Recommend Energy Efficient Desktop CPU

[Scopus]

Just curious, what makes you want/need less power draw than a r5 1600 or i5 8400? They both use so little power that even cutting it in half would never be noticeable on your power bill.
My 1600 hastily OC'd at 3850mhz/1.4v, is sitting at 45 C with a decent cooler, can't even hear the fan, and that's under load!

Planning to get 8700k in a clevo laptop, which due to inadequate cooling can overheat. Even at stock settings high chance of running too hot. On previous gen kaby lake many people report throttling during testing unless delidded and such. And i want it to be silent which im thinking got to lower the tdp way down to do.

 

[Zucker2k]

Planning to get 8700k in a clevo laptop, which due to inadequate cooling can overheat. Even at stock settings high chance of running too hot. On previous gen kaby lake many people report throttling during testing unless delidded and such. And i want it to be silent which im thinking got to lower the tdp way down to do.

8700k in a laptop should be bare die already so no need foe delidding.

 

[Scopus]

8700k in a laptop should be bare die already so no need foe delidding.

These laptops use desktop cpus

 

[Insert_Nickname]

I'm curious what people are pulling from the wall on their PCs? The system in my signature, along with my 25" LED (and four case fans) pull right around 60-70w at desktop using a balanced profile. I understand that lower power CPUs can pull less, but I don't think it will really amount to much savings, right? I mean, my CPU down-clocks itself to 800 MHz while doing this. I'd think having a very energy efficient (platinum or titanium) would save more power than a lower performance CPU?

Very much this. The first step to an energy efficient PC is the PSU. Unless you already have a gold rated unit, that's the first thing one should consider changing. Another thing is you shouldn't go overboard on wattage. PSUs perform best if loaded to about 50%, and modern PCs use surprisingly little power.

I mean, I know Intel has 35w CPUs, but you give up performance for energy savings, but I guess I just want to understand what the energy difference really will be?

The T series are more about when cooling limited then power savings. Sure you'd save some power running one, but if it has to run longer to finish a task the net gain will be zero.

 

[wacki]

.... Not sure an R5 or R7 is necessary here.

I will say that the R7 will probably be most efficient due to quicker race to idle.

What do you mean race to idle? It simply gets the job done quicker while using the same amount of electricity?

 

[LTC8K6]

The perpetual argument over power consumption.

Does the faster chip really use more power since it gets the job done much quicker and is already back at idle, while the "efficient" chip is still consuming it's max power trying to finish the job.

 

[Topweasel]

The perpetual argument over power consumption.

Does the faster chip really use more power since it gets the job done much quicker and is already back at idle, while the "efficient" chip is still consuming it's max power trying to finish the job.

Consider video encoding. It's a long job. An R7 1700 will maintain it 3.3GHz. A 1800x 3.7GHZ. Roughly a 10% difference at a 30% increase in power usage. The 1700 so let's say takes 33 minutes (assuming linear performance) vs. 30 minutes for the 1800x. Let's say for the 3 minutes it uses only 10w. The 1700 uses 35whr of energy to do the job, the 1800x would use 45whr plus .5whr for the same 33 minutes of measurement.

This is rough napkin math with tons of holes. But even giving the 1800x the benefit of linear performance increase due to clocks and cores (if my example had a core increase). No just getting done sooner doesn't help if the time it completes the job isn't quicker than the increase in power usage.

This applies on the simple jobs as well. Quick to finish can work. But not if the percentage of power increase exceeds the percentage of time saved.

Which is why more cores matter more when doing video encoding. When you have 65w 6 and 8 core CPU's, it's going to take a lot of power savings on a 2 or even 4 core CPU to keep up. A 1700 is really a masterpiece of productivity 8 cores with such lower power usage. Not that any of this really matters. A 1800x/8700k at 90w constant load 24/7 equals 2.1 KwHr a day, 15.1KwHr a week, 63KwHr a month. That comes out to (at my price of $.23 a KwHr) $14.5 a month to run that CPU at 100%. vs. $10.5 for a 65w CPU. Not only is the cost of running the two at 100% really not that big of a thing (least for me in Mi) but there is a point that the extra 10% it's getting done running 24/7 might be worth the extra $4. On top of that, we are talking about running 24/7. If we are talking about sporadic or 8hour work day only the difference really disappears as far as cost is concerned.

