BSOD and Memory Errors

[Obsessionman]

The past few days I've suddenly been getting quite a few BSOD. My PC is a few years old now and have been stable to this point.

The WhoCrashed analysis of the dmp files all pointed to potential memory errors. I ran Memtest86 for about 4 hours this afternoon and it failed with 4 errors all in Test 8.

I haven't changed anything on the hardware end in quite some time and have used the same A-XMP profile 1 the whole time with no issues before.

I've attached the MemTest86 report and the dmp files from all the crashes.

 

[JWNoctis]

PTW, and does it go away if you disable A-XMP? It sounds to me that increased voltage on A-XMP and consequently increased wear - see recent Intel RMA debacle for something similar but much faster - might be catching up to the CPU's memory controller, or the memory modules themselves. In my (n=1) experience, A-XMP profiles by default overvolts both quite aggressively, and probably unnecessarily.

Others would know more than I do.

 

[Obsessionman]

I'll run another test with the profiles disabled tonight.

Another point of info is my PSU is getting pretty old now ~7-8 years so I don't know if this could have an affect.

 

[Laurijan]

just post a pic of the memtest result after 1 full run. zip attachments are not save.

 

[ir_cow]

I see your system specs on the phone. What are they?

 

[Chomiq]

I see your system specs on the phone. What are they?

Doesn't have them listed.

 

[Obsessionman]

System Specs:

Ryzen 5 5600x
MSI B550-A PRO
RTX 3070 FE
EVGA SuperNOVA 750 B2
4 x G.Skill F4-3600C18-8GVK

Crucial MX500 1TB
Crucial MX500 500GB
WD Blue SN550 1TB NVMe
WD Black SN770 1TB NVMe

LG 32UN650-W
Optix MAG274QRF

Razer Viper Mouse
HyperX Alloy Origins Keyboard

M-Audio M-Track Solo – USB Audio Interface

I'll run another test with the profiles disabled tonight.

Another point of info is my PSU is getting pretty old now ~7-8 years so I don't know if this could have an affect.

The test I ran with A-XMP disabled was completed successfully. It seems to be running at 2133 MHz and 1.2V

just post a pic of the memtest result after 1 full run. zip attachments are not save.

 

[A Computer Guy]

Post a ZenTimings screenshot so we can get an idea of the voltages being used.

 

[Obsessionman]

Post a ZenTimings screenshot so we can get an idea of the voltages being used.

 

[A Computer Guy]

View attachment 348568

Include one when you have XMP/A-XMP enabled.

 

[Obsessionman]

Include one when you have XMP/A-XMP enabled.

Was just going to do that sorry.

 

[Chomiq]

Was just going to do that sorry.

View attachment 348571

VDDP is way too high should be around 0.95V max, VDDG is ok'ish. Compare tRC with the actual XMP profile (use CPU-Z or Thaiphoon burner).

 

[Obsessionman]

VDDP is way too high should be around 0.95V max, VDDG is ok'ish. Compare tRC with the actual XMP profile (use CPU-Z or Thaiphoon burner).

Is this what you're talking about in CPU-Z?

 

[A Computer Guy]

Was just going to do that sorry.

View attachment 348571

When troubleshooting bumping VSOC to 1.1 can help. Try that and retest.

 

[Obsessionman]

When troubleshooting bumping VSOC to 1.1 can help. Try that and retest.

I don't have a lot of experience with manually changing settings. I've been using the Default A-XMP 1 Profile for the past few years because I thought it would be stable for the advertised speeds.

Am I just looking for a BIOS setting called VSOC and setting it to 1.1

 

[mstenholm]

Since it's Samsung b-die it should be safe to bump VDIMM a bit, start with 0.02 V and re-test.

 

[Obsessionman]

Should I be concerned that this is happening so suddenly? I've been stable for so long then 5 crashes in a short period out of nowhere seems troubling

 

[Chomiq]

Is this what you're talking about in CPU-Z?

View attachment 348572

Yeah exactly, notice that tRC is 64 in the XMP profile but your board set it to 85.

Since it's Samsung b-die it should be safe to bump VDIMM a bit, start with 0.02 V and re-test.

What makes you think this is Samsung B-die?

 

[mstenholm]

Yeah exactly, notice that tRC is 64 in the XMP profile but your board set it to 85.


What makes you think this is Samsung B-die?

I might have seen it on Redit

 

[Obsessionman]

Yeah exactly, notice that tRC is 64 in the XMP profile but your board set it to 85.


What makes you think this is Samsung B-die?

Is this a board issue then or just a misconfiguration. I've not familiar with what tRC is I've never done any manual memory tuning.

I noticed in HWMonitor that DRAM has the voltage listed at 1.344V is that a significant different from 1.35V

Also I have 4 sticks bought at different times. I have used them all together with no issue using A-XMP for about 19 months. 2 are listed as Samsung in CPU-Z and 2 are listed as SK Hynix. It was only recent BSOD that when googled suggested I may have memory issues.

