How to do Power Consumption Testing in CPU Reviews?

[TheoneandonlyMrK]

I guess "kilowatt-hours" would be the unit of energy quoted in the USA the most. But scientists would typically use Joules (aka watt-seconds) instead.

Hmmm... some discussion / polling on what makes sense to more people probably should be done. I would personally understand Joules. But maybe more people are familiar with "kilowatt-hours" instead??

-------

A Watt is literally a "Joule-per-second". So a Kilowatt-hour is "1000 Joules-per-second hours", which is more confusing IMO, despite being the commonly quoted figure in the USA. Also, given how much energy these benchmarks use (probably 300 seconds @ 100 Watts or so, or 30,000 Joules or so), that's "too small" to be quoting kilowatt hours (aka: a kilowatt hour is 3600000 Joules. So a typical benchmark would be 0.008 kilowatt-hours of energy)

What would look better in an article? 30000 Joules or 0.008 kw-hr ??

People know Watt's, I know them all, I'm an engineer and do not require schooling on this.

My point though is we have a unit of measure everyone knows.

The argument seams to me, is that some think parts draw less power in use and want their version of USE to be the only, or main version of Use endorsed by website's and Benchmarks.

I would compromise and do what is done now plus gaming power, due to gaming easily being one of the prevalent use cases an enthusiastic pc fan might engage in.

What we have now is fine for power user's.


As for settings I have opined for default CPU settings not Auto since that auto allows every motherboard to overclock a bit, though as sold testing is in a way fine with me too if OEM tweaks are mentioned.

 

[R0H1T]

what wattage is the norm for that 12900k 125 or 241 or the nearly 400watts its using during load?

That depends on the (work)load, load voltage & of course how good or bad your chip is. Lastly the cooling.

With ADL Intel (unlocked k) chips will use whatever the cooling & motherboard allows them to ~ so there's no real hard limit.

 

[dragontamer5788]

People know Watt's, I know them all, I'm an engineer and do not require schooling on this.

But Watts is power, not power-consumption (aka Energy). That's why I'm talking Joules, Watt-hours, or Kilowatt-hours. Watts are the wrong unit to use fundamentally.

Batteries are quoted on 'Milliamp Hours". Energy prices are quoted on Kilowatt-Hours. Scientists use Joules (aka: Watt-seconds). Those are our choices frankly, unless you wanna go full Imperial on people and use like, Acres (aka: horsepower-days) as a unit.

 

[TheoneandonlyMrK]

But Watts is power, not power-consumption (aka Energy).

That's why I'm talking Joules, Watt-hours, or Kilowatt-hours. Watts are the wrong unit to use fundamentally.

How so, the watts something pulls in the moment ,if then a continuous load is still proportional to the watt hour's.

 

[RandallFlagg]

I think everyone has been programmed to expect only what they are given.

Even when it is not relevant to how much power they actually use.

I've had HWInfo64 running since early this morning, normal work day, about 4 hours in, overclocked 10850K. This is what I got.

38W minimum, 108W peak, 44.6W average - just CPU package power. In other words, during work my CPU consumes about 6W average over idle, and again it is overclocked.

 

[Frick]

Calories dudes, and make silly grahps about how many pushups one has to do to consume the same amount, and tie it to the scoring system. "Too many pushups" become a negative.

 

[neatfeatguy]

I've seen lots of comments across multiple sites that some people don't realize that power draw for Adler Lake series is lower/equal (or wherever it does fall) to Ryzen's 5xxx series when it comes to gaming.

Perhaps just add a couple other charts, as you already suggested with CP2077 and maybe one or two others for programs people do make use of.

Personally, my normal use of my computer would be gaming and also the use of Handbrake to convert (to H.264) a .mkv file to .mp4 format. I tend to convert 20-30GB blu-ray movies as I get them, each one takes around 35 minutes on my 5900x for the settings I use. If handbrake was an option to use, you could just pick one of the preset options for quality; such as FAST 1080p30.

