Ryzen P-States info & ASRock Mainboard updates

Wenn it comes to LLC levels on the Taichi board.
I personally only concider level 4 or 5 to be really usable.
From what i have seen level 1, 2 and 3 allowed way too much over volting,
which shouldnt be nessesary.

Since the Gaming K4 basiclly uses the same pwm.
i suppose that you might see similar behavings with LLC levels with that board.

1 Like

I was surprised to find that XFR Boosts raise the voltage to over 1.45v at times.
Some have set the LLC to Auto with the OC and have been stable.

Has anyone tested the voltage on the board independently from what the BIOS/software is saying?

I can just see in a few months time AMD admitting that they offset it to report higher voltages, just like the temperatures :stuck_out_tongue:

Found a solution to the Asrock X370 p-state overclocking problem. It is linked to what CPU you have, I have the 1700 so the maximum voltage I can use is 1.1875 in as p-state 0 (the maximum voltage is the default voltage for p-state 0 for your CPU). So to increase the voltage, i just set the offset to 13125, which in-turn, allowed me to get voltage of 1.35v (for some reason it doesn't add exactly .1325v) and a clock speed of 3.85Ghz. I then set a p-state 1 to 3.65Ghz at 1.05v and p-state 2 to 3.4Ghz at .9125v. Cool'n'Quiet now works perfectly in windows.

i am curious... what is p state does this underclock the cpu when its not being used to reduice power? insted of having it run 100%? how does this effect the cpu? and when under load it ramps up to ur max oc?

when a cpu is under no load, the voltage will drop, allowing for less heat to be produced. Also a cpu can be damaged overtime with a high voltage and have shorter life span, so having it under clock when idle, it extends the life of the cpu.

so it be pointless for me to do it if i have a custom loop? kuz thats waht im doing...

with a high voltage, at any temperature, will decrease the cpu's lifespan

a pstate is a performance state, it is not a physical thing. a pstate is the frequency/voltage/multiplier settings that the CPU uses when it is under a specific load scenario - pstate0 is full load. The other pstates, that are set at lower performance levels are settings that the CPU uses in other scenarios like when it is idle etc. GPUs and Intel CPUs also have pstates and use them for power management as well.

Unlike Intel bioses, The Ryzen bios has just opened up the possibility to edit the settings for each state. Intel speedstep has them as well but there is no way to access them directly

Hi guys! First time poster, long time lurker. Thanks to this thread i got my p-state overclock working. Thank you! Didn't know why it won't go to p-state 0. I'm on a AsRock x370 Killer SLI 2.30 BIOS.
The goal was to bring my Ryzen's 1700 voltage down a bit. Didn't like the 1.37 Volts at XFR.
Current power states are:
P0 3.5 Ghz 1.1875V
P1 3.0 Ghz 0.9 V
P2 1.0 Ghz 0.6 V
All other settings default, except LLC for the Cpu and SOC witch are set to Level 1. Also voltage is set at Offset with it's setting on Auto. System is stable( also i was able to run it at 3.75Ghz at 1.2V).
Here comes the weird part:
Yesterday i opened the latest version of HWInfo and left it running in the background to see what the temp, voltages, frequencies are.
Aproaching midnight i opened the HWInfo window and saw max clockspeed for every core at 4.075GHz with 1.18V. Couldn't belive my eyes. Took a quick screenshot, and begun logging the sensors. I'll try it again this afternoon to see if the result will replicate.
What do you guys think? XFR kicking in?

1 Like

HWInfo is currently still completely glitched out with Ryzen.
Don't trust anything it says. If you're using the Ryzen Balanced Power management profile also, almost all hardware monitor solutions are completely unable to correctly read the CPU state (that includes, cores, voltages etc) Other power plans are similar. Internally the Ryzen chip does a lot of it's own resource and thermal management so you don't have to worry too much.

Also 4Ghz at 1.18V is impossible. Like completely impossible. You won't even be able to POST with that.
Hence reinforcing the reasoning that you are simply getting garbage readings.

The only tool you can trust as of right now, is Ryzen master and that doesn't give out as much info as the others. I mean on my system CPU-z and HWMonitor report 2.6V Vcore (Nuts) and HWInfo reports 1.55V all the time with the Ryzen balanced power plan (also glitched) I think there are quite a few problems with how all theses different utilities are directly polling the CPU sensors and blocking/interfering with each other.

The thought that it's not accurate has crossed my mind. However both HWInfo and HWMonitor have correct readings as far as i can tell. Voltage and clock speed match those on Ryzen master. SOC voltage, power draw etc all seem fine.
Yes, i'm on the Ryzen balanced plan with minimum processor state at 1%.
Nonetheless I'll keep on investigating.

CPU-Z

HWInfo Clock speed (It's not reading correctly, the CPU does actually scale up and down)

If I switch to the Windows balanced plan the sensors start updating again. Whatever Ryzen balanced does on my system, it blocks reading from the relevant addresses.

HWInfo Power consumption and Voltages

Ok, I'll post some screenshots when i'll get home. In about 3 hours or so.
Try adjusting the minimum processor state at 1% in Ryzen balanced plan.
CPU-z is flawed on mine as well.

Well whoops. Forgot to do that.
Was stuck at 90%. What a horrible default setting.

So it worked? How are your sensor readings now?

Clocks and voltages now show up correctly while at idle. However under load the Voltage still shows up as 0.81V at 3.6Ghz which is plain impossible under 100% CPU load.

Printscreen from last night 4075MHz:

Current no load:

Current full load:

I've begun logging sensor data. Will see if it reports higher than 3.5Ghz and when. If it does I'll change P-state 0 to a higher frequency, then log again to see if it gets past 4075 Mhz.
If there's other tests or prinscreen needed, just ask!

In your first picture it shows that your BUS/Base clock, (BCLK) went up to 116.4Mhz.

116.4Mhz * 35 (multiplier) = 4074Mhz.
This however makes no sense since the ASRock X370 Killer/SLI does not offer an external BCLK.
So either you do not have the board you say you have, or your HWinfo, or Board/CPU is misbehaving.

Either way if you take an actual BCLK of 100Mhz * 35 you get your actual clock.

http://www.asrock.com/mb/compare.asp?Models=X370%20Killer%20SLI/ac,Fatal1ty%20X370%20Gaming%20K4,X370%20Taichi

1 Like

Definitely not an XFR boost as it reads 4075 across multiple cores.
However, that voltage you reported for what you called an XFR boost was not at all high. XFR boosts often use 1.45v and even higher.

While AMD has not revealed the highest safe (sustainable) VDDCR_CPU (CCX) or VDDCR_SOC (data fabric & peripheral) voltage levels, it can be speculated that voltages higher than 1.4500V are generally not advisable for sustained use, at least in conditions / workloads which result in high power consumption (i.e. all cores fully stressed).

Despite it is true that the high-end models can have their default voltage set up to 1.47500V during their maximum single core boost (XFR) operation, the power consumption / dissipation, amount of current flowing and the temperatures are very different between the scenarios where only a single core is fully stressed or all of the cores are fully stressed.