• UNEXPECTED_KERNEL_MODE_TRAP (7f)
• KERNEL_SECURITY_CHECK_FAILURE (139)
• CRITICAL_STRUCTURE_CORRUPTION (109)
• PAGE_FAULT_IN_NONPAGED_AREA (50)
• ATTEMPTED_EXECUTE_OF_NOEXECUTE_MEMORY (fc)

 

[A Computer Guy]

I don't have a lot of experience with manually changing settings. I've been using the Default A-XMP 1 Profile for the past few years because I thought it would be stable for the advertised speeds.

Note the CPU spec is officially DDR4-3200. Most Zen3 CPU's can do DDR4-3600 as far as I know but that is a RAM overclock for your CPU so there is no guarantee.
One thing you can do is set the XMP/A-XMP then manually downclock to DDR4-3200 to see if that fixes your problem as well.

Am I just looking for a BIOS setting called VSOC and setting it to 1.1

There may be some differences between BIOS and what they labeled it. I'll have to look at mine to see what it's called. Perhaps if someone else here has an MSI UEFI/BIOS they can chime in to the name of it.

( edit )

Looking at some pictures online it looks MSI labels this as "CPU NB/SoC Voltage". If your UEFI/BIOS also shows the current voltage value this should be about the same as what is currently reported in ZenTimings.

Is this a board issue then or just a misconfiguration. I've not familiar with what tRC is I've never done any manual memory tuning.

I also noticed in HWMonitor that DRAM has the voltage listed at 1.344V is that a significant different from 1.35V

The voltage you input might not be the exact voltage you get. This occurs frequently.

 

[Obsessionman]

Note the CPU spec is officially DDR4-3200. Most Zen3 CPU's can do DDR4-3600 as far as I know but that is an overclock so there is no guarantee.
One thing you can do is set the XMP/A-XMP then manually downclock to DDR4-3200 to see if that fixes your problem as well.

There may be some differences between BIOS and what they labeled it. I'll have to look at mine to see what it's called. Perhaps if someone else here has an MSI UEFI/BIOS they can chime in to the name of it.

( edit )

Looking at some pictures online it looks to be called "CPU NB/SoC Voltage". If your UEFI/BIOS also shows the current voltage value this should be about the same as what is currently reported in ZenTimings.

View attachment 348611


The voltage you input might not be the exact voltage you get. This occurs frequently.

Okay, I found this and it was set to Auto like the picture and value was 1.105

 

[A Computer Guy]

Okay, I found this and it was set to Auto like the picture and value was 1.105

Sorry I just re-read one of your recent posts and missed before about you buying memory at different times. It seems you have likely mixed kits of different IC's (chips) which isn't the best configuration for compatibility. They likely have different settings for their timings. If you want to use them you will need to get the settings from each type of kit installed and probably set timings manually using the worst timings between each kit to try and get them compatible while also matching the kits in the same channel. For example the Samsung modules in A2+B2 while the SK Hynix in A1+B1. However if you purchased all the dimms individually at different times it's possible you might have even different versions of samsung and sk hynix making compatibility more complicated.

Ideally I would sell your two kits and get a new 32GB (2 dimm) kit instead for the best out of the box compatibility.

 

[Obsessionman]

Sorry I just re-read one of your recent posts and missed before about you buying memory at different times. It seems you have likely mixed kits of different IC's (chips) which isn't the best configuration for compatibility. They likely have different settings for their timings. If you want to use them you will need to get the settings from each type of kit installed and probably set timings manually using the worst timings between each kit to try and get them compatible while also matching the kits in the same channel. For example the Samsung modules in A2+B2 while the SK Hynix in A1+B1. However if you purchased all the dimms individually at different times it's possible you might have even different versions of samsung and sk hynix making compatibility more complicated.

Ideally I would sell your two kits and get a new 32GB (2 dimm) kit instead for the best out of the box compatibility.

If I've used this configuration without issue for almost 2 years though could my blue screen issue being coming from something else then? Or just possibly one of the sets is dying

If the timings stables are shown the same is the because that's what the motherboard is setting them to?

 

[Super Firm Tofu]

Good grief. I'm not sure what's going on in this thread, but the OP has been running this setup for years stable. The recent onset of BSOD's, and the confirmation of a failed memory test, all points to one of the sticks of RAM going bad.

@Obsessionman, you can test out the sticks by removing one set at a time and running the memtest again. Make sure you're leaving each set of ram in the 2nd and 4th slots (from the cpu socket). Test one matched set, and then remove and try the others in the same slots.

If both sets test out fine by themselves it's possible that it could be a memory slot on the board, but I'd put my money on one bad stick of RAM.