I'm not sure what other large file/data usage programs others use, but I remember someone suggesting zipping and unzipping large files. This way you add in 2 or 3 more charts for power consumption without really changing anything that you're already doing. I don't think it would add much extra time to your benching procedures.

 

[W1zzard]

What would look better in an article? 30000 Joules or 0.008 kw-hr ??

Or maybe we go with 8.3 watt-hours? Or 30-kilojoules ?

I felt like kJ was the most reasonably scaled unit, that's why I've been using it for many years in my CPU reviews

 

[Mussels]

Following this as i'm quite curious to where it ends up
After seeing JayZ screwing his youtube video up so badly, we need a website that takes power consumption tests seriously, that doesn't change to support whatever new product is being sponsored


Test setup:
Everything except CPU Cooling should be identical, even down to locked case/testbench fan RPMs. (Excluding CPU, GPU, chipset)
If a system requires bigger coolers to work, that should show in the results - a 3700x running on a wraith stealth is gunna be far more efficient than a 12900K on a360mm AIO, and that needs to be brought up.

A lot is already covered, and i don't know what to improve but i personally look for:


System as a whole results (The problem here is that one single hardware change makes these numbers irrelevant to an end user)
CPU/GPU alone results (HWinfo/self reported values are helpful here, or measuring at power connectors if possible to verify them)

ST full load benchmark consumption
MT full load benchmark consumption
Gaming full load benchmark consumption
Some kind of true low load test: An offline windows OS with a script that opens a locally saved website, and changes page every 3 minutes? Opens a word doc and copy pastes content?


The thing with high load tests like cinebench is, the more performance you have, the faster they complete.
Average PC tasks are time based instead - so a system could be super efficient at crunching high end workloads, but waste power doing light tasks over long periods of time.
Some people game for 8 hours straight one day, then barely touch the system til the next weekend. Others like me leave it on 24/7 typing up essays on really weird topics on TPU at all hours of the day and night.

 

[joemama]

Do people really need the power consumption for certain tasks?
I feel like the performance and the power draw would be enough for normal people to judge a CPU

 

[R0H1T]

Average PC tasks are time based instead - so a system could be super efficient at crunching high end workloads,

They can also be I/O or network bottlenecked, in terms of CPU efficiency I'd argue synthetic benches or something like a (looped) Cinebench is probably the best of the lot.

 

[chrcoluk]

Mussels raises some good points as well, I dont like that tests tend to get done in perfect conditions for the product been reviewed, its a pattern I have noticed in the industry, and is one of the reasons I felt Alder Lake should have not been reviewed on Windows 11, to highlight the fact its going to be more problematic on non optimised OS.

So yeah basically not the best air cooler on the market ramped up to static 100% fan speed which I have seen some reviewers do.

I would like to see non AAA games added to the testing mix, there is a lot of emphasis on new AAA titles, add some DX9 60hz vsync'd indie titles into the mix, they will consume more power then idle, but not the same as playing something like a modern FPS with uncapped frame rate. However these comments are more for the overall review, I expect the power testing is wanted to be kept as simple as possible. Thats why I suggested to have it basically as idle, turbo disabled, and under load with turbo on. Two load tests, one under gaming conditions, one for rendering.

As a consumer, I would place more importance on the rendering load as that informs me more on my choice of PSU, and gives me an idea of peak load to be expected, as there is no guarantees, future games/software wont be more demanding than today.

Yeah the jay power test video is lazy.

The problem is likely that he is trying to pump out videos as fast as possible as these youtubers have a mindset that getting out content early as possible is better. So he clearly cut corners in not dissembling the AMD rig to equal out the components.

Not to mention the clickbait titles.

I always prefer media like techpowerup, written reviews.

 

[W1zzard]

@chrcoluk you made some great posts that you deleted, i feel they add to this topic and are relevant. want me to restore them?

 

[qubit]

I'll trust whatever you decide to do.

