Ryzen CPU idle speeds

[natr0n]

Was wondering about this since different windows versions give me different idle speeds on balanced power profile. (Using Core Temp as reference for clocks) using a 5950x

(I'm not great at poll making)

 

[Psychoholic]

My 9950x jumps around alot but seems to settle around 2535mhz according to hwinfo64

 

[windwhirl]

Was wondering about this since different windows versions give me different idle speeds on balanced power profile.

(I'm not great at poll making)

Are you looking for a specific Zen generation or anything goes?

My 3600 can either shut down the idle cores or go as low as 300 Mhz, IIRC from Ryzen Master. I'll look into it again later

 

[natr0n]

Are you looking for a specific Zen generation or anything goes?

My 3600 can either shut down the idle cores or go as low as 300 Mhz, IIRC from Ryzen Master. I'll look into it again later

anything goes. I use core temp mostly so maybe Im wrong about true idle speeds.

 

[A Computer Guy]

This is a bit of a tough question. My 5950x will idle like 300MHz on a couple of cores (with most parked) while playing a YT video according to Ryzen Master.

 

[natr0n]

This is a bit of a tough question. My 5950x will idle like 300MHz on a couple of cores (with most parked) while playing a YT video according to Ryzen Master.

View attachment 369684

I wonder why some tools report different. I know low clocks show on ryzen master and hwinfo which has too much info imo.

Even aida64 never shows me those low clocks. Should be standardized reporting or something.

 

[Psychoholic]

I wonder why some tools report different. I know low clocks show on ryzen master and hwinfo which has too much info imo.

Even aida64 never shows me those low clocks. Should be standardized reporting or something.

Some tools may be reporting effective clocks vs real clocks.

 

[Vincero]

I've noticed there is some inconsistencies between different tools - part of this I attribute to different tools using different sampling methods, after all the more it works to get the data will impact the idle time on at least one core.

Ryzen master, hwinfo and cpu-z's hardware monitoring app can give fairly reliable information in terms of providing logging of the minimum and maximum but I always take the current info as a guide only.

One word about Windows task manager: lies... I've found to be useless at indicating idle and high utilisation clock speeds - even when all cores have boosted past 4GHz the speed in task manager is much lower.
I only mention this as people who don't know any better might believe it.

 

[ShrimpBrime]

5600G is at fixed all core 4ghz. on my sons rig. No idle time for his PC any ways. If it's on it's gaming

 

[AusWolf]

Windows can't properly monitor Ryzen idle speeds because they fluctuate on a per-core basis and too quickly to detect. What you see in Task Manager is bogus.

 

[DaemonForce]

It's cool. Humms away at 3980-3990 all day, idle, half load and max load.
Might be an issue locking BCLK or float error because I keep it pinned at 4000MHz.
Also these stupid ass memory leaks are directly related to updates. Otherwise Ryzen go BRRRRRRR.

 

[FreedomEclipse]

3149-4724mhz with my 7800X3D

 

[windwhirl]

The 1 Mhz clock speed made me laugh

 

[AusWolf]

Guys, calm down...

Let's not forget that there's no such thing as "idle". Even the smallest of background tasks fires up your CPU and throws your clocks into space for just a millisecond, making all readings invalid, especially in the ancient Windows Task Manager.

 

[natr0n]

Guys, calm down...

Let's not forget that there's no such thing as "idle". Even the smallest of background tasks fires up your CPU and throws your clocks into space for just a millisecond, making all readings invalid, especially in the ancient Windows Task Manager.

I believe this is more of a modern windows issue bloated with services to spy and whatever.

Im on my windows 8.1 drive watching youtube currently and solidly idle maybe 4 cores jump a tad bit when starting video goes idle instantly again.

 

[AusWolf]

I believe this is more of a modern windows issue bloated with services to spy and whatever.

Im on my windows 8.1 drive watching youtube currently and solidly idle maybe 4 cores jump a tad bit when starting video goes idle instantly again.