 

[LTC8K6]

Well, I'm glad I only pay 10 cents per kwh...

 

[Topweasel]

Well, I'm glad I only pay 10 cents per kwh...

Yeah mean you could cut all of those numbers in half. But I mean even if you go up to $.50 a kwh, it's still not a terrible number. I am sure for a lot of people every cent counts but there are a lot of ways to save some money each month that doesn't involving getting a slower CPU and wasting a bunch of time. Even if it's just home projects when you look at the times savings in the long run is it really worth saving the 2-3-4 bucks a month. Replace lights with LED's. Get a HE washer. Pre clean the dishes and use a colder water setting on the dishwasher. Get thermometer that you schedule when it is on. Get a newer LCD TV specially if you are using a Plasma.

Hell I got LED lights to replace all of my lights because I liked them better and they last longer. But I have a 120w light using a 20w LED. I can run my CPU at 100% for nearly 2 hours for the savings I get when running that guy for an 1 hour.

 

[maddogmcgee]

If your going to be using it all day everyday and doing serious work on it, a laptop with a desktop CPU is likely to be a horrible choice. It will get hot, which means the battery will die very quickly. It will likely use a not terribly efficient PSU which means it will use more power regardless of your CPU selection. It will also be harder to repair if something goes wrong. It will also be big enough its not really portable. As others have suggested, just get a 1600/1700 (or 8700 if you are wedded ton Intel) with a desktop case. Cooling will cease to be a problem.

 

[DrMrLordX]

What do you mean race to idle? It simply gets the job done quicker while using the same amount of electricity?

Ideally, if you have an "embarassingly parallel" workload, you get more efficiency out of the system that manages to host more cores in one socket, if all other factors remain the same.

No matter which CPU you use in an AM4 system, you are always going to have some baseline power consumption from the platform itself. Then add to that the video card, RAM, storage subsystem, etc. So when you choose fewer cores, the CPU power consumption will be lower, but that savings will be offset somewhat by total system power draw, of which CPU is only one part.

Furthermore, with any modern process, you will probably get the higher-core-count CPU running at a lower clockspeed where the process is closer to an ideal point in the voltage curve. Unless you buy a real discount quad or whatever, that comes in at a low clock.

Bottom line is that if you compare the total joules used by the system to complete a given encoding job, you will find that something like the R7 1700 will be more efficient than any of the R3s.

This applies on the simple jobs as well. Quick to finish can work. But not if the percentage of power increase exceeds the percentage of time saved.

Correct. What we are looking at is average power draw vs. total time to completion, which yields a total number of joules of electrical energy used to complete the task. In the case of the R7 1700 vs 1800x, the 1800x at stock clocks suffers from sitting in a less-efficient part of the 14nm LPP voltage curve. You burn too much power to get that extra 400 MHz for it to be "worth it" from a raw efficiency perspective.

But if you start comparing R3s to the same R7 1700 . . . well, see above.

 

[thecoolnessrune]

Depending on the Desktop setup, an idle machine is going to be less than a monitor. My old Dell U2711 that I use here consumes 110 Watts, which in and of itself is more power than my Ryzen 1700X + 1080 Ti Rig consumes at idle

 

[mikeymikec]

Depending on the Desktop setup, an idle machine is going to be less than a monitor. My old Dell U2711 that I use here consumes 110 Watts, which in and of itself is more power than my Ryzen 1700X + 1080 Ti Rig consumes at idle

Huh. My 21" monitor states 21W typical usage in its specs.

 

[thecoolnessrune]

Huh. My 21" monitor states 21W typical usage in its specs.

I'd believe that, especially if it's an Edge-Lit LED LCD monitor. A smaller Edge lit panel is going to be a hell of a lot more efficient than a large monitor using CCFLs.

 

[coercitiv]

I'd believe that, especially if it's an Edge-Lit LED LCD monitor. A smaller Edge lit panel is going to be a hell of a lot more efficient than a large monitor using CCFLs.

Yup, power usage for the new models has come down significantly even if we consider same screen size:

• 24" from 2007 - 57W typical / 110W max (CCFL lit, 1920x1200)
  
• 24" from 2017 - 20W typical / 62W max (LED lit, 2560x1440)