 

[AusWolf]

Somehow I've always found wall power data less usable than component-specific power consumption because
1. Every system is different, so wall power only describes the one specific system tested.
2. One needs to configure cooling for individual components, not whole systems.
3. It's easier to compare idle and in-use power consumption when you don't have to take the rest of the system into account.
4. A component swap (for example a new GPU used for CPU tests) makes old review data ucomparable.

Whole system power consumption is only useful when choosing a PSU, imo.

A game power consumption figure would be quite useful, though.

Edit: also, I think it would be useful to run tests and benchmarks at Intel/AMD default power settings, not just the motherboard's default "Auto" setting, which tends to set ridiculously high limits, especially on high-end boards.

 

[Mussels]

Do people really need the power consumption for certain tasks?
I feel like the performance and the power draw would be enough for normal people to judge a CPU

Not anymore, hybrid CPU cores and the modern advanced sleep states really throw that kind of thing out the window.

You could see a 16 core 32 thread CPU and go "Oh no, 150W!" when all but one core sleeps and it uses 40W gaming.

 

[nguyen]

Can you separate CPU power consumption into 3 scenario

1. Idle: youtube playback
2. Production: automated run-through of various softwares: Cinebench, office, 7-zip, code complilation, etc.....
3. Gaming

1 and 3 are avg power consumption, meanwhile 2 will be energy expenditure to complete the run-through

 

[Mussels]

Can you separate CPU power consumption into 3 scenario

1. Idle: youtube playback
2. Production: automated run-through of various softwares: Cinebench, office, 7-zip, code complilation, etc.....
3. Gaming

1 and 3 are avg power consumption, meanwhile 2 will be energy expenditure to complete the run-through

Local video playback would be far better, and easier to script/test - and quite possible to throw a relatively short 4K H264 video and force software decoding, too.

 

[W1zzard]

4K H264 video and force software decoding

i dont think anyone uses h264 software decoding in 2021

 

[Taraquin]

I vote for wall -power

 

[The King]

I vote for wall -power

Wall power consumption will vary according to how efficient the PSU used is.
This data will only apply to people who use the exact same PSU used in the test.

 

[Taraquin]

Wall power consumption will vary according to how efficient the PSU used is.
This data will only apply to people who use the exact same PSU used in the test.

Point of wall powerconsumption is to compare consumption between CPUs. TPU uses the same PSU on all test systems, MB can contribute a bit, but you get a general trend. I don't think anyone would expect to get the exact same pwrdraw as TPU unless system us identical. If 12900K uses 20W less than 11900K i CP2077 using same GPU, PSU and similar MB that is what's interesting in my opinion.

 

[Wirko]

Please don't abandon reporting idle power. Anandtech did, for reasons I don't understand. Luckily, it can still be read out from their power vs. time graphs.

 

[The King]

Point of wall powerconsumption is to compare consumption between CPUs. TPU uses the same PSU on all test systems, MB can contribute a bit, but you get a general trend. I don't think anyone would expect to get the exact same pwrdraw as TPU unless system us identical. If 12900K uses 20W less than 11900K i CP2077 using same GPU, PSU and similar MB that is what's interesting in my opinion.

If CPU power consumption is the goal.

Lets say purely as an example the CPU needs 100W from the PSU. But this PSU is not very good and it pulls 130W from the wall.

Does this mean the CPU is consuming 130W of power? This method seems more suited to test PSU power consumption and efficiency than CPU power consumption.

 

[Taraquin]

If CPU power consumption is the goal.

Lets say purely as an example the CPU needs 100W from the PSU. But this PSU is not very good and it pulls 130W from the wall.

Does this mean the CPU is consuming 130W of power? This method seems more suited to test PSU power consumption and efficiency than CPU power consumption.

This tread is about how we test CPU pwr consumption. As long as we control the other variables wallpwr is good at finding the difference in pwrdraw between CPUs, but we don't get an exact number. Total system usage on a normalized system is what matters most in my opinion, but if you prefer CPU pwr alone I respect that. Finding an exact consumption through software is not so interesting I think since different loadscenarios can increase MB consumption etc.