Confused about how TDP and Power Draw Works? Look no Further!

Hardware Hub Build a PC
1

Recently, I made a post regarding how TDP relates to actual power consumption, questioning whether "X" power supply would support "Y" CPU's, and got a lack of responses, even after multiple "bumps". Whether that was because nobody cared to help me out, or if nobody knew the answer, I don't know. However, I now have the answer (which was technically on Wikipedia)!

To make it quick, if you want to get your PC's actual power usage under load, find the TDP of the CPU, multiply it by 1.5, and do the same for the other components inside of the computer (GPU, HDD, etc). It's that simple.

I'll use my recently upgraded server as an example. Specs:
CPU: 2x Intel Xeon X5650 (6-core 2.66GHz) TDP: 95W
24GB ECC Registered RAM
No HDD's or PCIe devices used. Booted the system via Win 2 Go

So, let's pretend that the RAM uses no power (realistically, it does, but not a whole lot). The TDP of each Xeon x5650 is 95W, and there is two of them.

Math:
95*1.5 (1 CPU x the 1.5 value given above, and on Wikipedia) = 142.5W
Now, multiply 142.5 by 2, since there are two identical CPU's in the server
142.5*2=285W

While running Prime95 with the max stress test, the system pulls 280W from the wall (according to my EM100 Energy Meter).

That's only a 5W disparity, which I consider to be a within the margin of error.

The power usage of the system would go up if all 24GB of RAM was being used (while testing, only 1GB was actually being used), how, it won't be a significant amount (I'd guess probably 50-100W for my 24GB ECC Registered DDR3, which is about 2-4W per DIMM), which is why I didn't really account for it in my testing. In the future, I plan to do the same tests on my GPU (GTX 970) to see how accurate power draw will be on a system with a dedicated GPU.

So, there you go! Now you can accurately calculate system power draw using the given TDP values, to help you decide on how large a PSU to buy for an upcoming PC build. As Louis Rossmann would say, I hope you learned something! :D

To read further on TDP: https://en.wikipedia.org/wiki/Thermal_design_power

2

nice tests but did you use the value of power drawn from wall without taking into consideration psu efficiency,which is not a constant value and varies depending on voltage and load ? or did you include that also 😃Another thing is that even if you determined the" tdp multiplier "(1,5) it may be good only for this pair of cpus.things get more energy efficient whith each generation which indicates that tpd multiplier will be getting slightly lower. but generally its not a good idea to calculate it to the single digit value as it is not the best practice to run psu at full whistle.

3

I did miss over that.

According to HPE, my 460W PSU is 80+ Gold Efficiency, meaning that the actual power draw from the CPU's was closer to the 220-240W range, putting them much closer to the rated 190W TDP of the two CPU's.

My other system with the GTX 970 has an i5 2500, which is about 3 years newer than X5650. I did say that I plan to run more tests in the future, comparing TDP to actual power draw :p

What I wanted to basically show with these results is that the TDP given by a manufactuer will give a close value to the actual power draw of the system. The reason for the test that I did above, and the one I plan to do on my i5 2500 system is because I've heard countless times in the past of people saying that TDP ≠ Power Draw. From my results, it shows that it is true, however, the TDP does give quite a close estimate as to how much power the system will use.

Regarding the 1.5 multiplier, according to Wikipedia:
"Some sources state that the peak power for a microprocessor is usually 1.5 times the TDP rating.", which is why I chose to include the 1.5x multiplier.

Anyway, I basically wrote this all up so that uninformed people could make good judgements as to how large of a PSU that they should buy for a PC build. So let's say that the TDP of system is totalling 300W, they should buy at the minimum, a 450W PSU (because of the 1.5 multiplier), or even a 500W PSU if they plan to add more RAM or storage drives in the future.

Hopefully that cleared up all lose ends, and I plan to run the same tests on my i5 2500 system this week, to compare the results to my X5650 system.

3 Likes
4

I like the idea that you did the testing and proven that yourself.Many people spread false info that tdp =power draw of cpu.please share the rest of the results .

1 Like
5

Well yeah i had that discussion a couple of times allready.

TDP = Thermal Design power, aka heat output.
Its somewhat related to total powerdraw TPD, but they are definitelly not the same thing.

1 Like