Sure, modern Windows is bloated as heck, but still, true idle doesn't exist. You can get close to 0% CPU usage, but you'll never achieve true 0, that is 0.000000...%, especially with monitoring software open which also throws your readings off.

 

[natr0n]

Sure, modern Windows is bloated as heck, but still, true idle doesn't exist. You can get close to 0% CPU usage, but you'll never achieve true 0, that is 0.000000...%, especially with monitoring software open which also throws your readings off.

True of course, but I'm not meaning 0%. Just like not touching anything kb/mouse as if you play music/video and walk away scenario.

 

[AusWolf]

True of course, but I'm not meaning 0%. Just like not touching anything kb/mouse as if you play music/video and walk away scenario.

Yeah, but Ryzen changes clocks so quickly even at a 0.001% usage that monitoring software, especially our ancient Task Manager can't follow and gives you false readings. Not to mention per-core clocks which is a completely alien concept for Windows.

 

[natr0n]

Yeah, but Ryzen changes clocks so quickly even at a 0.001% usage that monitoring software, especially our ancient Task Manager can't follow and gives you false readings. Not to mention per-core clocks which is a completely alien concept for Windows.

I understand now what you mean.

 

[Chrispy_]

Was wondering about this since different windows versions give me different idle speeds on balanced power profile. (Using Core Temp as reference for clocks) using a 5950x

(I'm not great at poll making)

Thanks for reminding me to reinstall CoreTemp. I've just installed 24H2 on this because microsoft weren't offering it as an update and I need to validate a couple of rendering apps on this version.

As for your idle speeds question, "yes to all" which is technically correct but entirely unhelpful.

Your idle speeds have very little to do with your CPU or OS and far more to do with which and how many services and applications you have installed. The amount of background crap running on a modern PC is insane and even though most of it demands very little of your CPU, there are hundreds of services and background apps that mean your CPU will almost never be idle.

Windows task manager, and even tools like CoreTemp or HWInfo aren't fast enough to catch the very rapid frequency changes of Ryzen's rush-to-idle default behaviour in most power plans. Even AMD themselves say that Ryzen Master isn't fast enough, but it paints a much better picture of the true clocks at any given moment than the aforementioned third-party utilities.

What's actually happening is that Ryzen is suspending cores (effectively 0MHz) to save power, but they're so suspended that they're not even reporting their clocks which means utilities tend to just assume they're still at the last reported clock. Task manager takes an average of known clocks, which means it is already out of date for all suspended cores and therefore inaccurate. CoreTemp, I think, looks at the slowest active core and reports that, which is closer to accurate for any given instant but still likely missing the correct values for near-idle states.

Knowing your idle clockspeed is basically useless, because on any modern Ryzen, the real answer is "0 MHz for suspended cores and at least the base clock is for the 0%-1% active core that's currently awake and rushing-to-idle as designed". If the tool you're using reports a different value, you're just seeing the error of the tool.

 

[AMF]

i idle around 4.0-4.3 ghz

 

[Zach_01]

Was wondering about this since different windows versions give me different idle speeds on balanced power profile. (Using Core Temp as reference for clocks) using a 5950x

(I'm not great at poll making)

3300MHz using balanced power plan and HWiNFO64 for monitoring with “Snapshot CPU Polling” enabled.

The 1 Mhz clock speed made me laugh

View attachment 369708

This is a bit of a tough question. My 5950x will idle like 300MHz on a couple of cores (with most parked) while playing a YT video according to Ryzen Master.

View attachment 369684

I wonder why some tools report different. I know low clocks show on ryzen master and hwinfo which has too much info imo.

Even aida64 never shows me those low clocks. Should be standardized reporting or something.

Let’s not forget that after a point (clock going downwards) some cores are entering “halt state” and further down the “power down” state.

