed, so it boots normally on systems with Secure Boot turned on in UEFI. You generally don’t need to disable Secure Boot to run it, though if the USB drive doesn’t appear in your boot menu, temporarily disabling Secure Boot is a reasonable troubleshooting step.
How long should I run MemTest86 for the results to be reliable?
One full pass (all 13 default tests, all installed capacity) is a reasonable minimum, but many real-world intermittent failures only surface after 30 minutes or more of continuous load. For diagnosing genuine gaming crashes, running two full passes, ideally overnight, gives far more confidence than a quick 15-minute check.
Can bad RAM cause game crashes but not affect normal Windows use?
Yes, and this is common. Browsing and light desktop use touches a small, repetitive portion of memory addresses at low sustained throughput. Gaming touches far more addresses, far more aggressively, over a longer continuous period, which is exactly the kind of load that exposes marginal memory cells that idle desktop use never stresses.
What’s the difference between MemTest86 and MemTest86+?
They started from the same codebase but split into separate projects years ago. MemTest86 (version 11.7 Build 1000) is closed-source with Free, Pro, and Pro USB tiers and broader modern UEFI/Secure Boot support. MemTest86+ (v8.10) is fully open source under GPLv2 and additionally supports ARM AArch64 and LoongArch64 architectures alongside x86/x86-64. For a typical gaming PC, either works fine; MemTest86 Free is the more commonly recommended default for most builders.
Do I need to test RAM if I already ran Windows Memory Diagnostic?
It’s still worth running MemTest86 or MemTest86+ as well. Windows Memory Diagnostic’s Extended mode is reasonably thorough, but MemTest86’s test suite is generally regarded as more exhaustive for catching subtle, pattern-sensitive failures, and running both gives you a second, independently-coded opinion.
Can overclocked or XMP-enabled RAM show false errors in MemTest86?
The errors themselves are real in the sense that MemTest86 genuinely read back incorrect data, but the root cause may be an unstable overclock rather than a defective module. That’s exactly why Step 2 has you test at BIOS defaults first: a module that’s clean at JEDEC speeds but errors at XMP speeds isn’t faulty, it’s simply being asked to run faster than this specific board and CPU combination can sustain reliably.
Should I test RAM on a Steam Deck or ROG Ally the same way as a desktop?
Not exactly. Handheld memory is soldered directly to the board, so you can’t isolate individual modules by physically removing them. Use a Linux memory tester like memtester from desktop mode on SteamOS-based handhelds, or Windows’ built-in mdsched.exe on Windows-based handhelds like ROG Ally, and treat any confirmed error as a case for manufacturer service rather than a DIY component swap.
What error count means the RAM is definitely bad?
There’s no universal magic number, but as a practical rule, one or more confirmed, reproducible errors at default (non-XMP) BIOS settings, on a repeat pass, is generally treated as sufficient grounds to isolate and replace that module. Errors that only appear at manually overclocked or XMP/EXPO speeds point toward a tuning issue rather than a defective stick, and are usually resolved by dialing back speed or timings instead of an RMA.
Will testing RAM void my warranty or damage the modules?
No. MemTest86, MemTest86+, and Windows Memory Diagnostic are all read/write stress tests that run entirely in software, they don’t apply any voltage or configuration beyond what your BIOS is already set to send the modules. Running any of them, at any pass count, carries no risk to the hardware and has no effect on manufacturer warranty coverage. The only physical risk in this entire process is standard static-discharge precaution when you physically remove and reseat modules in Steps 10 and 12, the same precaution that applies to any RAM installation.
Related Coverage
- How to Fix NAT Type: Gaming Router Setup in 10 Steps [2026]
- How to Set Up 4K 120Hz VRR on PS5, Xbox, Switch 2: 12 Steps [2026]
- How to Overclock a GPU: RTX 5080/5090 & RX 9070 XT in 12 Steps [2026]
- Fix Stick Drift on PS5, Xbox, Switch, Deck: 12 Steps [2026]
- Tim Cook Steps Down as Apple CEO, Leaves $4T Legacy [2026]
