Status of DC and its worth?

[Exascaletech]

Was big into DC a long time ago and looking at maybe trying it again. Four quick questions come to mind I need answered.
1. I recall reading a letter from one of the BOINC founders a few years ago and him stating in essence what a failure or tons of issues that really reflected bad on the effort. I'll try to find the letter but anyone seriously into DC should know about it. I visited UC-Berkeley when I was running numerous systems for them. This has caused me to pause my interest.
2. The question was always there if we were really doing any good. Yeah, its all baby steps and no real breakthrough ever happened when I was part of it. Some would say it was anti-productive and a waste of resources/electricity and the DC community was really reduced to just ego's and rich players there only for the points. I saw much of that.
3. Part of this quest is to see what if any project is worthwhile to me. I mainly like medical or help the under-served world issues. BOINC and WCG come to mind.
4. Only after figuring these things out for me, I'll test out the project and if I like it, look to add used computers where it makes sense or build a dedicated one. A quick question of where to look for resources/forums and at what level cpu is it just a waste of energy due to it being extremely inefficient for the amount of electricity used?

Maybe I should have asked each question individually but having done this at a high level in the past, I'm sure some one can hit the mark.
Thanks

 

[Skillz]

Was big into DC a long time ago and looking at maybe trying it again. Four quick questions come to mind I need answered.
1. I recall reading a letter from one of the BOINC founders a few years ago and him stating in essence what a failure or tons of issues that really reflected bad on the effort. I'll try to find the letter but anyone seriously into DC should know about it. I visited UC-Berkeley when I was running numerous systems for them. This has caused me to pause my interest.

Most of the "failure" that he speaks about is due to the widespread use of BOINC. It's still such as niche hobby to enthusiasts and not a lot of people do it. It's not just BOINC though, it's distributed computing in general.

2. The question was always there if we were really doing any good. Yeah, its all baby steps and no real breakthrough ever happened when I was part of it. Some would say it was anti-productive and a waste of resources/electricity and the DC community was really reduced to just ego's and rich players there only for the points. I saw much of that.

Some projects you actually get credited for your host doing something significant such as finding a Prime number (PrimeGrid, Private GFN), discovering a Pulsar (Einstein@home) and some others I can't think of off the top of my head. Though most projects don't have individual recognitions such as that and finding their publications for what BOINC helped them with can be time consuming. I'd recommend just finding a project you are interested in and looking around. Most of them post their publications somewhere on the site.

3. Part of this quest is to see what if any project is worthwhile to me. I mainly like medical or help the under-served world issues. BOINC and WCG come to mind.

BOINC isn't a project. BOINC is the software.

WCG is a project short for World Community Grid.
Rosetta@Home another medical project.
Denis@home is another medical project.

You can see a decently updated list of BOINC projects here:

Choosing BOINC projects

boinc.berkeley.edu

4. Only after figuring these things out for me, I'll test out the project and if I like it, look to add used computers where it makes sense or build a dedicated one. A quick question of where to look for resources/forums and at what level cpu is it just a waste of energy due to it being extremely inefficient for the amount of electricity used?

Not all projects use both CPU and GPU.

For the projects that do use GPU, in most cases the GPU is going to "do more work" with the same amount of power as GPUs can be vastly superior to CPUs in some work loads. Asteroids@home and Numberfields@home are two exceptions where the work done by a GPU is not quite as efficient as the work done on CPUs. Though the Asteroid@home GPU app is being actively worked on so it can be a lot more efficient. No guarantee on that.

Some projects are CPU only though so you don't really have a choice on what hardware you want to run. CPU or nothing.

Maybe I should have asked each question individually but having done this at a high level in the past, I'm sure some one can hit the mark.
Thanks

 

[Exascaletech]

Outstanding answer. Thanks, this gives me a good start. The number of cpu cores available have changed a lot since I was last involved. Are most projects up to date in using higher end cores or not? In the past, many were not even taking advantage of 64-bit systems but that seems to have been fixed.

 

[Skillz]

Outstanding answer. Thanks, this gives me a good start. The number of cpu cores available have changed a lot since I was last involved. Are most projects up to date in using higher end cores or not? In the past, many were not even taking advantage of 64-bit systems but that seems to have been fixed.

Not all projects are "multi threaded" per se, but the BOINC client allows you to run many tasks concurrently on any number of physical or logical cores you may have on the system. As well as being able to run on GPU, CPU or both.

 

[Endgame124]

As for efficiency, some of the projects, like asteroids, will run on ARM - meaning you can run them on a raspberry pi. A single dedicated and headless (no attached monitor or USB devices) Pi 5 will consume around 4 watts and do proportionally as much work per watt as some of the higher end consumer hardware.

 

[Exascaletech]

Had no idea about the raspberry pi being so efficient. Can't wait to see how that is possible on 4W. Had been looking at energy efficient 12 or 16 core. Thanks.

