CPU undervolt, crashing games, tests run fine

[Springpaw]

Hey everyone, new to undervolting here. Following this guide

I tried undervolting my 7700x because it makes me uncomfortable to see cpu core temps hitting 95c. I ran through cinibench without failure and overall system stability is good with win11 crashing once in a blue moon (once every two weeks or so). The issue I’m having is that weirdly only games are crashing. I work in tech so I’ve done a fair bit of troubleshooting to diagnose the issue but results seem to remain the same. Drivers are up to date. Would love to know if anyone is encountering similar issues. Games I’ve been encountering issues with are Elden Ring, Star Citizen, and Darktide. Thanks for the input in advance and happy New Year.

Specs:
TUF x670e
32gb 6000mhz ddr5
7700x
FTW 3080

 

[3x0]

Download this tool https://github.com/sp00n/corecycler/releases/tag/v0.9.3.0 and let it run overnight to see if the system crashes

 

[Springpaw]

Download this tool https://github.com/sp00n/corecycler/releases/tag/v0.9.3.0 and let it run overnight to see if the system crashes

Thanks I’ll give it a shot

 

[27MaD]

Well if it's crashing once every 2 weeks or so and crashing more often while gaming than your undervolt is likely to be unstable, Pump the core voltage a little bit and see if that fixes things up, Stress testing alone is usually not enough to test the stability of your system.

 

[Springpaw]

Download this tool https://github.com/sp00n/corecycler/releases/tag/v0.9.3.0 and let it run overnight to see if the system crashes

While running the test everything seems to be failing, I'm not sure how to read this test.

 

[3x0]

It means you need to decrease the Curve Optimizer values or vCore offset. It's very unstable if it can't pass even a single cycle/iteration

 

[Springpaw]

It means you need to decrease the Curve Optimizer values or vCore offset. It's very unstable if it can't pass even a single cycle/iteration

I set the CO down to 0 and it’s still failing every test with similar results

 

[3x0]

Hm, it's possible Prime95 is running with an improper config. Open config.ini file in notepad and set stressTestProgram = YCRUNCHER
Try again, it should work now. I don't have Zen 4 so I don't know what's the best config to test.

 

[Springpaw]

Hm, it's possible Prime95 is running with an improper config. Open config.ini file in notepad and set stressTestProgram = YCRUNCHER
Try again, it should work now. I don't have Zen 4 so I don't know what's the best config to test.

Test is running smoothly now at -30 really wonder wtf is going on lol. I’ll let it run for the weekend and report back. Extremely appreciate the help.

 

[HD64G]

If I was in your shoes I would just put a thermal target limit in the UEFI and that would be the end of your adventure. It will undervolt by itself that way. 85C is great to allow 99% of the performance without having problems with noise and heat.

 

[GerKNG]

I set the CO down to 0 and it’s still failing every test with similar results

Interesting...
my 7950X had exactly the same issue.

 

[Springpaw]

Interesting...
my 7950X had exactly the same issue.

Try changing the test to YCRUNCHER in default like mentioned above. Test is still running and passing every iteration.

 

[GerKNG]

Try changing the test to YCRUNCHER in default like mentioned above. Test is still running and passing every iteration.

because it's relatively light load. ended up returning the CPU and bought a 13700k. (having to deal with this stuff again isn't worth it imo.)
you might tweak your system now for days or weeks, wait for bioses etc. the whole AM5 Platform is currently broken as hell.
my thread was closed...

well the 7950X didn't last long...

so... i got my 7950X around a week ago and had often BSODs and general instability across the board. (Games, Low loads, idle) i thought my Curve Optimizer was a bit too aggressive. dropped it from -25 to -20, then to -15 and then to -10. still crashes. verified memory stability over night...

www.techpowerup.com

 

[3x0]

@Springpaw When you finish the light load testing with CoreCycler, you can switch to testing heavy all core load.
1. http://numberworld.org/y-cruncher/#Download Start the program, type 1-7-0 to start testing all
2. https://www.techpowerup.com/download/linpack-xtreme/ Start the program, 2 (stress test), 5 (10GB RAM), run at least 20 cycles with all threads enabled.

