What could be causing these errors in MemTest86?

[DoranVuTye]

Hi everyone!

 

I have a system here that has been seeing some general system instability lately such as giving a Memory Management BSOD error. I've done some troubleshooting and have narrowed it down to either my motherboard (unlikely), memory sticks, or processor as MemTest86 shows errors in a test. However I don't have enough knowledge to understand which component specifically may be causing the issue. 

 

The system is,

 

Asus x570 Crosshair VIII Hero WiFi

AMD 3900x

32 GB (4x8) Team TForce Dark Pro 3200 CL14

MSI GTX 1070

 

The sticks should be the same with only very minor differences which can be seen attached.

 

I've attached the results of the memtest below and will upload another test afterwards once I am finished. 

 

I plan on turning off docp/xmp and resorting the memory sticks to see if the errors persist and weather they appear at different times which could indicate a memory stick issue I believe. 

 

However, if I am wrong and there are better suggestions. Please let me know.

 

Another situation it could be is that the sticks aren't making proper contact with the board; as in my situation the AIO that I'm using is pushing against one of my sticks of memory to the point that you can see a difference in space between the sticks. However when I installed it the sticks still managed to seat properly from my opinion (including the audible click sound from the levers). If this is the case then it would also be somewhat tested when I reseat the memory. A photo can be seen attached.

 

Unfortunately I do not have extra sticks of memory nor another processor to swap them out and check that way.

 

Best case scenario it's just the sticks which allows me to RMA them as they come with a lifetime warranty; however if it's the memory controller on the processor then that would require me to go through AMD warranty which I'm not too sure on.

[Haro]

2 minutes ago, ıllıllı ĐÖŖÄŅ VŮ ȚŸË ıllıllı said:

Another situation it could be is that the sticks aren't making proper contact with the board

It will likely not boot if that's the case. 

 

2 minutes ago, ıllıllı ĐÖŖÄŅ VŮ ȚŸË ıllıllı said:

however if it's the memory controller on the processor then that would require me to go through AMD warranty which I'm not too sure on.

Clear the cmos and get rid of xmp/docp and then try memtest86..

[LIGISTX]

14 minutes ago, ıllıllı ĐÖŖÄŅ VŮ ȚŸË ıllıllı said:

Hi everyone!

 

I have a system here that has been seeing some general system instability lately such as giving a Memory Management BSOD error. I've done some troubleshooting and have narrowed it down to either my motherboard (unlikely), memory sticks, or processor as MemTest86 shows errors in a test. However I don't have enough knowledge to understand which component specifically may be causing the issue. 

 

The system is,

 

Asus x570 Crosshair VIII Hero WiFi

AMD 3900x

32 GB (4x8) Team TForce Dark Pro 3200 CL14

MSI GTX 1070

 

The sticks should be the same with only very minor differences which can be seen attached.

 

I've attached the results of the memtest below and will upload another test afterwards once I am finished. 

 

I plan on turning off docp/xmp and resorting the memory sticks to see if the errors persist and weather they appear at different times which could indicate a memory stick issue I believe. 

 

However, if I am wrong and there are better suggestions. Please let me know.

 

Another situation it could be is that the sticks aren't making proper contact with the board; as in my situation the AIO that I'm using is pushing against one of my sticks of memory to the point that you can see a difference in space between the sticks. However when I installed it the sticks still managed to seat properly from my opinion (including the audible click sound from the levers). If this is the case then it would also be somewhat tested when I reseat the memory. A photo can be seen attached.

 

Unfortunately I do not have extra sticks of memory nor another processor to swap them out and check that way.

 

Best case scenario it's just the sticks which allows me to RMA them as they come with a lifetime warranty; however if it's the memory controller on the processor then that would require me to go through AMD warranty which I'm not too sure on.

 

 

 

 

 

Have you tried to narrow down if a single stick is causing issues? Or a single slot? I wouldn't be surprised if that stick closest to the AIO is causing issues. Would be easy enough to pop that stick out and run memtest again and see what happens... Realistically, you should try and figure out if its just a single stick or a single slot via pulling sticks out and testing just 1 or two at a time, keeping track of what you have tried.

