CPU Power draw at low load?

[Gimpymoo]

If a CPU TPD is rated at say 100w and CPU is running at a low load of say 10%, is it possible to calculate what the power draw would be or is it dependent on too many variables?

Or does the CPU need there about that power to make it function?

 

[Hagar]

At this moment in time, the Ryzen 1700 (TDP 65W) shows 26W (package) and 1.8W (total cores) just browsing and idling mostly at 1550MHz with 3-5% loading. CPU current 2-3A, vcore 0.875V.

It would be a difficult calculation to perform.

 

[LuckyBenski]

Many many, too many variables. The voltage and clock speed are microadjusted constantly (many many times per second). Combine this with the fact you have 2-16 cores each capable of being clocked from say 500 milion to 4500 million times per second.

Energy dissipation is basically a function of clock speed times voltage. So, the energy dissipation varies from millisecond to millisecond. It's all an average.

Is there a reason you need to know the CPU's momentary consumption specifically? TDP is a good guideline for system design and that's about it. Motherboard obviously consumes power, so does RAM, etc

 

[Gimpymoo]

All stats/specifications point to "Peak" power draw, which is fine.

What I would like to know is "low load" power draw, if that makes sense?

Specifically:

https://ark.intel.com/products/129951/Intel-Pentium-Gold-G5400-Processor-4M-Cache-3_70-GHz

Vs.

https://ark.intel.com/products/129949/Intel-Pentium-Gold-G5400T-Processor-4M-Cache-3_10-GHz
(Running at 25W TDP at 2.1Ghz).

If both CPU's are running at 10% load for example, what would the difference in Power draw be.

Basically, if a CPU is using 10% of its available resources, how much wattage does it consume, all things being equal as they can be in relation to its rated peak wattage.

Also, as in this instance, if a chip has a lower TDP, does this change fluctuate in line with the maximum, thus the "low powered" CPU using lower wattage than the "high powered" CPU.

Sounds obvious but you never know.

 

[LuckyBenski]

All things being equal is still difficult. Easy to compare two chips under the same load, but difficult to quantify outright. Some CPU instructions use more energy per clock for example. I guess the closest thing to what you're looking for is an actual measurement from the chips in question. You're very unlikely to see a general quantified estimate across all architectures

 

[LuckyBenski]

I wonder if you could get information on the performance per watt and extrapolate it?

 

[Gimpymoo]

What I am trying to ascertain is that for "low load" purposes, is it worth paying more for a low power chip or does the high wattage CPU use the same juice under low loads anyway?

 

[Rroff]

My 4820K albeit not using the max power saving options and still boosting a fair bit uses about 30 Watt idle and 39 Watt average under low load. Compared to easily hitting 100 Watt during gaming, etc.

(This is with all cores at 4GHz minimum as well not stock).

 

[LuckyBenski]

What I am trying to ascertain is that for "low load" purposes, is it worth paying more for a low power chip or does the high wattage CPU use the same juice under low loads anyway?

In my experience T and S series CPUs are just the same silicon with lower stock clocks. I've heard some may use lower voltages too but not sure. So, more than likely power consumption won't be much different at low loads.

Interestingly my i5-3470S has lower stock clocks but single core boost is the same as a standard 3470. So its single core performance doesn't really suffer.

Several chips use the same trick to differentiate models - i5-4430, 4440 and 4460 are all the same silicon with different clock speeds. Performance varies by exactly as much as you'd expect for a 100Mhz difference.

 

[Gimpymoo]

My 4820K albeit not using the max power saving options and still boosting a fair bit uses about 30 Watt idle and 39 Watt average under low load. Compared to easily hitting 100 Watt during gaming, etc.

(This is with all cores at 4GHz minimum as well not stock).

Thanks, that is good to know.

It makes me wonder that if your not going to be pushing a CPU, a 54W variant can run just as low power/cool as a low powered 25W counterpart with the potential for faster speeds w

It is only when you start putting a CPU under load do you see differences.

Hoping I am wrong though as that would mean I have paid more for a CPU which for low load conditions, is a gimped version of something cheaper and faster.

 

[smilingcrow]

What I am trying to ascertain is that for "low load" purposes, is it worth paying more for a low power chip or does the high wattage CPU use the same juice under low loads anyway?

The difference at low load levels will be insignificant
My i5-8400 desktop with 32GB of RAM and two SSDs idles at around 25W as measured at the wall socket.
Under light loads that varies between late 20s and mid 30s so there is not much room for savings.
Most of the power savings come by not using higher clock speeds as the voltage increases when the clock speed rises.
Since power consumption is related to the square of the voltage and directly to the frequency keeping the voltage down is the key.
Buying a regular CPU and lowering the speed and voltage should give you almost all the power savings. Plus you have the extra headroom if required in the future and they tend to be cheaper and easier to source than the lower TDP versions