Question BSOD when use both sticks of ram

[Naturu]

Hi, I built a new computer about 4 months ago and over the past month have been getting random BSODs and also occasionally, the computer won't post and the motherboard will have a constant orange led light (This usually happens after a BSOD, but sometimes it can happen even when it is properly shut down the previous day). Sometimes, post BSOD, I will get an error code 0xc0000428 and be asked to use a recovery tool, but for some reason I don't really need it because it would just fix itself when I restart it. I have noticed that these issues all disappear if I just use a single stick of ram, or at least it went 5 days without any errors so far. Putting in both doesn't immediately cause a BSOD or the posting issue as well, it usually happens only after a few hours of use. I also don't think its temperature as it has crashed on idle before and at the time I was tweaking the fan speeds. I distinctly remember the system temps, CPU temps were at ~50C, RAM was at ~32C and GPU was at ~30C. I am not overclocking anything as well.

The most recent BSOD was also a bit weird, usually it just completely freezes the screen and the sound becomes weird. This time, it started with google chrome having some error before it showed the BSOD.

I have tried the following:
Memtest86 with each ram stick separately 4 passes each. No errors
Memtest86 with both ram sticks in, 4 passes. No errors.
Prime95 stress test, 17 hours. No errors.
Updated Bios latest 2803 version.
Reset Windows 10 (3 times)
Check for errors with Samsung Magician. No errors.
Tried the other dual channels. (Based on manual)
I increased voltage of ram to 1.25V (1.20V is default) hoping to increase stability.

Specs:
CPU: Ryzen 5 5600X
CPU Cooler: Cooler Master ML240L RGB
Mobo: ASUS ROG Strix B550-F Gaming (WiFi 6)
GPU: EVGA RTX 3080 10GB FTW3 ULTRA G 12G
Ram: Corsair Vengeance Pro SL 16Gb x 2
Storage: Samsung 970 EVO Plus Series - 1TB
PSU: Corsair RM Series RM850
Chassis: Corsair 4000D


Here is a compressed folder of the recent BSOD dump files: https://drive.google.com/file/d/1o_sLTGmlaBKn3251XvoR3IeuZWJwESPi/view?usp=sharing

If you notice, between 12/9/22 and 17/9/22 there wasn't any BSOD. During that time, I was only using 1 Ram. If anyone has any ideas I'm all ears. Thanks

 

[Naturu]

I have not done a clean install of windows yet ever since the BSODs started, I only did the full reset thing that windows have. I will do a clean install then check back later.

I am still using the same ram kit, have not changed it yet.

EDIT: One of my friend passed me his old ram that he isn't using at the moment to test, though its more of a waiting for problems to happen? VENGEANCE® RGB PRO 16GB (2 x 8GB) DDR4 DRAM 3200MHz C16 Memory Kit (cmw16gx4m2e3200c16). I have checked on the corsair site, it is compatible with it, though I don't think it was tested on asus side.

 

[Naturu]

Hi, it has been awhile, I have been busy recently and haven't had time to continue troubleshooting. During this time, I have been going with just 1 ram stick and so far it has been stable. DOCP profile also works.

I have not done a clean install yet as it takes quite a bit of time to reinstall things.

However I have tried reseating my CPU and that some how fixed the ram issue, I could even enable DOCP profile with no issues. This however was only temporary. Within 4 days the problem comes back. I tried reseating it again and it temporarily fixes itself again.

Does this point to anything specific? Otherwise I plan to do a clean install of windows this weekend when I got more free time.

 

[Colif]

it seems strange it fixes itself and then comes back again.

Do you have fast startup turned on?
https://www.tenforums.com/tutorials/4189-turn-off-fast-startup-windows-10-a.html

I only ask as if it was off, it would make the 4 day period very bizarre.
If its on, it could be software to blame.

Most of the bsod looked like drivers so a clean install could fix it.

 

[Naturu]

Fast startup is off.

I forgot to mention, I also tried changing the cpu cooler to the stock cooler, cause I thought maybe the cooler was causing mounting pressure (the AIO pipes are a a little taut) but the problem came back still.

I guess Ill try doing a clean install this weekend and update again after a few days.

 

[Darkbreeze]

And be sure that you DO a CLEAN install, which means choosing the "Custom" method during the installation, deleting all existing partitions including the hidden ones, then installing to the bare drive without creating new partitions or doing any formatting. Windows will automatically create the required partitions and perform any necessary formatting, automatically. Also, be sure to disconnect ALL secondary drives from the system beforehand, to avoid problems. You can reconnect them after the installation is complete. It is important to do this to avoid any potential for problems with the boot partition being accidentally created on another drive, or allowing the system to use a hidden boot partition on another drive that might have had Windows on it in the past and never had it removed since it is a hidden partition that cannot be seen in Windows explorer.

