Calculating Power Draw with HWInfo Confusion?

[Dave1001]

I have an 1000w PSU with an RTX 3070 and a 5800x. the max power draw for the 3070 is somewhere between 220-250w, and the 5800x is rated for 140w? however I have Eco Mode 65w active, so it draws even less than that

I’m trying to calculate my power draw from the PSU between my cpu and gpu, but HWINFO has a lot of wattages on display and adding their maximums gives me an incredibly large number


the attached image is the GPU and when the watts spiked up under gaming load (the current watts were taken at idle so just pay attention to the Maximum column). they tend to spike at the same time and CPU does the same. adding up all those maximum values gives me a number between 600-700w…that doesn’t seem right?

furthermore, the CPU wattages listed are:

CPU Package Power
Cores 0-7 Power
CPU Core Power
CPU SoC Power
Core + SoC Power
and CPU PPT

if I then add all the maximum values from the CPU, combined with the GPU, my power draw is apparently 980w?? this can’t be right, since realistically you should only need a 600-750w PSU to run a 3070/5800x setup

what am I calculating wrong here? are all these wattages taken from the PSU, or add to the total power draw? or should I only be adding the CPU Package Power + GPU Power? I have no overclock or anything to drastically increase power draw, and this doesn’t even take into account my case fans so 980w total cannot be right. even if the CPU wattages spiked at different times, the total power draw here would still be around 800w. and the GPU alone is allegedly consuming up to 700w?

some insight would be appreciated, really confused here

 

[Martin]

Calculating total power consumption of a system is possible only if the PSU supports such capability as only that can monitor all rails distributed in the whole system.
Otherwise there are several rails in the system that don't support monitoring of their consumption so you can only make certain estimations.
As for the CPU, the "CPU Package Power" should be the total power consumed by the CPU and you should not not sum that up with other CPU power values as they cover only certain sub-parts that are already included in the total/package value. There are multiple rails driven to the CPU and only some of them are capable of reliable monitoring. Some others (usually minor) are estimated. Similar for GPU.

 

[Dave1001]

Calculating total power consumption of a system is possible only if the PSU supports such capability as only that can monitor all rails distributed in the whole system.
Otherwise there are several rails in the system that don't support monitoring of their consumption so you can only make certain estimations.
As for the CPU, the "CPU Package Power" should be the total power consumed by the CPU and you should not not sum that up with other CPU power values as they cover only certain sub-parts that are already included in the total/package value. There are multiple rails driven to the CPU and only some of them are capable of reliable monitoring. Some others (usually minor) are estimated. Similar for GPU.

I also got a similar answer from someone on reddit, and they also said that I shouldn't add them all up because they aren't all exclusive. so if I wanted to gain an "estimate", I would basically want to only add CPU Package Power and GPU Power, right

 

[Dalai]

so if I wanted to gain an "estimate", I would basically want to only add CPU Package Power and GPU Power, right

That only gives the power of the "core" components. You still have to add other components such as the motherboard, drives, fans, pumps (if any) and anything else that uses power. HDDs can easily consume around 4-15W, SATA SSDs also have a wide range of power usage, so have fans. One simple way to see how much power a PC draws is to use an accurate power/electricity meter that plugs between the wall outlet and the PC, such as the (probably well-known) Kill-A-Watt:

Kill A Watt - Wikipedia

en.wikipedia.org

Regards
Dalai

 

[LeDoyen]

With the CPU and GPU you basically have a good estimate already.
Fans don't use much power at all. Once they have started, at steady speed the may use 0.2 - 0.3W a piece. SSDs use power mostly when reading/writing obviously. Standby power is negligible. But RGB lighting oh boy.. you got to count that depending on the build it can draw as much as a mid-range CPU.

The wall meter is good if you are interested about power usage and costs, but since it also factors in the PSU efficiency losses, it's not ideal either.
The wall meter would give you the high estimate, the GPU+CPU the low estimate. the right figure for components power draw is somewhere in between. that can give a good ballpark. Obviously, a PSU with power readings would be the best

 

[Dalai]

Fans don't use much power at all. Once they have started, at steady speed the may use 0.2 - 0.3W a piece.

It depends on the fans, of course. Fast-spinning fans can easily draw 6 Watts or more. Even regular fans (Arctic F12 or Noctua NF-B9 Redux-1600) draw 0.12 A which is around 1.4W at 12V. Sure, it also depends on the actual fan speed, i.e. whether or not they're controlled by some means like a fan controller or the motherboard. But a lot of such small pieces can add up to a significant number pretty fast...

SSDs use power mostly when reading/writing obviously. Standby power is negligible.

Again, it depends on the SSD. Drives with a high idle power consumption of several Watts exist.

The wall meter is good if you are interested about power usage and costs, but since it also factors in the PSU efficiency losses, it's not ideal either.

True. However, you pay for what is drawn from the wall, regardless of where the power is going, be it converted into compute power or movement (in case of fans, pumps and HDDs) or wasted into heat due to efficiency losses.

Regards
Dalai

 

[LeDoyen]

For the fans the figure on the label is max current at max speed, and even then it's often inflated a bit to have some margin.
but of course, power depends on the fan speed, so if you run them at 100% all the time it's worth adding to the power used.
For example i have 3 Arctic P120 on a radiator (given for 0.13A or 1.56W). at full speed they draw 4W for 3 fans, or 1.3W a pop (not far). at 50% speed which is more in line with day to day use, it falls down to 1W for 3 fans or 0.3W. Same kind of figures for the other models of fans i have (Bequiet Light wings, EK FPT, Noctua NF-A9), they draw peanuts in normal use, but power goes up exponentially past 70-75% speed.
I digress ^^ but the point was, even with fans blasting, unless you use high speed fans, it adds pennies to the pound, only a few watts, compared to the hundreds drawn by CPU and GPU, or even the efficiency losses of the PSU. RGB lighting can sometimes use way more ( another example with my case, lighting on white full brightness is about 34 Watts. 15W with a toned down discreete lighting scheme, which is still 7 times more than what the 13 fans draw).

Interesting topic once you start to measure all the things!