Computer crashes playing games and RAM issues.

[bahamut204]

I'm going to apologise in advance. This is going to be long to be able to describe the situation so far. Better to get as much out as I can remember now before people start telling me to try things that I've already tried.

 

After building multiple custom PCs over the years, both for myself and others. I am experiencing my first nightmare build. After a month of testing, waiting for replacement parts, retesting, waiting, then retesting again. I am ready to admit I need help and a lot of it.

 

Specs:
Windows 10 Pro 64bit
AMD Ryzen 5 5600X
MSI MPG X570 Gaming Plus (Latest non beta bios and latest AMD chipset drivers)
Gigabyte Aorus RTX 3070
4x Patriot Viper Steel 32GB DDR4 3200 (PC4 25600) CL16 (more later)
Corsair RMX Series RM750X 80 Plus Gold
ASUS (PCE-AC56) Dual-Band Wireless AC1300 PCIE Adaptor
500 GB Samsung 980 Pro PCIE 4.0 NVME (OS installed here)
1TB WD_Black SN750 PCIE 3.0 NVME
2x 4TB Samsung 870 QVO SSD
2x 4TB Western Digital HDD
Fractal Meshify 2 Case

 

For the RAM there are 2 different model numbers:
4 sticks from Newegg, sold in packs of 2: PVS464G320C6K
3 sticks from Best Buy, sold in packs of 1: PVS432G320C6
Same specs and name. Using CPU-Z it shows the stick from Newegg were manufactured by Samsung and the sticks from Best Buy manufactured by SPECTEK.
I will be referring to them as Samsung and Spectek ram respectively to save writing the model numbers over and over again.
Since I've had it the longest and it's passed the most testing. I'll also refer to the Samsung RAM as being known good sticks. I don't have access to other RAM to test.

 

Games tested: Far Cry 5, Escape From Tarkov, COD: Warzone, Evil Genius 2, Outriders (All have the same issues.)

The computer has not blue screened through any of this so there are no crash dumps to report.

 

Situation:
First when building the PC. It would not POST with all 4 sticks. Upon testing individually it was found that the system would not POST with 2 of them. Regardless of slot or number of times reseated.
I RMA'D the 2 faulty sticks and used the system for about a week without a noticeable incident. (I had no reason to check event viewer or CPUZ at the time. So I am unaware if the following issues were happening in the background during this time.)
I recieved a refund from Newegg after about 10 days on the basis that they weren't selling that product anymore. (I am aware that they are now selling it again at a higher price.)
I then bought the sticks from Best Buy as they were the same brand, specs, and the only place I could get replacements.
I'm not sure about the next part as 2 things happened at nearly the same time and before I could game again.
I installed the Spectek RAM and a Windows Update happened shorty after turning the PC on.
Since then every game crashes within 20mins. Either a straight crash to desktop with no error, a crash with event viewer showing the faulting module as the dx11 or dx12 dll file, or the screen turns black and event viewer logs that dwm.exe crashed.
These issues were fixed with a reinstall of Windows. But a new set of errors started immediately afterwards.
Now all games crash between 20-60mins of play time. Either crashing to desktop with no error or the computer freezes completely and parts of the screen go black while the color inverts on the rest of the screen.
Have to do a hard shutdown in that case. When the computer freezes there are 2 netBT errors reported in event viewer.
The issues are only happening while playing games. The freezing and crashing are not happening otherwise. Also ran Heaven, Cinebench, and 3DMark for multiple passes in a row. No issues.
I have exhausted every possible software cause that I could think of or find on the internet already.
Started testing with MemTest86. Stopped the test partway through the first pass once it passed 300 errors with the 4 sticks in there.
Testing each stick in multiple slots showed that one of the Spectek sticks had an intermittent falure every 2-3 passes.
RMA'D the stick and waited for replacement. Ran the remaining sticks in slots 1,2,4. Forcing single channel usage.
Tested the games again for fun. Was surprised to find no errors. Other than losing some performance I was able to game for 3h+ multiple times without issue.
Got the new Spectek stick (now my 7th RAM stick total) and started memtest immediately with all 4 sticks installed. They failed on the 2nd or 3 passes that were done.
Booted the computer to check CPUZ to confirm that the stick was a Spectek stick and noticed 2 things odd. First, while the Spectek sticks were installed in slots 1,3 and Samsung in 2,4. CPUZ sometimes reports the Spectek in slots 1,2 and Samsung in 3,4. Other times it reports in correctly. Checking event viewer there was the following errors on startup. Previous shutdown was unexpected, System rebooted without cleanly shutting down first, Audit events have been dropped by transport, Windows failed fast startup, and 2 netBT errors. The system was booting from a clean shutdown.
I removed the new stick and ran with system with the remaining 3 sticks in slots 2,3,4. Again single channel and again a day of gaming and a night of memtest with no errors.
Moving to test dual channel specifically with the 2 'good' sticks. Slots 2 and 4 generated no errors with memtest after 4 passes. Slots 1 and 3 had intermittent errors again.
With both configs I booted into Windows next and all of the previous errors mentioned where there.
I went to test the newest stick on its own and the computer will not POST with just that stick in there. Regardless of slot or number of times reseated.
I have not mentioned this yet. But during the entire process of testing different configs of RAM. There was about a 50% chance that the computer would actually POST. The rest of the time it either sat on the RAM indicator or power cycled 3 times then sat on the RAM indicator. It would sometimes take multiple reseats to get it to POST. Even though I knew it was seated properly.
It was also randomly failing to load the XMP profile and kicking me into the bios. Sometimes a reboot would get it to show 3200mhz. Most times it shows 2666mhz. CPUZ also shows on most boots that it is running at 2666mhz even with the xmp proflie enabled.
The timings have never been what was advertised. It's supposed to be 16-18-18-36 and it's 16-20-20-40.

