BSOD With Ram

[matthew509]

---

Hello everyone,

 

Hope you’re all having a great day. I’ve run into a strange issue with my RAM. I’ve been using 64GB since the start of the year (upgraded from 32GB), and after a few months my PC started BSOD’ing with Memory Management (0x1A) and Kernel Security Check Failure errors.

I checked the system and it seemed to be RAM-related. I went into the BIOS and loaded the “Optimized Defaults,” and while that stopped the crashes, my performance dropped heavily. Anything running in the background like YouTube while gaming lags a lot.

 

Here’s the important part:
I noticed my RAM was running at 2133 MHz, even though it’s rated for 3600 MHz. When I enable DOCP and set it to the correct speed, I get the same crashes as before. I tested each stick individually and they all work fine.

What confuses me is this:

Even if I set the RAM to lower speeds like 2933 or 3200, the PC will still crash after a while under load but after it crashes its perfectly fine. However at 2133 MHz with DOCP off, everything is stable. I’ve tried many combinations DOCP off, manual speed changes, adjusting voltage, enabling/disabling Gear Down Mode, etc. and nothing seems to fix it.

 

Specs (main components):

• CPU: Ryzen 9 3900 (non-X)

• RAM: Corsair Vengeance LPX 3600 MHz (2×32GB)

• Motherboard: ROG STRIX X570-F Gaming

I did use ChatGPT to make this sound clearer since my original explanation was all over the place I'm so sorry.

[MadAnt250]

Are the sticks in the correct slots?(Pretty much not next to each other)

Try adjusting the timings, maybe turn the tRAS up a few numbers.

 

Also, try updating the BIOS first.

[StDragon]

https://www.memtest86.com/download.htm

 

Run Memtest86 from a USB flashdrive. If the memory errors are in random addresses with each test, you've got either have wrong memory timings, voltages, or a CPU problem. Don't attempt to change voltages yourself unless you know what you're doing as that could damage the CPU and/or DIMMs. From what you've stated so far, you've done good testing.

However, if the test fails in a specific memory address space over and over (repeatable), that means you've got a bad DIMM. It's easy to isolate and just run with the working good DIMM until you can get it replaced under warranty.

And yes, update your BIOS if you haven't already done so. But I highly recommend you set conservative memory timings to ensure the system is without error before flashing the BIOS. You could write-back a corrupted BIOS image making matters far worse.

[Kisai]

3 hours ago, matthew509 said:

---

 

 

Here’s the important part:
I noticed my RAM was running at 2133 MHz, even though it’s rated for 3600 MHz. When I enable DOCP and set it to the correct speed, I get the same crashes as before. I tested each stick individually and they all work fine.

What confuses me is this:

 

That's what happens when you reset the settings. DOCP/XMP is basically an overclocking mechanic.

If you can not get the DOCP/XMP to work then either the RAM is failing or the CPU is. It should always work at it's speed.

 

So 2133Mhz the correct speed for it. In theory you should be able to operate it at 3600 under ideal conditions, but in practice I've found no memory modules ever operate at DOCP/XMP for very long. Usually it works for a few months when all the parts are new. If it operates at 2133, and Memtest86 passes at 2133 but fails at 3600, you could email Corsair to see if they will warranty the RAM. If it passes both but just doesn't work under windows at 3600, then it's likely another setting (eg power management) that is destabilizing it.

 

 

 

[Imprezzion]

There's no chance of running dual rank 64GB on a 3900 at 3600Mhz stable with just docp without a whole bunch of manual tuning of voltages and such. 

 

Manual timings and voltages might get you there but more than likely it will top out at like, 3200. 

[matthew509]

So Update I ran memtest for like 1 hour and in that time i got 100 or so errors (this is with both ram in) i reduced my speed to 3200 with D.O.C.P off and crashed under load again until tomorrow I updated bios to latest stable build below will be pictures from phone if it helps in anyway, I also set voltage to auto but set timings to what it says on the ram itself. Thanks for all the help so far! 

 

FYI: First image is 8 minutes in and got 48 errors forgot to take picture of same screen 1 hour later

 

 

 

[Sawa Takahashi]

39 minutes ago, matthew509 said:

FYI: First image is 8 minutes in and got 48 errors forgot to take picture of same screen 1 hour later

RAM errors are usually indicative of a faulty stick.

You should try doing the same test again with only 1 RAM stick at a time.

Good luck !

[Bjoolz]

15 hours ago, Kisai said:

If you can not get the DOCP/XMP to work then either the RAM is failing or the CPU is. It should always work at it's speed.

