[SOLVED] Computer crashes an hour into gaming but only for DX12

GIGgle_Hertz

Distinguished
Nov 3, 2012
16
0
18,520
Got an interesting issue or at least I think so. Hopefully this is the appropriate thread, let me know if not. Have been troubleshooting for a month, have made progress but now feel like I’m chasing a red herring maybe.

So I had built a fairly new PC, Used some Parts from my previous build (like my m.2 drive, hard drive and Power supply)

CPU: AMD 5800x
Motherboard: Asrock B550 Steel legend
Memory: G.skill Trident Z Neo 2x16GB 3600 (F4-3600C18D-32GTZN)
GPU: EVGA 2060 XC 6GB
Powersupply: EVGA supernova 850 G2 Gold

Everything on the build is at stock settings Except for the memory, I have enabled the XMP profile for 3600 speed.

When I first built this, like many others, I did not have a new GPU for it yet. So I was using my previous GPU, the EVGA 2060 XC 6GB.
Everything is working great. I passed the 3Dmark testing and 8 hour test(only have dx11) and CPU benchmarks. I can play games for hours with no issue.
Finally, I win (not really win) a Newegg shuffle and I get an ASUS 3070 8GB v2 Gaming KO. I installed my new card and again everything seems great. I pass 3Dmark and pass testing at 8 hours. When I started to game though, things seemed fine at first.

I first noticed when playing Battlefield 2042, typically within an hour of playing, my PC would crash and reboot. But then after that crash, I could resume playing for multiple hours. The crash can sometimes be a bluescreen reboot, other times just a black screen reboot.
I also noticed that certain games would not trigger the crash and could be played for hours. Like Hell let loose, or total war: Warhammer 2. That’s when I started marking what games crash and which don’t. Through my testing I determined that only DX12 games were crashing within an hour of playing. While Dx11 games could be played for hours without issue.

I began troubleshooting my issue. I was leaning on my Card as it was the most recent change and none of the crashes would occur with the 2060 on the same games.
  1. I reinstalled my Graphics driver- issue occurs
  2. Reseated my card – issue occurs
  3. Made sure to use 2 separate 8pin cables- issue occurs
  4. I clean uninstalled all my drivers and re-installed them. - issue occurs (this included a newer AMD chipset driver than wasn’t listed on my Motherboard site)
  5. I disabled PBO in my motherboard – issue occurs
  6. Finally I disabled XMP on my memory –no crashes occur it seems.
So now I’m thinking it’s the memory.
  1. Set my memory at 3600 with out XMP – issue occurs
  2. Update bios to the latest and tried XMP again – issue occurs
  3. Set memory at 3600 and 1.35v – issue occurs
  4. Set XMP and raise voltage to 1.36v – SO FAR, has not had the issue but haven’t be able to test a lot yet to say for sure. Will update once I know more.

I also ran a full mem test on my memory with XMP enabled for 3600 just to see if it would fail or crash. successful.

So this is where I am currently at, seeing if the issue will re-occur. But as I’m sitting here, I’m really starting to doubt the progress I’ve made. None of this occurred with the 2060, Only the new 3070. Should I be going to ASUS support? It is the obvious change in my system before issues start occurring. And only Dx12 Games are affected so is it even more of a GPU problem?
Is 3600 on my memory Bin die just a pipe dream? My memory was not on my motherboard Qualified list but WAS on G.Skill’s Qualified Motherboard list. Researching other peoples builds, they do seem to hit 3600.
Is it a power draw issue?

I guess I am just looking to confirm my path is true or if I am off course or any other suggestions.
 
Solution
I swear i wouldn't let this die out in the great ether. I forgot to come back in the fury of troubleshooting since the last update.

UPDATE:
I really jinxed myself last time.