 

One last odd thing that I noticed. Depending on the slots and sticks. The RAM's bandwitdh varied in memtest. 21816-22359 MB/s in single channel and 36971-38729 MB/s in dual channel. Is this normal?

 

That's about all I can remember for now.

[Applefreak]

26 minutes ago, bahamut204 said:

I'm going to apologise in advance. This is going to be long to be able to describe the situation so far. Better to get as much out as I can remember now before people start telling me to try things that I've already tried.

 

After building multiple custom PCs over the years, both for myself and others. I am experiencing my first nightmare build. After a month of testing, waiting for replacement parts, retesting, waiting, then retesting again. I am ready to admit I need help and a lot of it.

 

Specs:
Windows 10 Pro 64bit
AMD Ryzen 5 5600X
MSI MPG X570 Gaming Plus (Latest non beta bios and latest AMD chipset drivers)
Gigabyte Aorus RTX 3070
4x Patriot Viper Steel 32GB DDR4 3200 (PC4 25600) CL16 (more later)
Corsair RMX Series RM750X 80 Plus Gold
ASUS (PCE-AC56) Dual-Band Wireless AC1300 PCIE Adaptor
500 GB Samsung 980 Pro PCIE 4.0 NVME (OS installed here)
1TB WD_Black SN750 PCIE 3.0 NVME
2x 4TB Samsung 870 QVO SSD
2x 4TB Western Digital HDD
Fractal Meshify 2 Case

 

For the RAM there are 2 different model numbers:
4 sticks from Newegg, sold in packs of 2: PVS464G320C6K
3 sticks from Best Buy, sold in packs of 1: PVS432G320C6
Same specs and name. Using CPU-Z it shows the stick from Newegg were manufactured by Samsung and the sticks from Best Buy manufactured by SPECTEK.
I will be referring to them as Samsung and Spectek ram respectively to save writing the model numbers over and over again.
Since I've had it the longest and it's passed the most testing. I'll also refer to the Samsung RAM as being known good sticks. I don't have access to other RAM to test.

 

Games tested: Far Cry 5, Escape From Tarkov, COD: Warzone, Evil Genius 2, Outriders (All have the same issues.)

The computer has not blue screened through any of this so there are no crash dumps to report.

 

