Idle CoreParking vCore // 1700x+X370TaiChi

I can add more info here.
I came home in the afternoon, booted cold and the system would not park.

Looking at HWiNFO:

Core #X VID (where X is 0-7 for the 8 cores) is at 0.400v to 1.500v. They are lively and jumping around a lot, despite an idle desktop even after waiting for stuff to settle.

vCore SV12 TFN is around 1.28v but spends a fair time at 1.35v and does go to around 1.45v. Rarely drops below the 1.28, and when it does it's only down to 1.18v anyway.

VR OUT from the IR35204 chip is very similar to SV12 TFN vCore.

vCore (from the motherboard) drops to 0.400v for very brief periods but mainly sits 0.9v or above. We all know this is not vCore, but something like the requested voltage from the motherboard or some jazz, I forget. It's not a number to care about unless it's 1.5v or something crazy.

I tried a sleep cycle, just setting Windows to sleep, then bringing it back up. No change. Let Windows settle after waking, and it's still stuck in higher voltage, responsive no-parking mode.

Reboot....

After a reboot and letting Windows settle.

Core # VID: Mostly at 0.400v but occasionally jumping up to 1.500v or 1.475v on any particular core at max. Mostly activity is seen on any particular core going to 0.894v or 0.900v.
Behaviour seems odd if most cores are parked, why is voltage fluctuation over 8 cores? I think OS tools are BS-ing me!

vCore SV12 TFN sits mostly at 0.794v-0.800v but jumps up to 1.281v and 1.400v max.
VR Out follow similar trend.


Under both conditions temperature jumps momentarily to around 36°C but rapidly falls back to 28°C as the baseline temperature.
Noctua NH-U14s cooler is set to have the fan running at 30% as minimum speed, so plenty of cooling power is available, and easily keeps tiny spikes cool.

At a loss as to why this happens.
I guess, in a perfect world, I'd love to see software tell me exactly what is going on where and how, but it seems a lot of the info on how the chip operates is not really public domain.

I have previously experimented witch disabling 4 cores completely and running a 4c/8t CPU, but it saved no watts at the wall, none that I could measure enough to say there was a difference. So I am happy to park the cores that I don't need when browsing, idle on desktop, checking emails etc. but still have the rest kick in when doing something more strenuous like A/V transcoding.

@cx5

Hi thanks for your suggestions, but I have no desire to overclock my system at present. I intend for this system to last about 10 years, as that is my usual upgrade cycle.

Chasing fps for games is not something I care about, and have actively limited the GPU to 60fps (monitor cannot display any faster) and around 75% power.

Stability and longevity is what I wish. Everything is plenty fast enough.


As for the BIOS version. I stopped at P3.10.
It offered the fan control I needed over previous versions and saw no reason to update, in fact I saw reasons not to go with higher versions.


An update to the original post.
I did have the 100MHz clock set to Auto in BIOS. I set it to manual and 100MHz, and so far I haven't seen the cores unpark. This is a first cold boot this morning. More time will tell, as it seems to randomly occur but only on the first cold boot.

cx5 wrote: Last night I flash v5.67 (the last of the mohicans = last beta unofficial pcie-4 for x370 taichi) and ram cannot run 2933mhz using previously successful timing. Tonight will try to find new timing. New calculator timing also not working, sad sad.

Clarification, in case this info could help someone.

Two nights ago I used the wrong timings hence 2933 couldn't work. Last night realized my mistake, hence quickly revert to timings in my signature and uefi 5.67 works immediately.

HOORAY, pcie-4 beta v5.67 uefi = the last of the mohicans, now lets wait for this price to come down [URL=https://www.tomshardware.com/news/phison-pcie-4.0-ssd-ps5018-e18-e13t-e16,40131.html]7GB/s pcie-4 PHISON E18 coming soon[/URL]

@VUMeter, My ram default 1.2v jedec runs 2133, as well as 2400 and 2666. The timings are default within the jedec library/rom thing.

they even up the vddcr_soc by themselves, for 2666mhz soc is 0.875v

Runs like a charm. Always works regardless of uefi version.

In the past didn't manage to make it work because I choose XMP first (which has incompatible timing with cpu) then change speed to 2666, that's why always fail.

But if I don't choose xmp at all, just choose 2666mhz directly, it will always boot successfully.

Overclock without voltage up, good prospect to you???

I installed ProcessLasso and aggressively shoved all programs onto the first core and it's two threads 0 and 1.
Whilst there was very little activity jumping beyond those two threads, the others would not park.

Checking out Power Profiles, it is almost as if "Ryzen Balanced" is permanently engaged on cold boot. The current seems higher and vCore doesn't drop much.
If I reboot and use Ryzen Balanced profile, the behaviour is fairly similar to that of a cold boot with ANY profile.

The fact that voltage and current seem to be higher, this can't be a good thing? The on die temperature seems about the same though, so perhaps the numbers are a bit of BS. Voltage + Current = Power. Power = Heat. But there isn't any real heat, not enough for me to really declare it obvious - maybe the cooler is just too efficient for that.

Rebooting on cold start really cannot be good for this machine!

Linux requires installing so that I can add ntc6775 to the /etc/modules file. This needs to be done and the system rebooted. Cannot be done with a 'live' Linux USB stick as it wont stay on reboot.

So much for that.
What it looks like is the CPU being stuck in P0 state.
All cores were at 3500MHz and a 35x multiplier, not coming down!
Voltage around the 1.2v which is what we get when all cores are hit with Prime95.
Voltage jumps for the odd single core boosting to 3900MHz.

At this point, I think I need a CMOS reset and set up things neatly again. I think something has gotten borked somewhere.

I am half tempted to try the ZenStates program, but as it's an unoffical tool and has been fiddled with so it works on other platforms than Asus motherboards, I fear damage.

CMOS reset and I'll report back.