  1. The replacement card arrived and the issue STILL OCCURRED. - Moral was low at this point. Logically it made since as other cards ran with out issue but alas the problem continued.
  2. I then re-seated everything - still issue.
  3. tried new 850 gold PSU - still issue.
  4. Got asus b450 board and built a secondary PC from spare parts(1700x, the new PSU, Same memory and the 3070) - no issue!!!!!
  5. Took the 5800x out of the original PC and put it in the secondary PC. - THE ISSUE RETURNS!
  6. i swap back to the 1700x on the b450 and the issue goes away again...
dx12 is more hardware demanding or in another words, its utilising hardware better then dx11 does...

if gpu itself isnt faulty, id start with testing on another PSU, just make sure you dont mix PSU cables from different PSU units, even if they fit, they arent compatible with eachother

you should also run DDU in safe mode and try reinstalling gpu drivers

for bsod, you can upload minidumps from c:\windows\minidumps into some file hosting web service (google drive/onedrive/etc) then post link here

3600mhz on ryzen might need extra voltage on infinity fabric, post screenshot from zentimings to see how your mainboard configured voltages
 
You probably did this but just in case. When installing the new RTX 3070, did you use DDU to do a proper clean install? Also in BIOS load the optimised defaults and re-engage XMP along with rebar....but it does seem like a driver issue if all is good on DX11 and not DX12 though reading around this there are a fair few issues with DX12.
 

Zerk2012

Titan
Ambassador
Got an interesting issue or at least I think so. Hopefully this is the appropriate thread, let me know if not. Have been troubleshooting for a month, have made progress but now feel like I’m chasing a red herring maybe.

So I had built a fairly new PC, Used some Parts from my previous build (like my m.2 drive, hard drive and Power supply)

CPU: AMD 5800x
Motherboard: Asrock B550 Steel legend
Memory: G.skill Trident Z Neo 2x16GB 3600 (F4-3600C18D-32GTZN)
GPU: EVGA 2060 XC 6GB
Powersupply: EVGA supernova 850 G2 Gold

Everything on the build is at stock settings Except for the memory, I have enabled the XMP profile for 3600 speed.

When I first built this, like many others, I did not have a new GPU for it yet. So I was using my previous GPU, the EVGA 2060 XC 6GB.
Everything is working great. I passed the 3Dmark testing and 8 hour test(only have dx11) and CPU benchmarks. I can play games for hours with no issue.
Finally, I win (not really win) a Newegg shuffle and I get an ASUS 3070 8GB v2 Gaming KO. I installed my new card and again everything seems great. I pass 3Dmark and pass testing at 8 hours. When I started to game though, things seemed fine at first.

I first noticed when playing Battlefield 2042, typically within an hour of playing, my PC would crash and reboot. But then after that crash, I could resume playing for multiple hours. The crash can sometimes be a bluescreen reboot, other times just a black screen reboot.
I also noticed that certain games would not trigger the crash and could be played for hours. Like Hell let loose, or total war: Warhammer 2. That’s when I started marking what games crash and which don’t. Through my testing I determined that only DX12 games were crashing within an hour of playing. While Dx11 games could be played for hours without issue.

I began troubleshooting my issue. I was leaning on my Card as it was the most recent change and none of the crashes would occur with the 2060 on the same games.
  1. I reinstalled my Graphics driver- issue occurs
  2. Reseated my card – issue occurs
  3. Made sure to use 2 separate 8pin cables- issue occurs
  4. I clean uninstalled all my drivers and re-installed them. - issue occurs (this included a newer AMD chipset driver than wasn’t listed on my Motherboard site)
  5. I disabled PBO in my motherboard – issue occurs
  6. Finally I disabled XMP on my memory –no crashes occur it seems.
So now I’m thinking it’s the memory.
  1. Set my memory at 3600 with out XMP – issue occurs
  2. Update bios to the latest and tried XMP again – issue occurs
  3. Set memory at 3600 and 1.35v – issue occurs
  4. Set XMP and raise voltage to 1.36v – SO FAR, has not had the issue but haven’t be able to test a lot yet to say for sure. Will update once I know more.
I also ran a full mem test on my memory with XMP enabled for 3600 just to see if it would fail or crash. successful.