 

[Springpaw]

because it's relatively light load.
my thread was closed...

well the 7950X didn't last long...

so... i got my 7950X around a week ago and had often BSODs and general instability across the board. (Games, Low loads, idle) i thought my Curve Optimizer was a bit too aggressive. dropped it from -25 to -20, then to -15 and then to -10. still crashes. verified memory stability over night...

www.techpowerup.com

Hmm, I’m not really having an issue with bsod, maybe you got bad silicon? I’ve been aggressively researching this issue, you could try looking into the this thread and the one linked in it. https://www.reddit.com/r/AMDHelp/comments/zubjet

 

[GerKNG]

well BSODs or not... the Chip is unstable out of the box (you have errors across almost all cores.)
the BSODs will come earlier or later when the crash in a game or application is heavy enough.

 

[Springpaw]

If I was in your shoes I would just put a thermal target limit in the UEFI and that would be the end of your adventure. It will undervolt by itself that way. 85C is great to allow 99% of the performance without having problems with noise and heat.

That’s fair, I’m just determined to solve this issue. Only for that it’s an uncomfortably high temp to run the chip at and it being relatively new tech worries me on the it’s lifespan.

well BSODs or not... the Chip is unstable out of the box (you have errors across almost all cores.)
the BSODs will come earlier or later when the crash in a game or application is heavy enough.

Hopefully not but the game instability is why I’m here. I’m just trying to isolate the issue exactly, I’m just not entirely sold if this is tied to the undervolt or if I’ve got another issue. It’s just very strange this is isolated only to games and specifically multithreaded games it seems.

 

[GerKNG]

That’s fair, I’m just determined to solve this issue. Only for that it’s an uncomfortably high temp to run the chip at and it being relatively new tech worries me on the it’s lifespan.


Hopefully not but the game instability is why I’m here. I’m just trying to isolate the issue exactly, I’m just not entirely sold if this is tied to the undervolt or if I’ve got another issue. It’s just very strange this is isolated only to games and specifically multithreaded games it seems.

well that's the thing.
"Light multithreaded work" like modern games are pushing the highest possible boost clocks out of the chip and make it unstable.
as long as your ram is stable it's another dud like mine and many others.

 

[lsevald]

I managed to fix my system somehow (7950x, Asus B650E-E). Prime95 AVX2 loads would crash my PC instant reboot/BSOD or just error out in P95 (even at BIOS defaults, and much worse if I touched PBO and CO). It would also reboot randomly while gaming, or just idling in Win11 and launching Chrome. I have tried all the BIOS files for my board back and forth, little helped, until yesterday. Sadly I have tried so many things now that I'm not entirely sure what I did to fix it. But things started to improve a lot after I did this (if memory serves me right):

1. Flashed fresh BIOS (latest non beta 821)
2. Loaded optimized defaults and configured my RAM, but like usual, the system was still AVX2 unstable at this point (RAM tests fine)
4. Went into Curve Optimizer->All Core, and set it to something like +5 save and reboot
5. Went into Curve Optimizer->All Core again, and set it to the default Positive and 0, save and reboot
6. Did the same for Curve Optimizer->Per Core (Set all to +5, Save&Reboot, then set all back to default Positive 0, Save and Reboot)
7. Disabled PBO (default is Auto), also disabled Asus "Medium Boost It" (default is Auto) and Save&Reboot (this alone didn't fix anything before)
8. System was fairly AVX2 stable at this point (two cores threw an error after 2 hours or so)
9. Applied Vcore Offset -0.0100 in BIOS as I know my CPU likes it a bit lower

Started Core Cycler with P95 Smallest FFT AVX2 with no hyperthreading for max boost, left for work, and came back to:

---------------------
------ Summary ------
---------------------
The script ran for 06 hours, 08 minutes, 44 seconds
No core has thrown an error

After that I have played games that used to crash a lot for 4-5hours, not a hitch. So now I'm at least stable at stock, and the system scores ~2050/38500ish in CB23, which seems normal. So what's going on here? Are some settings left behind in Curve Optimizer and applied when not supposed to? Have anyone of you used the Ryzen Master-> Curve Optimizer -> Start Optimizing feature? I had some wierd stuff going on after trying that. My system refused to BIOS post, and I had to clear CMOS to get it back up and running again. Does Ryzen Master mess with BIOS stuff? I don't know, maybe I will try to break it again to see if I can replicate the issue

 

[Springpaw]