CPU states
C0: active state
C1: halt state
C6: power down state

Traditional core clock metric only refers to the active C0 state. So the min 3300MHz I’m seeing tells absolutely nothing about true core state.
Today CPUs/GPUs are so dynamic that another metric is necessary.
Enter “Effective Clocks”

RyzenMaster using AMD’s proprietary method to calculate core clocks including all states, and no one really knows how it works.
To my understanding AMD is showing whatever they want. Not just for clock but for temp, power, voltage and rest.

HWiNFO64 sensors mode uses, let’s say a “universal” method for effective clock.
In the HWiNFO64 you can monitor all these clocks and the C0/1/6 states also in C-States residency section.
Do not set polling interval below 1000ms as it’s messing up with halt/power down states.
Use “Snapshot CPU Polling” (for Ryzen) to minimize the observer effect.

Yes Ryzen can alter clocks, voltage and states at least 50times per second but leaving monitoring long enough you getting more and more accurate readings.

It’s not strange to see a few MHz as core clock speed on effective clocks. Below 50 or maybe 20MHz, the core has entered the C6 (power down state).

EDIT:
Its around 50MHz

FEATURE - Effective clock vs instant (discrete) clock

It has become a common practice for several years to report instant (discrete) clock values for CPUs. This method is based on knowledge of the actual bus clock (BCLK) and sampling of core ratios at specific time points. The resulting clock is then a simple result of ratio * BCLK. Such approach...

www.hwinfo.com

 

[AusWolf]

My 7800X3D hovers at 1.5 GHz with dips even around 1 GHz at 0.5% usage with Steam, Heroic launcher and Chrome running in the background on Bazzite (Fedora Linux). I'm impressed!

But just like I said above: take this with a grain of salt.

 

[_roman_]

I'm not sure if these are valid numbers.
Idle with vivaldi browser and some shell. The Virtual Cores (hyperthreading) are also reported 6+6

Sienna_Cichlid /home/roman # grep model\ name /proc/cpuinfo |head -n 1; grep MHz /proc/cpuinfo
model name    : AMD Ryzen 5 7600X 6-Core Processor
cpu MHz        : 545.000
cpu MHz        : 483.520
cpu MHz        : 594.080
cpu MHz        : 545.000
cpu MHz        : 526.780
cpu MHz        : 545.000
cpu MHz        : 545.000
cpu MHz        : 529.448
cpu MHz        : 520.965
cpu MHz        : 503.828
cpu MHz        : 545.000
cpu MHz        : 486.180

 

[Chrispy_]

I'm not sure if these are valid numbers.
Idle with vivaldi browser and some shell. The Virtual Cores (hyperthreading) are also reported 6+6

Sienna_Cichlid /home/roman # grep model\ name /proc/cpuinfo |head -n 1; grep MHz /proc/cpuinfo
model name    : AMD Ryzen 5 7600X 6-Core Processor
cpu MHz        : 545.000
cpu MHz        : 483.520
cpu MHz        : 594.080
cpu MHz        : 545.000
cpu MHz        : 526.780
cpu MHz        : 545.000
cpu MHz        : 545.000
cpu MHz        : 529.448
cpu MHz        : 520.965
cpu MHz        : 503.828
cpu MHz        : 545.000
cpu MHz        : 486.180

I'm fairly sure those numbers are invalid for one very obvious reason; You have 6 cores, yet there are more than 6 different clock speeds listed which means that two threads running on the same core are reporting different speeds which would be impossible.

Reaslitically, Ryzen Master seems to show that only one or two cores are active at any point when clocks are down in the triple digits, and that's probably only showing more than one because the activity has jumped between cores during the sampling period. To have all cores loaded, but none above 600MHz implies that your cpuinfo is showing the last reported speed for suspended cores, rather than your CPU usage actually being 12-15% at idle. Chances are good it's actually reporting 0-2% usage at idle, which makes sense at 1*500MHz out of a total 100% usage of 6*4700MHz