So this is where I am currently at, seeing if the issue will re-occur. But as I’m sitting here, I’m really starting to doubt the progress I’ve made. None of this occurred with the 2060, Only the new 3070. Should I be going to ASUS support? It is the obvious change in my system before issues start occurring. And only Dx12 Games are affected so is it even more of a GPU problem?
Is 3600 on my memory Bin die just a pipe dream? My memory was not on my motherboard Qualified list but WAS on G.Skill’s Qualified Motherboard list. Researching other peoples builds, they do seem to hit 3600.
Is it a power draw issue?

I guess I am just looking to confirm my path is true or if I am off course or any other suggestions.
So now I’m thinking it’s the memory.
  1. Set my memory at 3600 with out XMP – issue occurs
  2. Update bios to the latest and tried XMP again – issue occurs
  3. Set memory at 3600 and 1.35v – issue occurs
  4. Set XMP and raise voltage to 1.36v – SO FAR, has not had the issue but haven’t be able to test a lot yet to say for sure. Will update once I know more.
I also ran a full mem test on my memory with XMP enabled for 3600 just to see if it would fail or crash. successful.

When you ran memtest was it with the voltage at 1.36?
Did you do a fresh install of windows when you upgraded hardware?
If it's not crashing at 1.36 I would raise it to 1.38 and just use it.
 

GIGgle_Hertz

Distinguished
Nov 3, 2012
16
0
18,520
UPDATE: i got a crash playing chivalry 2 @ the 1.36v. It seems my memory got reset to default this time around, XMP was disabled, which was not happening before. Its having me doubt myself though and what i had set. So i am going to run memory at default (no XMP) just to confirm its not crashing with stock. Sorry to muddy the waters.
BUT if it does crash at stock, that may simplify the troubleshooting some, if not a power issue.

dx12 is more hardware demanding or in another words, its utilising hardware better then dx11 does...

if gpu itself isnt faulty, id start with testing on another PSU, just make sure you dont mix PSU cables from different PSU units, even if they fit, they arent compatible with eachother

you should also run DDU in safe mode and try reinstalling gpu drivers

for bsod, you can upload minidumps from c:\windows\minidumps into some file hosting web service (google drive/onedrive/etc) then post link here

3600mhz on ryzen might need extra voltage on infinity fabric, post screenshot from zentimings to see how your mainboard configured voltages

So thats the thing, the 2060 could do the Dx12 games fine with XMP at 3600. Battlefield and chivalry 2 ran fine till the 3070 was put in.
Ill have to double check if i have a spare PSU, otherwise thankfully they are chearper than other parts.
I did not do DDU, the 2060 and 3070 technically used the same driver package, but I will perform one after my default test/verification.
I will also look into posting any of the blue screens i got. If i recall when i dug into them originally, they were all for amdsys processes. But the black screens crashes/reboots would make me think power.
Ill also look into zentiming.

---------------------------------------------------------------------------------------------------
You probably did this but just in case. When installing the new RTX 3070, did you use DDU to do a proper clean install? Also in BIOS load the optimised defaults and re-engage XMP along with rebar....but it does seem like a driver issue if all is good on DX11 and not DX12 though reading around this there are a fair few issues with DX12.

I did not, it seemed more related to if i ever had an AMD card in it. when i originally went to download the driver, i saw it was the same driver as the 2060. But i will make that my next step for sure after the next memory verify/test.
Ill also looking into the optimised defualts. After i updated the BIOs, i did do a default reload, but not sure if it was optimised default. Not sure im familar with rebar? unless im missing an acrynom.

-----------------------------------------------------------------------------------------------------------------
So now I’m thinking it’s the memory.
  1. Set my memory at 3600 with out XMP – issue occurs
  2. Update bios to the latest and tried XMP again – issue occurs
  3. Set memory at 3600 and 1.35v – issue occurs
  4. Set XMP and raise voltage to 1.36v – SO FAR, has not had the issue but haven’t be able to test a lot yet to say for sure. Will update once I know more.
I also ran a full mem test on my memory with XMP enabled for 3600 just to see if it would fail or crash. successful.

When you ran memtest was it with the voltage at 1.36?
Did you do a fresh install of windows when you upgraded hardware?
If it's not crashing at 1.36 I would raise it to 1.38 and just use it.