 

[Naturu]

I have just finished doing a clean install of windows. I have updated drivers through windows update, as well as those mentioned earlier from previous posts. And I have also done a clean install of my display drivers.

I also reseated my CPU again. So right now, I have both ram sticks in. Should I enable DOCP profile, or just let it run default first and see how it goes for a week?

 

[Colif]

i would run at default ram speed for a week and see how you go.

 

[Darkbreeze]

After you determine that it is ok running at the default speed, then enable DOCP.

 

[Naturu]

Hi, unfortunately it did not even last 4 days this time. Today, when I turned it on, it was stuck on the orange light and could not post again. After restarting 2 times (holding power button to turn off and turning it on after), it managed to post and get to the desktop screen. However, within seconds the screen froze, but there was no blue screen. Subsequently, I tried restarting it 5 times and it never posted.

Currently, I have removed 1 stick so that I can turn it on and continue doing stuff. There was no bluescreen dump file. Is there anything else I can try?

 

[Darkbreeze]

Pull the CPU and check for bent pins. If there are none, then I think I'd RMA the motherboard.

 

[Naturu]

Each time I reseat my CPU, I have always checked for bent pins. The most i see is some thermal paste which I wiped away with alcohol wipes. So the likely problem is motherboard? Is there any chance that its still a ram issue despite it passing memtest86?


I still have the ram my friend passed me and I haven't tested it yet. VENGEANCE® RGB PRO 16GB (2 x 8GB) DDR4 DRAM 3200MHz C16 Memory Kit (cmw16gx4m2e3200c16). I have checked on the corsair site, it is compatible with it, though I don't think it was tested on asus side. If this works and doesnt crash, will this rule out motherboard as the issue? Or even if it works, motherboard can still be the issue?

 

[Darkbreeze]

Yeah, I'd try it with the other RAM first. Clearly it might simply be a problem with that memory and using different memory could show that. I'd try it and see what happens.

 

[Naturu]

It has been about 5 days since I used my friend's ram and so far there hasn't been any crashes or freezes. Going to observe 1 more week, and if its still stable, I think it might be the ram.

In the mean time, will try to pass my friend my ram and see if he crashes with it as well.

 

[johnbl]

fyi: i looked at your old dumps:
the results are consistent with memory timing errors. I see one bit corruptions in key files, and kernel address that look valid but are not.
generally, you would update the bios to get the best memory timings or slow down the memory slightly if the problem is not isolated to one memory stick.
you will find that not letting the computer sleep will reduce the number of bugchecks. full reboot will also help until you get the timing issue resolved.
(file image copied into ram, gets corrupted by one bit, the system memory manager then copies the modified driver image back to the pagefile.sys later when the modified image is used it causes a bugcheck.)

I have also seen this with bugs in chipset drivers that result in corrupted pagefile.sys. common issue when bios setting for secondary memory timings are incorrect. (like command delay set to 1 clock when It should be set to 2 to allow the electronics not violate its setup and hold time and give the memory controller time to stabilize the address lines. (takes longer when using large memory modules or more than one. )

memtest86 would generally be a good test to hammer out these issues.

you might consider a minor increase in the ram voltage.

this tool can show RAM timings CPU-Z | Softwares | CPUID



------------------------


(most current to oldest bugcheck)

bugcheck from 9/17 single bit memory corruption in ndis driver stored in memory.
win32kbase.sys wind32k.sys and win32kfull.sys modified.
-----------

bugcheck (bad kernel address called)
win32kfull.sys modified
----------
bugcheck (bad kernel address called)
one bit corruption in nt kernel
win32k.sys modified
-----------
bugcheck (bad memory address used by win32kfull)
----------
bugcheck bad memory address used by some service

--------------

bank 1 and 3 have corsair ch32gx4m2d3600c18 ddr4 16gb at 2133MHz

 

[Naturu]

At the moment I do not know which stick is the issue. Using either 1 individually works just fine. Using both is when the issues occur. After doing a clean reset of windows, I also did not get any BSOD, the screen just freezes and I have to hold the power button to shut down and boot it again.

What do u mean by sleep? Do u mean sleep mode? My computer has never been in sleep mode. I turn off the auto sleep in X minutes. And what is full reboot? Do u mean like turning off fast start up? Or do u mean just shutting it down and turning it on?

Memtest86 did not reveal any issues with the ram sticks in default settings, individually and also when paired. Though I only did 4 passes each. I have also increased ram voltage and it did nothing (I did not dare go beyond 1.39V as I unfamiliar with messing with these). For the secondary timings, when u mean incorrect, what do u mean?