Not true. The rated speed for the RAM is just for the RAM. RAM compatibility is a triangle between the motherboard, CPU and RAM. We have 7500MT/s RAM now, which there is no way in hell you'll get working on early DDR5 CPUs. The memory controllers just can't handle that speed. Any speed higher than the officially supported speed by the CPU is overclocking, and as with all overclocking, your milage WILL vary. 

 

@matthew509 Go to C:\Windows\Minidump (Sometimes Minidumps for some reason) and check if you have any minidump files. If you do, go back to the Windows folder and copy the Minidump folder itself to the Downloads folder (You can use the desktop if you don't have OneDrive syncing files). Zip the copied folder and attach it to a post.

 

This is likely just going to look like memory though, just covering all bases. 

[Kisai]

51 minutes ago, Bjoolz said:

Not true. The rated speed for the RAM is just for the RAM. RAM compatibility is a triangle between the motherboard, CPU and RAM.

Yes, but if you read the OP's first post they said it worked fine for a long period of time, which means that the problem is likely age-related (eg CPU, MB or RAM) rather than simply "oops, one part stopped working suddenly at the OC speed it's rated for" since it still worked at the intended clock speed.

 

If the question was "I got 2133 memory that has a XMP rating of 3200 but it never, ever, works at 3200" then people would be telling the user to fiddle with the voltages since 3200 is the recommended RAM speed for a R9 3900.

 

Like I said, every time I've tried to use RAM at the clock speed that was only attainable via XMP, even with earlier types of RAM, usually it would work for a while, and then out of the blue will stop working at that speed but be fine at it's rated setting.

 

Which is why the OP should run memtest86 at both the 2133 and the DOCP 3200 to verify that the RAM is failing and not simply falling out of it's socket.

 

When you have a 2x(size) configuration, it should always work at the XMP/DOCP. If you have a 4x(size) configuration, if it was bought as a set it should work at the XMP/DOCP setting, but if it was bought as two separate kits or two different size kits, it likely will never work at that setting and only the slowest stated JDEC setting, since the slowest chip will set the speed for all of it.

 

You can test a single module as well to simply rule out which module is failing, but if it was purchased as a 2x kit, then you would need to warranty both.

 

 

[Bjoolz]

19 hours ago, Kisai said:

When you have a 2x(size) configuration, it should always work at the XMP/DOCP. If you have a 4x(size) configuration, if it was bought as a set it should work at the XMP/DOCP setting, but if it was bought as two separate kits or two different size kits, it likely will never work at that setting and only the slowest stated JDEC setting, since the slowest chip will set the speed for all of it.

This is what I was referring to, this is incorrect. 

[Superxdave]

I was getting similar errors using memtest86 and it shows which CPU core associated with the issue in the RED list. Since I am undervolting my AMD 5800X CPU I raised the value by a few numbers for that specific core and it went away. -23 to -21 on the undervolt setting. Looks like your issue is on core 0 and 1.

[matthew509]

Sorry for late post was memtesting so I'll give a quick run down i tested both of them together and they was showing the errors i mentioned earlier, then i tested one stick in slot 2 a no errors for 1 hour+ then i tested the other stick in slot 2 again no errors then i tested again with both (slot 2 and 4) no errors so i was like woah its fixed maybe i re-seated it, next day worked fine, worked on star citizen etc then my discord reopened, another hour later game crash, then another game crash 1 hour later then boom i got the bsod with "Kmode exception not handled" like i said i lowered my speed to 3200 here are the minidump files (I also did turn DOCP back on after memtest) 

Minidump.zip

[Bjoolz]

On 12/5/2025 at 3:03 AM, matthew509 said:

Sorry for late post was memtesting so I'll give a quick run down i tested both of them together and they was showing the errors i mentioned earlier, then i tested one stick in slot 2 a no errors for 1 hour+ then i tested the other stick in slot 2 again no errors then i tested again with both (slot 2 and 4) no errors so i was like woah its fixed maybe i re-seated it, next day worked fine, worked on star citizen etc then my discord reopened, another hour later game crash, then another game crash 1 hour later then boom i got the bsod with "Kmode exception not handled" like i said i lowered my speed to 3200 here are the minidump files (I also did turn DOCP back on after memtest) 

Minidump.zip 2.01 MB · 0 downloads

Yeah, these just look like memory. Memory doesn't have to mean RAM, but it's usually the main suspect. Windows puts low priority data from RAM into the page file and loads it back in when needed so storage can look like memory (And memory can look like storage). The memory controller is in the CPU and if this fails it will just look like memory.

 

When it's from storage you usually get about half the dump files directly blaming storage or blaming storage drivers. I don't see here so it's likely not storage. And storage wouldn't give errors in Memtest so in your case it's 100% not storage, was just speaking in general.