 

If it is the AIO ports causing the issue, just rotate the AIO. It won't look quite as "correct", but thats sorta how the cookie crumbles unfortunately (assuming that turns out to be the issue.).

 

If you can narrow down which stick or DIMM slot is bad, you would be able to RMA the RAM, or the mobo no problem

[Master Disaster]

Unfortunately this is one of those cases where we are in the same boat as you and are guessing at the cause. Your assumption is exactly the same as mine would be, board, CPU or RAM, narrowing it down beyond that requires actual hardware to be able to try swapping things in and out. If I was forced to guess I would say its more likely to be RAM over the other 2 though.

 

As suggested above my first move would be to unplug that one stick and try memtest again.

 

If that doesn't work there is another thing you can try. Head into BIOS and increase the SOC Voltage on the CPU from its default of 1.1v up to 1.2v then try running memtest again and see if that clears it up.

 

If neither of these things work then you have something that you probably cannot diagnose at home. You would need to either pay a shop to test a different CPU in the board or just RMA all 3.

 

You have updated the BIOS to the latest version, correct?

[DoranVuTye]

1 hour ago, TofuHaroto said:

It will likely not boot if that's the case. 

 

Clear the cmos and get rid of xmp/docp and then try memtest86..

Ok so after disabling DOCP and leaving my ram sticks as they were; the errors seem to have cleared up. From my guess this tells me that it's a ram stick problem. Would you agree?

22 minutes ago, Master Disaster said:

Unfortunately this is one of those cases where we are in the same boat as you and are guessing at the cause. Your assumption is exactly the same as mine would be, board, CPU or RAM, narrowing it down beyond that requires actual hardware to be able to try swapping things in and out. If I was forced to guess I would say its more likely to be RAM over the other 2 though.

 

As suggested above my first move would be to unplug that one stick and try memtest again.

 

If that doesn't work there is another thing you can try. Head into BIOS and increase the SOC Voltage on the CPU from its default of 1.1v up to 1.2v then try running memtest again and see if that clears it up.

 

If neither of these things work then you have something that you probably cannot diagnose at home. You would need to either pay a shop to test a different CPU in the board or just RMA all 3.

 

You have updated the BIOS to the latest version, correct?

Yea my BIOS is updated to the latest version, and I've tested with the same setup but without DOCP as I stated to the other dude above and the errors no longer appear. I think it may be an issue with my memory sticks then as with the DOCP settings the sticks are automatically bumped up to 1.35v which leads me to believe that the sticks may be faulty since they're rated to run at the speeds.

[Haro]

4 minutes ago, ıllıllı ĐÖŖÄŅ VŮ ȚŸË ıllıllı said:

Would you agree?

Not entirely. 

Set the soc to 1.1 and dram to 1.35. After that set the frequency to 3200 with xmp/docp. Then use memtest to make sure it's stable. If it is go for 3600 with xmp/docp with the values said above and use memtest and check..

[DoranVuTye]

9 minutes ago, TofuHaroto said:

Not entirely. 

Set the soc to 1.1 and dram to 1.35. After that set the frequency to 3200 with xmp/docp. Then use memtest to make sure it's stable. If it is go for 3600 with xmp/docp with the values said above and use memtest and check..

SOC is already set to 1.1 and dram is already set to 1.35v under DOCP.

[Haro]

1 minute ago, ıllıllı ĐÖŖÄŅ VŮ ȚŸË ıllıllı said:

under DOCP.

Then try 3200 and then work your way up to 3600. 

4 sticks while they have the same bandwidth, it adds more pressure on the imc and the interconnect. You can also try manually setting the FCLK to 1600 for 3200 and 1800 for 3600. (Though that might now help). 

[Master Disaster]

18 minutes ago, ıllıllı ĐÖŖÄŅ VŮ ȚŸË ıllıllı said:

SOC is already set to 1.1 and dram is already set to 1.35v under DOCP.

Set the SOC to 1.2v, 1.35v for the RAM is correct.