Situation:
First when building the PC. It would not POST with all 4 sticks. Upon testing individually it was found that the system would not POST with 2 of them. Regardless of slot or number of times reseated.
I RMA'D the 2 faulty sticks and used the system for about a week without a noticeable incident. (I had no reason to check event viewer or CPUZ at the time. So I am unaware if the following issues were happening in the background during this time.)
I recieved a refund from Newegg after about 10 days on the basis that they weren't selling that product anymore. (I am aware that they are now selling it again at a higher price.)
I then bought the sticks from Best Buy as they were the same brand, specs, and the only place I could get replacements.
I'm not sure about the next part as 2 things happened at nearly the same time and before I could game again.
I installed the Spectek RAM and a Windows Update happened shorty after turning the PC on.
Since then every game crashes within 20mins. Either a straight crash to desktop with no error, a crash with event viewer showing the faulting module as the dx11 or dx12 dll file, or the screen turns black and event viewer logs that dwm.exe crashed.
These issues were fixed with a reinstall of Windows. But a new set of errors started immediately afterwards.
Now all games crash between 20-60mins of play time. Either crashing to desktop with no error or the computer freezes completely and parts of the screen go black while the color inverts on the rest of the screen.
Have to do a hard shutdown in that case. When the computer freezes there are 2 netBT errors reported in event viewer.
The issues are only happening while playing games. The freezing and crashing are not happening otherwise. Also ran Heaven, Cinebench, and 3DMark for multiple passes in a row. No issues.
I have exhausted every possible software cause that I could think of or find on the internet already.
Started testing with MemTest86. Stopped the test partway through the first pass once it passed 300 errors with the 4 sticks in there.
Testing each stick in multiple slots showed that one of the Spectek sticks had an intermittent falure every 2-3 passes.
RMA'D the stick and waited for replacement. Ran the remaining sticks in slots 1,2,4. Forcing single channel usage.
Tested the games again for fun. Was surprised to find no errors. Other than losing some performance I was able to game for 3h+ multiple times without issue.
Got the new Spectek stick (now my 7th RAM stick total) and started memtest immediately with all 4 sticks installed. They failed on the 2nd or 3 passes that were done.
Booted the computer to check CPUZ to confirm that the stick was a Spectek stick and noticed 2 things odd. First, while the Spectek sticks were installed in slots 1,3 and Samsung in 2,4. CPUZ sometimes reports the Spectek in slots 1,2 and Samsung in 3,4. Other times it reports in correctly. Checking event viewer there was the following errors on startup. Previous shutdown was unexpected, System rebooted without cleanly shutting down first, Audit events have been dropped by transport, Windows failed fast startup, and 2 netBT errors. The system was booting from a clean shutdown.
I removed the new stick and ran with system with the remaining 3 sticks in slots 2,3,4. Again single channel and again a day of gaming and a night of memtest with no errors.
Moving to test dual channel specifically with the 2 'good' sticks. Slots 2 and 4 generated no errors with memtest after 4 passes. Slots 1 and 3 had intermittent errors again.
With both configs I booted into Windows next and all of the previous errors mentioned where there.
I went to test the newest stick on its own and the computer will not POST with just that stick in there. Regardless of slot or number of times reseated.
I have not mentioned this yet. But during the entire process of testing different configs of RAM. There was about a 50% chance that the computer would actually POST. The rest of the time it either sat on the RAM indicator or power cycled 3 times then sat on the RAM indicator. It would sometimes take multiple reseats to get it to POST. Even though I knew it was seated properly.
It was also randomly failing to load the XMP profile and kicking me into the bios. Sometimes a reboot would get it to show 3200mhz. Most times it shows 2666mhz. CPUZ also shows on most boots that it is running at 2666mhz even with the xmp proflie enabled.
The timings have never been what was advertised. It's supposed to be 16-18-18-36 and it's 16-20-20-40.

 

One last odd thing that I noticed. Depending on the slots and sticks. The RAM's bandwitdh varied in memtest. 21816-22359 MB/s in single channel and 36971-38729 MB/s in dual channel. Is this normal?

 

That's about all I can remember for now.

Wait, you are using 32 GB sticks, right? Those are hit and miss most of the time. Very few boards run fine with that much memory in a single slot. Unless the board maker lists those modules as supported, don't expect them to work. You may have to adjust timings and voltages by hand to get them to boot properly.  

[bahamut204]

MSI's spec page lists 128GB max supported RAM. Which is 32GB/stick and the speed is on the list as well. I do not know where to find a list of specific models of RAM that are supported if there is one.

 

I do not know about the Spectek sticks that I've returned. But I did get an email from Newegg stating that they confirmed that the 2 Samsung sticks I sent them were faulty before they issued my refund. I do not know if they really did or not.

 

Could there be something wrong with the CPU?

I'm beginning to wonder if it's a motherboard issue possibly in addition to faulty RAM. Most of the actual crashing happens in dual channel mode. Is it possible there is something wrong with how the motherboard is handling dual channel mode?

[Heliian]

1 hour ago, bahamut204 said:

MSI's spec page lists 128GB max supported RAM. Which is 32GB/stick and the speed is on the list as well. I do not know where to find a list of specific models of RAM that are supported if there is one.

 

I do not know about the Spectek sticks that I've returned. But I did get an email from Newegg stating that they confirmed that the 2 Samsung sticks I sent them were faulty before they issued my refund. I do not know if they really did or not.

 

Could there be something wrong with the CPU?

I'm beginning to wonder if it's a motherboard issue possibly in addition to faulty RAM. Most of the actual crashing happens in dual channel mode. Is it possible there is something wrong with how the motherboard is handling dual channel mode?

@Applefreak is on to it.

 

The ram you're trying to use isn't on their compatibility list.  https://www.msi.com/Motherboard/support/MPG-X570-GAMING-PLUS#support-mem-21