I managed to fix my system somehow (7950x, Asus B650E-E). Prime95 AVX2 loads would crash my PC instant reboot/BSOD or just error out in P95 (even at BIOS defaults, and much worse if I touched PBO and CO). It would also reboot randomly while gaming, or just idling in Win11 and launching Chrome. I have tried all the BIOS files for my board back and forth, little helped, until yesterday. Sadly I have tried so many things now that I'm not entirely sure what I did to fix it. But things started to improve a lot after I did this (if memory serves me right):

1. Flashed fresh BIOS (latest non beta 821)
2. Loaded optimized defaults and configured my RAM, but like usual, the system was still AVX2 unstable at this point (RAM tests fine)
4. Went into Curve Optimizer->All Core, and set it to something like +5 save and reboot
5. Went into Curve Optimizer->All Core again, and set it to the default Positive and 0, save and reboot
6. Did the same for Curve Optimizer->Per Core (Set all to +5, Save&Reboot, then set all back to default Positive 0, Save and Reboot)
7. Disabled PBO (default is Auto), also disabled Asus "Medium Boost It" (default is Auto) and Save&Reboot (this alone didn't fix anything before)
8. System was fairly AVX2 stable at this point (two cores threw an error after 2 hours or so)
9. Applied Vcore Offset -0.0100 in BIOS as I know my CPU likes it a bit lower

Started Core Cycler with P95 Smallest FFT AVX2 with no hyperthreading for max boost, left for work, and came back to:

---------------------
------ Summary ------
---------------------
The script ran for 06 hours, 08 minutes, 44 seconds
No core has thrown an error

After that I have played games that used to crash a lot for 4-5hours, not a hitch. So now I'm at least stable at stock, and the system scores ~2050/38500ish in CB23, which seems normal. So what's going on here? Are some settings left behind in Curve Optimizer and applied when not supposed to? Have anyone of you used the Ryzen Master-> Curve Optimizer -> Start Optimizing feature? I had some wierd stuff going on after trying that. My system refused to BIOS post, and I had to clear CMOS to get it back up and running again. Does Ryzen Master mess with BIOS stuff? I don't know, maybe I will try to break it again to see if I can replicate the issue

What temps are you getting on average?

 

[lsevald]

At BIOS defaults it will go up to 95°C (P95 and other heavy multi threaded loads), but for normal gaming and desktop use, it's probably running 40-75°C most of the time. I can probably set a 80°C limiter in BIOS and I won't notice it performance wise.

 

[Mussels]

You've just undervolted too far, simple as that

Ryzen CPU's when you run the curve optimiser too low, will crash in low-load states, usually when transitioning from low to high

AVX loads have the highest power draw, so they'll show it up the fastest

Ryzen master applies settings to the BIOS, you'll need to reset your BIOS to get things back to defaults

 

[Springpaw]

You've just undervolted too far, simple as that

Ryzen CPU's when you run the curve optimiser too low, will crash in low-load states, usually when transitioning from low to high

AVX loads have the highest power draw, so they'll show it up the fastest

Ryzen master applies settings to the BIOS, you'll need to reset your BIOS to get things back to defaults

By reset do you mean reflash or just reset to default? And I wouldn’t be able to fix the issue by by raising the curve optimizer?

At BIOS defaults it will go up to 95°C (P95 and other heavy multi threaded loads), but for normal gaming and desktop use, it's probably running 40-75°C most of the time. I can probably set a 80°C limiter in BIOS and I won't notice it performance wise.

For normal gaming without the undervolt I was hitting 95c gaming

 

[Mussels]

Load optimised defaults in the BIOS, or by removing the battery which does the same if the system wont post

You likely can fix it with changing the curve optimiser, but of course - only if the curve optimiser is the problem

For normal gaming without the undervolt I was hitting 95c gaming

That just plainly says your cooling is no good

 

[Springpaw]

Load optimised defaults in the BIOS, or by removing the battery which does the same if the system wont post

You likely can fix it with changing the curve optimiser, but of course - only if the curve optimiser is the problem

That just plainly says your cooling is no good

I’m using a a noctua NH-D15 with two 2 front case fans, two top case and one rear exhaust in a meshify-C. I also worked in hardware repair for 3 years so I’m confident my thermal application is nice and tidy.