Stop RAM Errors: Fix Memtest86 Failures on DDR4/DDR5

30-Second Fix: Reseat Your RAM Modules

I've seen this trip up more people than anything else. Your RAM might feel seated, but it's not fully clicked in. Shut down the PC, unplug the power cord, and press the release clips on both ends of the DIMM slots. Pull each stick straight out, then push it back in firmly until you hear a distinct click on both sides. Don't force it—if it doesn't click, you're not aligned right.

This works especially well if you moved your PC recently or swapped a cooler. The physical shift can nudge the RAM loose just enough to cause Memtest86 errors. I've fixed dozens of machines with this alone. If you're still seeing red bars in Memtest86 after this, move on.

5-Minute Fix: Test One Stick at a Time

Now we isolate the bad actor. Memtest86 errors can be one bad stick, a dead slot, or a timing mismatch. Here's the drill:

1. Remove all RAM except one stick in slot A2 (second from the CPU). That's the primary slot for dual-channel kits on most boards.
2. Run Memtest86 for at least one full pass. If it passes, that stick is good.
3. Swap to the next stick in the same slot. Test again.
4. If all sticks pass individually, the issue is either the motherboard slot or the combination of sticks. Try each stick in slot B2 next.

Real-world example: Last week a guy with a Ryzen 7800X3D and G.Skill DDR5 was getting errors only in slot A1. Swapping to A2 fixed it. That board just hated A1 with his kit. It happens.

If you find one stick that fails in any slot, that's your RMA candidate. Mark that stick and skip to the advanced section for the motherboard test.

15-Minute Fix: Check CPU Cooler Pressure and XMP

This is where things get weird but real. RAM is sensitive to physical pressure. A cooler that's too tight on one corner can warp the CPU socket just enough to cause memory training errors. Loosen your cooler screws a half turn each, in a cross pattern. Then retighten to just snug—don't torque them down. Test again with Memtest86.

Also disable XMP or EXPO in BIOS. I know, I know, that's where the speed is. But some memory controllers can't handle the rated speeds with four sticks or with certain motherboard layouts. Run Memtest86 at default JEDEC speeds (usually 2133/2400/4800). If errors vanish, your RAM kit is fine, but your board or CPU can't handle the overclock. You can try manually setting the voltage to 1.35V for DDR4 or 1.25V for DDR5, and loosen timings by one step (e.g., CL16 to CL18).

Specific scenario: On ASUS B650 boards, I've seen DDR5 at 6000MHz fail Memtest86 even though it passed on an MSI board. The fix? Set FCLK to 2000MHz manually instead of auto. That alone cleared errors for three separate builds I worked on.

Advanced 15+ Minute Fix: Motherboard Slot Test and RMA Steps

If you're still here, you've eliminated the sticks. Time to blame the slots. Use the one good stick you confirmed earlier and test it in every slot: A2, B2, A1, B1. Run one pass of Memtest86 per slot. If it fails in slot B1 but passes in A2, that slot is dead. Check your motherboard manual for warranty—most have 3-5 year coverage. Contact the manufacturer for an RMA.

If all slots pass, the problem is the combination of sticks. Try swapping their positions. Sometimes the memory controller just doesn't like a specific order. I had a Corsair Vengeance kit that only passed Memtest86 when the sticks were in slots A2 and B1, not A2 and B2. Strange but true.

Last resort: Update your BIOS. Yes, it's cliché, but it really does fix memory compatibility issues on modern boards. Check the motherboard vendor's site for beta BIOS versions that mention 'memory stability'. On my MSI Z790 board, going from BIOS 7E04v16 to 7E04v18 fixed DDR5 errors that had plagued me for weeks.

After all this, if you still get errors, the RAM is faulty. RMA it. Don't try to 'band-aid' with looser timings—that's a waste of performance. Most vendors have a 15-30 day return window for DOA, and lifetime warranty for defects. Get a replacement, then test again.

One last thing: Memtest86 errors are not always RAM. If you have a dying PSU, it can cause memory errors too. But that's a different headache for another day. Start with these steps, and you'll find the culprit 9 times out of 10.