9900k. can not control VCCSA Voltage. fixed.

[K0NG]

the problem was that i got some BSODS with machine check exceptions and Memory Management errors when playing BFV.

then i saw that i had my VCCSA and VCCIO Voltage manually set to 1v which is not that much for 4x8GB 3600 Mhz Memory.

i bumped it up to 1.1v for both and everything was/is fine now.

the problem is since 2 days the VCCSA Voltage goes up after every boot.

the first time it was at 1.108v
then the next day it was 1.120v
then 1.128v
yesterday 1.14v
Now it is at 1.17v

i can not do anything against it.
when i go to the bios and pick Auto the voltage goes up after every reboot.
when i say 1V flat it is still at 1.17v and the next time a few Mlilivolts more.

i flashed the bios (there was an update for microcode) and cleared the CMOS.

it worked until and the voltage was at 1.108v

now i rebooted the PC two times and the voltage went to 1.128 and now to 1.168v again.

It's a Z390 Extreme4.

 

[K0NG]

just a minute later i rebootet again and now it is at 1.184v

[Jurrunio]

WTF asrock

 

You manually locked it to a certain number in the bios right? You read the number from there?

[K0NG]

Just now, Jurrunio said:

You manually locked it to a certain number in the bios right? You read the number from there?

it's Manual 1.11V in the Bios.

and it climbs and climbs all the time now i am at 1.21v

[K0NG]

i literally can watch it live in HWINFO64.

around 16mv per 5 Minutes.

[Mister Woof]

Gremlins

[K0NG]

now it gets creepy and dangerous.

the System Agent jumped up to 1.6V now fluctuates a couple hundred! milivolts.

[xg32]

44 minutes ago, K0NG said:

now it gets creepy and dangerous.

the System Agent jumped up to 1.6V now fluctuates a couple hundred! milivolts.

the orignal voltage settings on the asrock boards were a joke, i'd update the bios and see if that helps, 1.6v is ridonk

[Mister Woof]

i got my money on busted sensor

[kla3]

Probably reading a false value, or as someone before me said "Gremlins"

[K0NG]

well the reading seems to be correct since it is 1.5v+ in the bios too.

it was perfectly fine a few days ago.
now i am back to 1.12v at 2133 Mhz on the Ram.

as soon as i (even manually) go to the XMP specs the SA freaks out and almost kills itself within seconds after booting.

[cried22]

HWinfo reports 1.48v to my 9900ks when I really only have 1.31v. It is not accurate for all hardware setups for some reason.

[kla3]

If i were to trust my hwinfo numbers, my laptop would be boiling now

Do any parts get really hot? thats usually the first sign of to much voltage

 

My best guess is false values tbh

[K0NG]

29 minutes ago, kla3 said:

If i were to trust my hwinfo numbers, my laptop would be boiling now

Do any parts get really hot? thats usually the first sign of to much voltage

 

My best guess is false values tbh

again.
the Bios Voltage says 1.5v Too
and the CPU is at least 15°C warmer in Idle when the 1.5v kicks in.

[K0NG]

After CMOS Reset and freshly installed Bios (latest one)After 30 Minutes on Desktop. you Can see that VCCIO (Which as the identical fixed voltage as the SA in Bios) stayed normal. Only the VCCSA Voltage shoots up to oblivion and sticks there until i reset the CMOS Again. (IN BIOS The Voltage Readout is the same as in HWINFO64. it says 1.584v on VCCSA!)

[kla3]

9 hours ago, K0NG said:

the Bios Voltage says 1.5v Too

have you tried using Intel Tuning utility/throttlestop to try to set the voltage, just as an experiment to see if its a BIOS bug,

[K0NG]

1 minute ago, kla3 said:

have you tried using Intel Tuning utility/throttlestop to try to set the voltage, just as an experiment to see if its a BIOS bug,

i'll try it

[K0NG]

There is no way to change vccio/sa voltages in XTU

[kla3]

9 hours ago, K0NG said:

'll try it

Also; is your CPU set to manual OC or do you just run it with Turbo Boost?

VCCSA can go up if CPU is set to auto, and it starts to get hot, since the CPU gets less efficient when hot

[kla3]

9 hours ago, K0NG said:

There is no way to change vccio/sa voltages in XTU

Does vcore scale as the VCCSA go up?

EDIT: do you use LLC?

[K0NG]

8 minutes ago, kla3 said:

Does vcore scale as the VCCSA go up?

EDIT: do you use LLC?

my system is manually overclocked like all the other ones i had before.

i use a moderate LLC to compensate the transient dips and spikes instead of pushing 0 mOhm.

The CPU has a fixed voltage, like the ram,VCCIO/SA and Chipset.

No Vcore does not scale with it.

it is ONLY the Systemagent and ONLY since yesterday. it NEVER was a problem (since i monitor voltages and temperatures almost non stop.)

i went back to 4.20 now (4.30 is the new one released in "early 2020" so a few days ago.)

[kla3]

9 hours ago, K0NG said:

my system is manually overclocked like all the other ones i had before.

i use a moderate LLC to compensate the transient dips and spikes instead of pushing 0 mOhm.

The CPU has a fixed voltage, like the ram,VCCIO/SA and Chipset.

No Vcore does not scale with it.

it is ONLY the Systemagent and ONLY since yesterday. it NEVER was a problem (since i monitor voltages and temperatures almost non stop.)

i went back to 4.20 now (4.30 is the new one released in "early 2020" so a few days ago.)

Oh man, then im out of clues really!

Is it time to tell asrock to get their shit together? lol

[K0NG]

So. 4.20 is installed since 5 minutes now and it already went up to 1.136v (from 1.120v)

[kla3]

some fluctuation is normal, but it seems odd that it varies that much when you tell it to be locked!

Have you tried running some high load on just the CPU, to see if it triggers the voltage spike?

[K0NG]

when i run Aida64 torture test with cpu,fpu,cache and ram it stays to that voltage.


now it's at 1.142v (like i said before.. it slowly climbs until around 1.6v and only can get reversed when i clear the CMOS)