[SOLVED] How to know what CPU power is?

[drea.drechsler]

I've read that CPU power readings in most utilities aren't accurate. But I was wondering about Ryzenmaster...is it providing an accurate reading?

I'd want to think AMD knows the internal workings of their own CPU's enough to provide an accurate read out from the utility they provide, but I'm not sure.

 

[jay32267]

It can be pretty difficult to confirm the readings from the software unless you are electrically inclined.

I would download HWInfo and compare the readings. I find HWInfo to be pretty good.

 

[R4DIO4CT1VE]

I would assume it is accurate enough. I know Ryzenmaster has some 20c offset on their readings for some reason, not sure if they changed that or not though.

 

[drea.drechsler]

I would assume it is accurate enough. I know Ryzenmaster has some 20c offset on their readings for some reason, not sure if they changed that or not though.

That's not just Ryzenmaster, it's how the certain Ryzen 1000 and 2000 processors reports internal temperature. You have to know your utility and how it deals with that offset. And it does not apply to Ryzen 3000 processors as they report actual now.

 

[Finstar]

If you want a 100% accurate reading then get one of these bad boys and slap it around your cpu power cable:
https://www.amazon.com/Fluke-323-Tr...mp+multimeter&qid=1565012051&s=gateway&sr=8-4

 

[drea.drechsler]

It can be pretty difficult to confirm the readings from the software unless you are electrically inclined.

I would download HWInfo and compare the readings. I find HWInfo to be pretty good.

I have compared to HWInfo (the readings are 3-5 W different with a 3700x on a B450m Mortar)...and I've read some of Martin's (HWInfo author) background info on his web page about the readings provided by HWInfo readouts. He does say that the SVI2 voltage readings are the "most" accurate but that current and power are iffy as they require calibration to specific motherboards for accuracy.

But that's his utility, I'm more concerned with Ryzenmaster. I'm aware there is a lot of internal processor operating data available but the means to query it is locked up by NDA's...something AMD probably doesn't have to worry about for their own utility (which is why it alone can work for adjusting clocks from within Windows). That's what gives me some hope the power and current readings on RM could be accurate.

 

[drea.drechsler]

If you want a 100% accurate reading then get one of these bad boys and slap it around your cpu power cable:
https://www.amazon.com/Fluke-323-Tr...mp+multimeter&qid=1565012051&s=gateway&sr=8-4

Except that would include VRM power losses, which can be substantial else it wouldn't get toasty warm during heavy encoding sessions.

Also, I'm not worried about 100% accurate. +/- 10% would be quite enough. With my 1700@3.9 Ghz the CPU power readout was laughably high: 160-180 watts when encoding a movie, H.264. But temps were still in the 55-60C range. With this 3700x on the same board power is reasonable at 100-104 W while running at 4.15-4.175 Ghz steady doing a similar H.264 encode. And temps are 72-73C under same 240mm AIO cooler. I think the difference is RM now has to know a correct processor power load for PBO2 to effectively control boost within the defined limits.

 

[Finstar]

Except that would include VRM power losses, which can be substantial else it wouldn't get toasty warm during heavy encoding sessions.

Also, I'm not worried about 100% accurate. +/- 10% would be quite enough.

Isn't including the whole power draw the point?
What use would the cpu draw only be, for example when choosing a new PSU?
The total draw is exactly what i'd want and what has the most relevant information.

 

[drea.drechsler]

Except that would include VRM power losses, which can be substantial else it wouldn't get toasty warm during heavy encoding sessions.

Isn't including the whole power draw the point?
What use would the cpu draw only be, for example when choosing a new PSU?
The total draw is exactly what i'd want and what has the most relevant information.

I'm interested because I want to limit power draw of the CPU through PBO. I don't like letting the CPU bounce off the 95 C Tjmax limiter should a power virus (like P95 small FFT) get launched. Anything with tightly looped AVX instructions could do it. Ryzen has no BIOS settings to force a down throttle as Intel processors do, so setting a power limit in PBO seems the most reasonable way to do it. But I want to know what true high-performing and reasonable power draw is in something like H.264 encodes and Folding@Home. Both of those apps use AVX extensively - but responsibly so the CPU's don't go into thermal meltdown - so their performance will be used as the power trip threshold.

 

[Finstar]

I'm interested because I want to limit power draw of the CPU through PBO. I don't like letting the CPU bounce off the 95 C Tjmax limiter should a power virus (like P95 small FFT) get launched. Anything with tightly looped AVX instructions could do it. Ryzen has no BIOS settings to force a down throttle as Intel processors do, so setting a power limit in PBO seems the most reasonable way to do it. But I want to know what true high-performing and reasonable power draw is in something like H.264 encodes and Folding@Home. Both of those apps use AVX extensively - but responsibly so the CPU's don't go into thermal meltdown - so their performance will be used as the trip threshold.

In that case the Ryzen Master numbers are 100% accurate, even though they might not be in reality, it's those numbers that determine when the cpu throttles down.

 

[drea.drechsler]

In that case the Ryzen Master numbers are 100% accurate, even though they might not be in reality, it's those numbers that determine when the cpu throttles down.

Thanks