 

I believe you might be suffering from a similar issue as I had a few weeks back. The SOC on the CPU is responsible for running the Infinity Fabric Clock (FClock) which is the internal bus that the RAM controller, PCIe controller and NVMe controller (since NVMe is just PCIe) are all connected to. The base FClock is 1200Mhz. When you enable DOCP it automatically OCs the FClock to the same speed as your RAM (so in your case 1600Mhz) which puts additional strain on the SOC since its now having to run the PCIe bus at 1600Mhz too. Because you actually do have a PCIe 4.0 GPU connected I believe trying to run 4 sticks of RAM and a GPU at 1600Mhz is causing instability in the SOC.

 

1.1v on the SOC is stock, anything up to 1.2v on the SOC is fine for 24/7 use but do not go beyond that for periods of more than a few hours.

 

Can I ask what your SSD is?

[DoranVuTye]

26 minutes ago, Master Disaster said:

Set the SOC to 1.2v, 1.35v for the RAM is correct.

 

I believe you might be suffering from a similar issue as I had a few weeks back. The SOC on the CPU is responsible for running the Infinity Fabric Clock (FClock) which is the internal bus that the RAM controller, PCIe controller and NVMe controller (since NVMe is just PCIe) are all connected to. The base FClock is 1200Mhz. When you enable DOCP is automatically OCs the FClock to the same speed as your RAM (so in your case 1600Mhz) which puts additional strain on the SOC since its now having to run the PCIe bus at 1600Mhz too. Because you actually do have a PCIe 4.0 device connected I believe trying to run 4 sticks of RAM and a GPU at 1600Mhz is causing instability in the SOC.

 

31 minutes ago, TofuHaroto said:

Then try 3200 and then work your way up to 3600. 

4 sticks while they have the same bandwidth, it adds more pressure on the imc and the interconnect. You can also try manually setting the FCLK to 1600 for 3200 and 1800 for 3600. (Though that might now help). 

Here's my bios settings with DOCP enabled.

 

I want my system to run with the DOCP settings at 3200 MHz however the exhibit MemTest86 errors with those settings.

 

I don't think I have a PCIE 4.0 device connected but I'll try setting my SOC voltage to 1.2v and seeing if it exhibits errors with DOCP enabled. Should I try an offset or a manual fixed voltage?

[DoranVuTye]

27 minutes ago, Master Disaster said:

Set the SOC to 1.2v, 1.35v for the RAM is correct.

 

I believe you might be suffering from a similar issue as I had a few weeks back. The SOC on the CPU is responsible for running the Infinity Fabric Clock (FClock) which is the internal bus that the RAM controller, PCIe controller and NVMe controller (since NVMe is just PCIe) are all connected to. The base FClock is 1200Mhz. When you enable DOCP it automatically OCs the FClock to the same speed as your RAM (so in your case 1600Mhz) which puts additional strain on the SOC since its now having to run the PCIe bus at 1600Mhz too. Because you actually do have a PCIe 4.0 GPU connected I believe trying to run 4 sticks of RAM and a GPU at 1600Mhz is causing instability in the SOC.

 

1.1v on the SOC is stock, anything up to 1.2v on the SOC is fine for 24/7 use but do not go beyond that for periods of more than a few hours.

 

Can I ask what your SSD is?

SSD is an OEM drive. Model is Samsung XP491 256 GB SSD.

[Master Disaster]

6 minutes ago, ıllıllı ĐÖŖÄŅ VŮ ȚŸË ıllıllı said:

 

Here's my bios settings with DOCP enabled.

 

I want my system to run with the DOCP settings at 3200 MHz however the exhibit MemTest86 errors with those settings.

 

I don't think I have a PCIE 4.0 device connected but I'll try setting my SOC voltage to 1.2v and seeing if it exhibits errors with DOCP enabled. Should I try an offset or a manual fixed voltage?

Apologies, I thought I had read you had a 5700XT connected.

 

TBH 1.08v on the SOC is a little on the low side. Try setting it to manual mode and upping it to 1.15v.

[DoranVuTye]

Just now, Master Disaster said:

Apologies, I thought I had read you had a 5700XT connected.

 

TBH 1.08v on the SOC is a little on the low side. Try setting it to manual mode and upping it to 1.15v.

Will try that now.

 

The board auto defaults to 1.08 so no idea why it's that low.

[Master Disaster]

Just now, ıllıllı ĐÖŖÄŅ VŮ ȚŸË ıllıllı said:

Will try that now.