So when i ran the Memtest, it was just for the XMP enabled. No voltage adjustment, just what the profile does.
I did do a fresh install of windows when i upgraded my board, CPU, memory etc. Not after the 3070
I will look at trying that voltage after verifying the default no XMP is infact working with out issue.
 
It would be interesting to see how it runs once you use DDU for a clean nvidia driver re-install.. Even if it was using an RTX 2060, the drivers can get muddy and DDU does an exceptional job at cleaning and helped me once in the past.

Rebar is just the newish Re-sizable bar feature now available on Nvidia GPU's giving full access to memory, helps sometimes giving a boost in games but I doubt it has anything to do with the issue but good to have on.

I would also for the time being run the memory tuning it manually to DDR4 3600 18-22-22-42 and making sure the voltage is at 1.35v. This memory is supposed to be good and fully tested on the Ryzen CPU's but have to admit, I haven't used AMD since the Athlon days going way, way back so I am a bit weak in this area.

So now I’m thinking it’s the memory.
  1. Set my memory at 3600 with out XMP – issue occurs
  2. Update bios to the latest and tried XMP again – issue occurs
  3. Set memory at 3600 and 1.35v – issue occurs
  4. Set XMP and raise voltage to 1.36v – SO FAR, has not had the issue but haven’t be able to test a lot yet to say for sure. Will update once I know more.
I also ran a full mem test on my memory with XMP enabled for 3600 just to see if it would fail or crash. successful.

When you ran memtest was it with the voltage at 1.36?
Did you do a fresh install of windows when you upgraded hardware?
If it's not crashing at 1.36 I would raise it to 1.38 and just use it.
 

GIGgle_Hertz

Distinguished
Nov 3, 2012
16
0
18,520
Update: Confirmed i get no crashes with the default 2133. Which confirms the crash before was at 1.36v with xmp set to 3600

I booted into safe mode and ran DDU, then installed the latest drivers in safe mode. Rebooted, set defaults, set only XMP to 3600. saved and exited
Booted, played battledfield and crashed like before.
It was a blue screen this time. The mini dump is here if i got drop box setup right.
https://www.dropbox.com/s/e23y7sqtg9ldbf3/050622-7000-01.dmp?dl=0

Gonna load default and set XMP to 3600, then grabbing a zentimings shot.
 

GIGgle_Hertz

Distinguished
Nov 3, 2012
16
0
18,520
I tried Voltage at 1.380v and still got the crash.

Decided to try 3200 speed and see if thats even possible
I tried xmp with speed set at 3200, got the crash

Currently just have defaults loaded and 3200 set for the memory speed and will see.

If that fails, may try to reseat the memory as i cant re call if i already did that. Pretty sure i did but that was in the beginning of it all.
If nothing then, i may try re-seating the CPU as well.

If still issue after that, i think i just need to bite the bullet and start a support call with one of the vendors. I think i will start with ASUS for the GPU first, Logically it makes sense to me as that was the last change that occurred before the issue started.
 

GIGgle_Hertz

Distinguished
Nov 3, 2012
16
0
18,520
Think this is turning into a dead thread judging from the lack of comments other than myself. But I will continue so it is logged in the great ether, hope this will help another in search of an answer.

Update: manually set memory speed at 3200 failed with the crash.

I got tired of trying and loaded defaults (memory at 2133) to be able to play a game. I recently loaded Evil genius 2 (a directx12 game), upon playing that game for 30mins I got the same crash.
This was honestly a relief, as it kinda solidifies that something is wrong with the GPU.

I started a case with ASUS and provided all of my troubleshooting info, they said to RMA it to them for repair/replacement. Upon removing the 3070 for mailing, i put in my old RX 480 8GB card (i had sold the 2060) and tested the same games with memory XMP set to 3600.
No crashes way past an hour of gaming.
So both the evga 2060, and the RX 480 had no crashes for the same games. Even further confirming, that i think the issue is GPU related.

