For those of us, that are doing distributed computing, mining, crunching, whatever...,
Saving on a power bill is always interesting.
 
To save a little on system power draw, without suffering much of a performance penalty; has anyone done some testing in how low the voltage of DDR3 and DDR4 RAM, as well as CPUs can be lowered, without experiencing errors?
 
And does just lowering voltage in bios on RAM or CPU, really save power?
Or, on a CPU, is perhaps disabling turbo boost, or setting a hard cap on speed, a better solution? Or doesn't it affect power draw at all?
Like, will a (eg) 2,1Ghz CPU, with 3,7 Boost use less energy, when disabling the boost and cap the speed to 2Ghz? Or will it need to go paired with lowering the cpu voltage?

Nope, Max Power 100% of the Time for the Max Completion and Fastest Completion Times.
And as above I need to Heat my home in the Winter.
Crunching@EVGA A Fellowship of All Walks of Life Looking to Improve Our Communities' Future One Byte at a Time...
 
And No Mining under Our Team but you can ask the Miners under here Cryptocurrency
They are all about Saving Power Costs and Reducing the Power of all their Hardware.

I envy you guys in that!
In the south, it's always too hot. Even in the winter, the heat generated by my GPUs is just blown outside.
But surely, those of you who crunch/fold/whatever in summer, will want to find ways to save some $$?
Besides, if you can save watts (say 200W) on larger (1000W) systems, would mean you could add an additional small system to increase performance?

It's an interesting idea to under clock/volt the CPU and memory, especially during the summer months.  I haven't gone down that road (yet). 
I do overclock my CPU during the cooler months, and bump it back to stock when it gets hot outside.  I also cut back my GPU usage in the summer.
As a way to save on the power bill while still crunching 24/7, I focus on having efficient PSUs (80 Plus Gold or better).

Crunching apps that we run need the Ghz and fast memory access. Lowering power just crimps the crunching. I've given up overclocking my systems and just let them run at stick on platinum and titanium power supplies. The Watts saved by a few millivolts just don't add up enough in my opinion.
Most CPUs are power efficient without an OC, only extreme OCs really get them at really high wattage.
Speed boost and Multicore enhancements only add a little bit to the power budget.
Same for GPUs, buy midrange if you're concerned about efficiency and power draw. I'm not convinced it is a good idea to power starve a GPU long term.

ipkha
Crunching apps that we run need the Ghz and fast memory access. 
Same for GPUs, buy midrange if you're concerned about efficiency and power draw. I'm not convinced it is a good idea to power starve a GPU long term.

For 24/7 GPU crunching, the results are obvious. Power capping and overclocking really are the way to go!
Performance is reduced by only a few percent, heat is reduced by tens of degrees C, electric consumption is down by >20% (for large server farms with 8+ GPUs running at 1500W, this means I can simply add 1 or 2 GPUs to the mains, consuming the same amount of overall wall power).
I have been running some of my GPUs in this state for almost a year, 24/7, without ill effects. I actually believe it's better for GPUs to do this, than running them at stock power draw levels, due to reduced heat.
 
For GPU crunching, power limiting a cpu, could result in meagre power savings, in the likes of 5-10Watts (less if they are modern CPUs), but the speed reduction can cause larger GPU speed drops of 10+%.
So definitely not recommend on GPU farms.
I will yet have to test to see how much cutting turbo boost frequencies on the CPU, will affect power consumption ,and how much it affects folding speed, in a case scenario where the CPU>3Ghz without boost (3Ghz of cpu speed, is the minimum needed for RTX2080Tis, 2Ghz for RTX2060s).
 
But in terms of crunching on cpu, I'm currently running boinc, Rosetta, on CPU. I'll need a few weeks of data, before I'll be able to answer my own question here.

My username is the same as here. I just started, so there's not a lot of data yet. Plus, I started on a netbook, tried out the Android client (without success), momentarily played with the idea of running boinc on my raspberry pis, but realizing I had a 2 GHz 10core 20 Thread Xeon laying around in the living room, I tried out the windows client, and quickly jumped over to Linux.
My system just got active since last night.

I noticed using newer CPUs, x99,z370 and so on, I noticed they do not throw off as much heat as the older Sandy and older CPUs
So now my electric bill is as much or more and not getting the heat I used to. Kind of defeats the who purpose of what I want to do.
 
Why I OC to get things warmed up.
 

bill1024
I noticed using newer CPUs, x99,z370 and so on, I noticed they do not throw off as much heat as the older Sandy and older CPUs
So now my electric bill is as much or more and not getting the heat I used to. Kind of defeats the who purpose of what I want to do.
 
Why I OC to get things warmed up.

Same Here

Efficiency is allowing you to get more work done, for the same amount of heat. Reducing the heat (and therefor the electric bill) allows you to add to your hardware, and produce more work.
The heat over here is a byproduct I don't need.
I could only wished I lived in colder regions!

The world is not waiting..
If you can do 20% more work, by changing a few settings, why not?
Most people don't do it, because they don't know how, think it might be a lengthy, difficult, or dangerous procedure, but it really isn't.

We're only here for another few decades at most, before we have to pass this information along to another generation, who might, or might not be equipped to deal with that information.
The more we can get done today, the quicker we'll have answers, and the less people will have to die meaninglessly to some foreign diseases.
 
There are a lot of projects where crunching speed doesn't matter. Their outcome doesn't change much (eg: find new prime numbers, math games, etc...).
 
Science, and on the medical field, every finished task/wu/batch ahead of time matters!
Having computers understand and being able to predict proteins, find cures on their own, matters!
We're still far from there, but I feel that doing as much as we can in these areas, matter!
 
It's not about a virtual score; the reward is in the real world.

bill1024
 
One thing you can do with the heat if you do not need it. Make a water loop and have the radiator outside the house.
Air to air or bury some pipe underground and do ground source cooling of the loop.

Or you can install a hybrid heat pump water heater that will absorb the heat you paid to create & supply you with hot water. There are ducted versions as well.

My input:
I run CPU at maximum stable overclock and GPU I will run at stock clock and lower the power target to keep temps down.
I see a much bigger performance advantage vs. heat/power usage with CPU overclocking.
 
Getting a good OC takes time and careful testing, but heat management is what it always comes down to for me.

I've posted this thread for several weeks but couldn't get any satisfactory answer. Hoping that the DCers here can help. I'm trying to do extreme undervolting on my EVGA RTX 3070 Ti XC Ultra to save power bill. I've EVGA RTX 3060 Ti FTW and EVGA RTX 3080 Ti XC hybrid that I can easily undervolt to around 0.75 to 0.80V for optimum power crunching efficiency in terms of PPD/(gpu watt). 
 
I'm wondering how low can the gpu voltage on other 3070 Ti models can be adjusted. I think my 3070 Ti XC Ultra gpu voltage seems high even when running at very low type of load (<40% gpu power limit). The best way to test this out is to use MSIAB and edit the frequency voltage curve as I have shown in the other thread. I think this is very likely to be vbios limitation but wondering if other cards have much less aggressive vbios in term of gpu voltage vs frequency performance. Thanks!