Cheapest rig for Distributed computing


Aug 27, 2014
There are a few Distributed computing programs out there.
One being Folding At Home, the other being Boinc.
Folding at home is a program that will analyze proteins and folds them. It needs tens to hundreds of thousands of computing hours, just to display a fragment of a video of only a few milliseconds long, for scientists to study.
They have a variety of diseases they study.

The other is Boinc.
Boinc is a program that runs a variety of projects, that can range from solving Mathematical problems, studying science related issues, AI, or biology, or study weather patterns, among a few...

Folding at home works best with the fastest GPUs.
If you want a fast rig? Just make sure it's relatively modern, and is equipped with at least 1x FAST GPU. And I mean FAST!
There are people running 6 to 9x RTX 2080Ti or RTX TITAN GPUs, considered one of the fastest GPUs on the market today.
You don't need to cash out that much though. With a single GTX 1650 Super, a 1660 or 1660 Ti, you can already significantly contribute!
I would not go lower than these, as FAH really is based on Quick Return Bonus, meaning, the faster the hardware, disproportionally the higher your PPD scores will be.

For Boinc, speed isn't as much of an issue, because you won't get any extra points for faster hardware.
That leaves the end user a lot more freedom to choose what hardware would work best.
For the past few years, most projects were depending on GTX 1050 to GTX 1060 hardware for GPUs, and at least a quad core CPU.
Today however, you can easily buy an AMD Ryzen 9 3000 series CPU, with 8, 12, or 16 cores + an additional threads.
The extra cores really help speed up processing the data!
To be competitive in the CPU market, you really need at least an octacore. However, currently, if you're looking for a sweet spot, it would probably be a Ryzen 3 3900X. Buying 2 entire systems (with motherboard, RAM, and watercooling) costs about the same as a Threadripper 3900 series CPU of the same amount of threads.
GPU wise, most projects now focus their WUs on an RTX 2060 to RTX 2070.
Faster is better, but you'll have some configuration manipulation to do, to make them run as efficient as possible.

GPUs will always crunch data faster than CPUs.
Even the fastest Threadripper CPU, is only capable of processing data about as fast as a budget GTX 1050, or 1060 GPU.
For that reason, most people support Boinc with GPUs.

Projects that give the highest points scores, are mathematic projects, like GPUGrid, PrimeGrid, and The Collatz Conjecture.
These projects are for "PPD hunters". They don't really contribute much to science.
Projects that quite often offer the most contribution to Science, are the CPU projects.
If your aim is a high position on the PPD scoreboard, go for GPUs.
If your aim is a higher contribution to science that matters, spend a little more to the CPU instead.

"So why not AMD?", you may ask?
AMD is still a great alternative, and affordable.
AMD is best for when initial purchase cost of the GPU is important for you.
And when your contributions will be limited (eg: a few hours a week).
However, if you want to support GPU projects continuously (as much 24/7 as possible), Nvidia is the way to go; simply because the added cost of Nvidia's initial purchase, is easily offset by the electric cost to run it.
Nvidia GPUs consume on average 10-20% less power than AMD, and can be tuned to run between 20-40% even lower power.
Meaning, for nearly the same performance, you could be consuming about half the power of an AMD GPU.

Lastly, for Boinc, there are the cheapest systems overall.
If you're looking for contributing to Boinc, and you're 'as cheap as it goes', look online for an Atomic Pi, or a refurbed or second hand laptop or desktop. Or, alternatively, look for AMLOGIC TV boxes (Specific the S905X3, which can be found for about $35 per unit; or the S922x which can be found for about $80).
Anything going from between $30-$99 is a great budget to start folding/crunching data.

And once you get more serious with folding, use the cheap system to install linux on, and test settings with, which you can then transport to your more expensive system.

Good luck!
Thread starter Similar threads Forum Replies Date
P Distributed Computing 1