The only other factor i could consider, is if my PSU is not delivering the wattage its suppose too. As the 3070 does draw more power than the other 2 cards. But my PSU should be more than enough for a 3070, recommended is 650w and i have 850w.

Will keep you posted on the results from Asus RMA. Hoping to have the conclusion soon.
 
Think this is turning into a dead thread judging from the lack of comments other than myself. But I will continue so it is logged in the great ether, hope this will help another in search of an answer.

Update: manually set memory speed at 3200 failed with the crash.

I got tired of trying and loaded defaults (memory at 2133) to be able to play a game. I recently loaded Evil genius 2 (a directx12 game), upon playing that game for 30mins I got the same crash.
This was honestly a relief, as it kinda solidifies that something is wrong with the GPU.

I started a case with ASUS and provided all of my troubleshooting info, they said to RMA it to them for repair/replacement. Upon removing the 3070 for mailing, i put in my old RX 480 8GB card (i had sold the 2060) and tested the same games with memory XMP set to 3600.
No crashes way past an hour of gaming.
So both the evga 2060, and the RX 480 had no crashes for the same games. Even further confirming, that i think the issue is GPU related.

The only other factor i could consider, is if my PSU is not delivering the wattage its suppose too. As the 3070 does draw more power than the other 2 cards. But my PSU should be more than enough for a 3070, recommended is 650w and i have 850w.

Will keep you posted on the results from Asus RMA. Hoping to have the conclusion soon.
Have you tried 1.35v?
 

GIGgle_Hertz

Distinguished
Nov 3, 2012
16
0
18,520
UPDATE!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

Asus determined the card could not be repaired and have sent me a replacement card. Will have to test once i receive it but i should be good.
Its a relief to see my troubleshooting steps come to fruition after such a long and methodical process.
It was quite the doozy of an issue to trouble shoot, a first for me in the 20 years of building computers.

I have no specifcs of the exact issue of the card for any curious. Support just said they determined it was a hardware issue, they got the same results i did in there testing. Ill post if they provide any other details.
Ill circle back around once i confirm the replacement card.
 

GIGgle_Hertz

Distinguished
Nov 3, 2012
16
0
18,520
I swear i wouldn't let this die out in the great ether. I forgot to come back in the fury of troubleshooting since the last update.

UPDATE:
I really jinxed myself last time.

  1. The replacement card arrived and the issue STILL OCCURRED. - Moral was low at this point. Logically it made since as other cards ran with out issue but alas the problem continued.
  2. I then re-seated everything - still issue.
  3. tried new 850 gold PSU - still issue.
  4. Got asus b450 board and built a secondary PC from spare parts(1700x, the new PSU, Same memory and the 3070) - no issue!!!!!
  5. Took the 5800x out of the original PC and put it in the secondary PC. - THE ISSUE RETURNS!
  6. i swap back to the 1700x on the b450 and the issue goes away again.
  7. I started my RMA process and got the replacement CPU in a week. NO ISSUE ON THE REPLACEMENT CPU!
In all my years of troubleshooting, this one has been a real pain in the Arse. I had a bad processor that ran fine when paired with a rx 480 or 2060 but when paired with a 3070 would crash playing games. Same processor passed multiple cinebench tests and 3dmark tests.
A real interesting one for sure, my only guess is the combo of higher power draw of the 3070 and the use of directx12 revealed the processors flaw.

Hopefully this help someone out there.
 
Solution
different cpu - different memory controller
your voltages are much lower then on mine ryzen at 3600MHz ram, but on the other hand, mine ryzen doesnt even POST at 3600 with stock soc voltage (3533max)
here you can see mine voltages, mem controller on zen2/zen3 is basicly same
BwDQpSp.png



its not really cpu flaw as cpu is advertised for 3200 ram clock, you got cpu with better memory controller now which needs less voltages to run stable which fits your mainboard soc/if settings (vsoc/vddg)
 

GIGgle_Hertz

Distinguished
Nov 3, 2012
16
0
18,520
Late follow up haha, no worries though.

To be clear, The old 5800x would crash with memory at stock 2133hz.
So not sure if what you say fully applies. It could be the old 5800x did have a faulty memory controller.