[SOLVED] Ryzen 7 1700x Having Kernel_Power 41 issues

[Mr4Derop]

So recently (since february 14th), i've been receiving some temperature problems.(I've had kernel_power 41 crashes in total of 32.)
My ryzen 7 1700x has been crashing at ~75C, which i think is really not good.
All of that crashing was on 3.78ghz on 1.35V.
I have raised the voltage to 1.4 and it crashed at ~82 or 83 on cinebench multi-core.
Can someone please tell me what i can do without raising voltage?
I'm apparently going to try to lower it to 3.4 or 3.6 GHZ.
But the stock turbo boost is UP to 3.8 ghz.
Full specs:
Ryzen 7 1700x 8-core 3.77 GHZ
AMD Radeon RX 580 GIGABYTE 8GB
32 Gigabytes of 2666mhz HyperX Fury RAM
Some zalman 600W power supply? (zalman wattbit somethingsomething)
Asus PRIME B450M-A
I can specify more specs if you want me to.

 

[drea.drechsler]

A 1700X shouId be good for 3.9-4.0Ghz so I have to think it should be able to hold a mild overclock of 3.8Ghz at 1.35V and definitely at 1.4V. Does it stay stable when at stock settings? I'd do a cmos reset to make sure you're at full default settings to determine that.

How are you determining voltage though? Don't just set it in the BIOS and expect that's what the CPU is actually getting. Use HWInfo64 and look at the CPU Core Voltage (SVI2 TFN) sensor. That the actual core voltage the CPU's seeing and it may vary considerably from no-load to heavy load. In particular it may be dropping under 1.25V when running Cinebench even though VCore is set to 1.35V in BIOS. For a 1st gen SVI2/TFN can safely go as high as 1.425V under 'light loaded' condition where temp is <70C, but try to stay under 1.375V under extreme heavy load with temps no higher than 85-90C.

If you have access to a better/newer PSU you might also try it out. That Zalman unit sounds a bit old and they weren't the best anyway; an aged sub-standard PSU can leave a lot of ripple on the +12V line that your motherboard VRM, already at it's limits with an 8 core CPU, just can't handle.

 

[punkncat]

What are you cooling this CPU with? Stock cooler?

What case is the system in?

From what I recall about my own experiences with 1st gen Ryzen, you shouldn't run them on manual settings where voltage would be over 1.4. My own experience attempting to OC a 1700 was best result 3.8 all core @1.35V. It wouldn't stay very stable at that, I turned it down to 3.7 all core @1.3125V which ended up being good long term. I feel that if I had better cooling than a fairly inexpensive tower cooler the results could have been higher. I saw issues with temps being much above 70C causing shutdown as well.

 

[Mr4Derop]

A 1700X shouId be good for 3.9-4.0Ghz so I have to think it should be able to hold a mild overclock of 3.8Ghz at 1.35V and definitely at 1.4V. Does it stay stable when at stock settings? I'd do a cmos reset to make sure you're at full default settings to determine that.

How are you determining voltage though? Don't just set it in the BIOS and expect that's what the CPU is actually getting. Use HWInfo64 and look at the CPU Core Voltage (SVI2 TFN) sensor. That the actual core voltage the CPU's seeing and it may vary considerably from no-load to heavy load. In particular it may be dropping under 1.25V when running Cinebench even though VCore is set to 1.35V in BIOS. For a 1st gen SVI2/TFN can safely go as high as 1.425V under 'light loaded' condition where temp is <70C, but try to stay under 1.375V under extreme heavy load with temps no higher than 85-90C.

If you have access to a better/newer PSU you might also try it out. That Zalman unit sounds a bit old and they weren't the best anyway; an aged sub-standard PSU can leave a lot of ripple on the +12V line that your motherboard VRM, already at it's limits with an 8 core CPU, just can't handle.

That is the problem, it doesn't even give 1.35V.
The cpu only received 1.281V in cinebench, and i'm assuming it's the PSU.
However i do have access to a 200W PSU which i can connect to my CPU instead and test it, and see if it does do something but i'm really unsure.
It runs fine on 3.4 GHZ clock, which is the base clock, but i do need more power for gaming.

 

[Mr4Derop]

What are you cooling this CPU with? Stock cooler?

What case is the system in?