 

The board auto defaults to 1.08 so no idea why it's that low.

Auto voltages can be weird sometimes, SOC should be 1.1v for stock but maybe the board has determined your CPUs SOC can handle running slightly below spec?

 

Its still possible we're miles off the mark and the issue isn't the SOC at all tbh.

[DoranVuTye]

Here's a screenshot of the tests with DOCP disabled showing that it runs with no errors.

[Haro]

10 hours ago, ıllıllı ĐÖŖÄŅ VŮ ȚŸË ıllıllı said:

so no idea why it's that low.

try 1.15 as said above. 

also try 2933 and work your way up. one thing you can do is try 2 sticks at the same time, as mix and matching ram can cause problems..

[DoranVuTye]

Just now, Master Disaster said:

Auto voltages can be weird sometimes, SOC should be 1.1v for stock but maybe the board has determined your CPUs SOC can handle running slightly below spec?

 

Its still possible we're miles off the mark and the issue isn't the SOC at all tbh.

True, I do run my processor with a undervolt vcore offset of -.0750 if that has anything to do with it.

[Master Disaster]

1 minute ago, ıllıllı ĐÖŖÄŅ VŮ ȚŸË ıllıllı said:

True, I do run my processor with a undervolt vcore offset of -.0750 if that has anything to do with it.

It shouldn't, I also run my 3800X with a -0.025v offset.

[DoranVuTye]

Running the test now it seems that the errors are still occuring and more often at 1.15 SOC voltage.

 

 

[Master Disaster]

Just now, ıllıllı ĐÖŖÄŅ VŮ ȚŸË ıllıllı said:

Running the test now it seems that the errors are still occuring and more often at 1.15 SOC voltage.

 

 

Then it seems as though @TofuHaroto is probably correct. Try pulling 2 of the sticks out and running just a matched pair in slots 2 and 4. There's absolutely zero reason running a dual channel matched kit should cause any issues, if it does you've got an issue somewhere else and at that point it has to be the CPU since the board isn't really doing anything except providing the bus between the CPU & RAM, the IMC is on the CPU.

[Master Disaster]

Just to be totally thorough, if you still get errors with only 2 sticks it would be wise to also test the other 2 on their own.

[DoranVuTye]

19 minutes ago, Master Disaster said:

Just to be totally thorough, if you still get errors with only 2 sticks it would be wise to also test the other 2 on their own.

 

30 minutes ago, Master Disaster said:

Then it seems as though @TofuHaroto is probably correct. Try pulling 2 of the sticks out and running just a matched pair in slots 2 and 4. There's absolutely zero reason running a dual channel matched kit should cause any issues, if it does you've got an issue somewhere else and at that point it has to be the CPU since the board isn't really doing anything except providing the bus between the CPU & RAM, the IMC is on the CPU.

Doing that right now, will test one pair and then the other afterwards. Will update after!

[DoranVuTye]

45 minutes ago, Master Disaster said:

Just to be totally thorough, if you still get errors with only 2 sticks it would be wise to also test the other 2 on their own.

 

1 hour ago, TofuHaroto said:

try 1.15 as said above. 

also try 2933 and work your way up. one thing you can do is try 2 sticks at the same time, as mix and matching ram can cause problems..

The first and oldest pair show errors in slots 2 and 4. 

 

I'll try the other pair now.

[Haro]

1 minute ago, ıllıllı ĐÖŖÄŅ VŮ ȚŸË ıllıllı said:

I'll try the other pair now.

Check if that one has errors as well. 

An extra step you can take us downclock the frequency a bit and then work your way up to it. 

If non of the pairs work..then it could be the board or the cpu. Though I doubt two kits will suffer the exact problem. 

[DoranVuTye]

10 minutes ago, TofuHaroto said:

Check if that one has errors as well. 

An extra step you can take us downclock the frequency a bit and then work your way up to it. 

If non of the pairs work..then it could be the board or the cpu. Though I doubt two kits will suffer the exact problem. 

Man, things just turned for the worse. I swapped out the pairs for the newer sticks and now the system won't even post at all. I'm getting debug code 0d and an orange debug light under dram.