 

[Darkbreeze]

You really can't even DO a full reboot if Fast startup or Hibernation are enabled. Myself, I always recommend (And do so on every system I work on) disabling hibernation on desktops anyhow, which will automatically disable Fast startup. The reason is that it just seems to be the root of too many problems in Windows which don't happen if it is disabled. If it is enabled, it isn't actually shutting down or restarting, it is simply saving the configuration and reloading it without ever actually fully cutting the power to everything. That is an oversimplification but is basically the deal.

Given all the testing you have done already on the memory, both individually and together, I really just think there is an incompatibility between that memory kit and your motherboard. They simply do not like each other. This could be something as simple as the board doesn't prefer some secondary or tertiary timing and reconfigures it to a different number which then causes problems with other memory timings or settings which creates an issue, OR that there is a setting that doesn't work well with the configuration and the motherboard DOESN'T reconfigure that setting, which has the same effect.

Or anything. I've read about extensive memory testing where some boards or CPUs simply do not like the TYPE of IC's (Memory chips) in use on a given memory module. There is a reason why memory and motherboard manufacturers DO compatibility testing and it isn't just because they like wasting time unnecessarily.

 

[johnbl]

try this, got into bios find the memory timings change the command rate timing to 2T or 2N if it is set to 1T or 1N.
boot into windows, turn off the system virtual memory, reboot and turn the system virtual memory back on. This will dump the pagefile.sys and remove any saved corruption.

see if the system is more stable.
command rate is often the 5th timing setting but it depends on the bios

 

[johnbl]

the command rate specifies the number of clock ticks are required after the RAM chip is selected to be used. This means the memory controller wants to select a chip to get data from and the 2 clock ticks are required to allow the electronics address controller to stabilize the address before the correct address will be set.

most bios will default to 1T but with larger RAM capacities or using more than one RAM stick
the circuit will act as a capacitor and slow down the timing. The timing also slows down depending on the slots that are used. (further from the cpu the slower the circuit) (often motherboard vendor will specify in the fine print to use 2T when you add more RAM)

violations of the electronics setup and hold time means the memory is being read before the memory address has been fully locked in and stabilized. This means windows might want to read from a memory address like 0xffff9a001c365180
which is equal to
1111 1111 1111 1111 1001 1010 0000 0000 0001 1100 0011 0110 0101 0001 1000 0000

this would be given to the address controller but if the address is not locked in at the correct timing in the electronics then any of the one bit could be read as a zero or a zero bit could be read as a 1 and the resulting memory address read from would be incorrect and read out memory data would be wrong. This could happen anywhere in the system. If it happens in checked data then the system will bugcheck. in unchecked data the windows memory manager will save the RAM data to disk in the pagefile.sys and the errors will build up over time until something critical gets corrupted and the system bugchecks.
(this is why you want to delete the pagefile if you have these type of issues)

anyway changing the secondary timings is a quick change that can help in this case.

edit: I should also point out the electronics timings tend to change as the electronics gets older. Your motherboard was designed sometime near 2014
so it is old.

manual also indicates that if you use more than 4GB ram you need to have better cooling (of ram, i would think)

 

[Darkbreeze]

anyway changing the secondary timings is a quick change that can help in this case.

change the command rate timing to 2T or 2N if it is set to 1T or 1N.

In some cases, this might also result in not being able to POST, so while it is still good advice as it very well might be the solution needed, be prepared for there to be a need to reset the BIOS if after doing so you find the system will no longer post OR is throwing constant errors/BSOD.

 

[Naturu]

Hi, it has been awhile again. Work kept me too busy to test things, but I finally got some time off.

I tried the secondary timing thing. In the bios, apparently the command rate timing defaults to auto, which I guess switches between 1T and 2T.

I have tried with default settings for everything aside from the command rate timing, all 3 settings, auto, 1T and 2T. All of them work on day 1 and did not freeze or BSOD, however, the next day, sometimes it will fail to boot and remain in the orange LED I talked about earlier. To boot, force shutting down and booting works sometimes, reseating the ram sticks usually works and removing 1 always works.

When it manages to boot however, I have not encountered any crash yet, which is kind of strange.

Regardless, is there any other things I could try?

Otherwise I think I might just conclude that this is probably ram issue and have to decide if I should just stick with 1 stick or buy another pair.

If I were to buy a pair of ram sticks, do u think it would be fine with 32gb (16x2) as long as it was tested by asus? I was wondering if the cause of my issues is it being 32gb, cause my friends 16gb (8x2) works fine, could even enable docp with no trouble.

 

[Darkbreeze]

It is not because of it being 32GB. Especially if that 32GB is made up of two DIMMs and not four. I'd stick to G.Skill sticks with a 3200-3600mhz speed and a CL16 or CL18 latency. Preferably G.Skill Neos, Ripjaws or Trident Z sticks. Stay away from the Corsair LPX sticks. Stay away from brands that don't validate per board model. Corsair, Crucial and G.Sill validate.