From what I recall about my own experiences with 1st gen Ryzen, you shouldn't run them on manual settings where voltage would be over 1.4. My own experience attempting to OC a 1700 was best result 3.8 all core @1.35V. It wouldn't stay very stable at that, I turned it down to 3.7 all core @1.3125V which ended up being good long term. I feel that if I had better cooling than a fairly inexpensive tower cooler the results could have been higher. I saw issues with temps being much above 70C causing shutdown as well.

Case: Zalman Z1 or something, i am running full-time open side of the case to increase performance since it's awfully filled with dust and i have no time to clean it.
Cooler is a Deep Cool no name, no idea what the model is.

 

[drea.drechsler]

That is the problem, it doesn't even give 1.35V.
...

But what are you using to measure CPU core voltage? You can't just set a voltage in BIOS and assume that the CPU is getting that while under load.

Even good AM4 motherboards can have considerable Vdroop, and yours is a low end model so you can expect a lot of it. Normally it's not an issue since Ryzen's boost algorithm actually expects it to a degree. But I had major problems with that when dialing in a fixed overclock of my 1700. I could set it at 1.375V in BIOS and it would merrily crash at a 3.7G OC under heavy stress test load because Vcore was dropping below 1.25V at the CPU cores and I didn't know it.

That was really early on with Ryzen when there weren't many AM4-aware utilities and not a lot of information available. But once I found out about HWInfo64 and what the SVI2 TFN voltage was reporting I was able to get a stable overclock at 3.9Ghz while staying within voltage and temp guidelines.

One other thing to keep in mind though. Even though my CPU was within temp limit (I had it under a 240mm AIO) with about 1.35-1.37V (SVI2) at 3.9G in Prime95 the VRM was running at 100-105C on an MSI B450 motherboard. Your low-end Asus Prime model board probably doesn't expose a temp sensor for the VRM FET's. You may not realize they are toasting away at 120C or so, maybe even worse since it's so dusty inside your case.

Too-high FET temps could also be your problem: while FET's Tjmax are usually around 125C the VRM's voltage stability will be very bad. So a big load variation (like Cinebench and Prime95 often present) can lead to a big voltage spike or sag...and crash.

 

[punkncat]

awfully filled with dust

-and-

Cooler is a Deep Cool no name

Well, probably found the bigger part of your problem then...

 

[Mr4Derop]

But what are you using to measure CPU core voltage? You can't just set a voltage in BIOS and assume that the CPU is getting that while under load.

Even good AM4 motherboards can have considerable Vdroop, and yours is a low end model so you can expect a lot of it. Normally it's not an issue since Ryzen's boost algorithm actually expects it to a degree. But I had major problems with that when dialing in a fixed overclock of my 1700. I could set it at 1.375V in BIOS and it would merrily crash at a 3.7G OC under heavy stress test load because Vcore was dropping below 1.25V at the CPU cores and I didn't know it.

That was really early on with Ryzen when there weren't many AM4-aware utilities and not a lot of information available. But once I found out about HWInfo64 and what the SVI2 TFN voltage was reporting I was able to get a stable overclock at 3.9Ghz while staying within voltage and temp guidelines.

One other thing to keep in mind though. Even though my CPU was within temp limit (I had it under a 240mm AIO) with about 1.35-1.37V (SVI2) at 3.9G in Prime95 the VRM was running at 100-105C on an MSI B450 motherboard. Your low-end Asus Prime model board probably doesn't expose a temp sensor for the VRM FET's. You may not realize they are toasting away at 120C or so, maybe even worse since it's so dusty inside your case.

Too-high FET temps could also be your problem: while FET's Tjmax are usually around 125C the VRM's voltage stability will be very bad. So a big load variation (like Cinebench and Prime95 often present) can lead to a big voltage spike or sag...and crash.

HWInfo64 is what i used, so probably it might be toasting away.
I will try another PSU when i can so that probably could fix it since right now i only have 100-200W PSU's laying around.

 

[drea.drechsler]

...
I will try another PSU when i can so that probably could fix it since right now i only have 100-200W PSU's laying around.

You can get the case dusted out now to see if that helps. If you have one, put a fan to blow on the area of the VRM to cool that off too. If your crashes go away you'll know you're on to something there.