 

[Endgame124]

Had no idea about the raspberry pi being so efficient. Can't wait to see how that is possible on 4W. Had been looking at energy efficient 12 or 16 core. Thanks.

I did a lot of testing and tweaking with the pi 4 in 2020. It is basically as efficient as an AMD 5950X in points per watt (but a lot fewer points per day). The pi 5 is more efficient than the pi 4, but the 7950x is more efficient than a 5950X.

I started a pi thread here, though I never really got it as nice as I wanted it to be.

The Raspberry Pi Thread

Table of Contents and Change Log Why use a Raspberry pi? -Overview -Pi Versions -Resources / Links Pi OS and installation. -Preferred OS for Boinc only -Other OS -Tweaking Raspbian Lite for DC use Pi Tweaks and Configuration. -Shutting down non essential hardware -Over / Under clocking...

forums.anandtech.com

 

[Exascaletech]

What a nice article. You put some time into that. Thanks for the effort. Didn't finish it yet, but will. I'm looking for a high end system. In short can you list what would be the crazy top performer set up in Pi? I was just earlier today looking at cpu's in the 1k range as a dedicated system and the whole MB issue and saw how silly the wattage was going to be.. Never would have thought of this.

 

[Skillz]

I've got half a dozen RPi4. Cool little things, but if I wanted to do as much work as the 7x 64 core EPYC setups I have it would be a management nightmare trying to manage 200+ RPIs. But that alone wouldn't even account for all the GPUs I have as well.

Still didn't stop me from buying the RPis and building a little cluster with them however.

Best use I've found for the RPi so far is running the Pi-Hole DNS setup on one of them and blocking ads on my whole network.

 

[Exascaletech]

I use a public DNS server through my vpn which has worked extremely well so far and I'm 100% Linux. Would be interested in a quick detail of your Pi DNS set up. Is there a site that compares on a rough scale how cpu's compare to each other on DC? I'm not a points guy, just what big bang for the buck and have zero interest in gaming so I can see a lot more decisions in my future than I originally planned. Looks like BOINC has a few projects I like and will look also into Rosetta and Denis as you mentioned so I'm making good progress.

 

[Skillz]

You can see more info on pihole here https://pi-hole.net/

I don't know of any sites that compares CPUs or GPUs for BOINC. Other than what people post on forums, discord or other stuff.

You can go to the projects web sites and check the stats of the top computers, but it's time consuming.

 

[Endgame124]

There is a site that compares hardware for folding@home. It’s not Boinc, but it’s still medical research and to a certain degree still shows how things stack up.

Realtime Folding@Home GPU/CPU PPD Database & Tools

folding.lar.systems

I think the “easy” decision for larger than pi 5 hardware right now is the AMD 7950X. It’s top end consumer CPU that will yield quite a bit of work at a solid return per Watt, and you can throw a GPU in it for any project that can use it.

In a few months the 8000 series is due out, and the 8959 will invariably replace the 7950 at the top of the stack.

 

[mmonnin03]

As for efficiency, some of the projects, like asteroids, will run on ARM - meaning you can run them on a raspberry pi. A single dedicated and headless (no attached monitor or USB devices) Pi 5 will consume around 4 watts and do proportionally as much work per watt as some of the higher end consumer hardware.

That has to be 4w per core right? The Pi5 needs a larger 25w power supply than the 4. 4w per core would put it around what a 32t AMD CPU uses.

 

[Endgame124]

That has to be 4w per core right? The Pi5 needs a larger 25w power supply than the 4. 4w per core would put it around what a 32t AMD CPU uses.

I haven’t extensively played with my pi 5 yet, and I haven’t gotten out a killawatt specifically for it, but as long as you under volt it, drop Ethernet to 100mbit, disable the GPU, disable USB, etc, it seems to be very similar to a pi 4 in total power draw. About 4w total.

 

[Exascaletech]

There is a site that compares hardware for folding@home. It’s not Boinc, but it’s still medical research and to a certain degree still shows how things stack up.

Realtime Folding@Home GPU/CPU PPD Database & Tools

folding.lar.systems

I think the “easy” decision for larger than pi 5 hardware right now is the AMD 7950X. It’s top end consumer CPU that will yield quite a bit of work at a solid return per Watt, and you can throw a GPU in it for any project that can use it.

In a few months the 8000 series is due out, and the 8959 will invariably replace the 7950 at the top of the stack.

Thanks for the link, it was exactly what I was looking for. I will put the 7950X on my watch list as its very close to what I'm looking for and the W is reasonable. Sadly, WCG, the best I've looked at so and am running has a questionable financial future as I think many do. Since this rig is going to be dedicated to DC, I need to find another program to look at.

 

[Skillz]

Thanks for the link, it was exactly what I was looking for. I will put the 7950X on my watch list as its very close to what I'm looking for and the W is reasonable. Sadly, WCG, the best I've looked at so and am running has a questionable financial future as I think many do. Since this rig is going to be dedicated to DC, I need to find another program to look at.

DENIS@home - Medical - CPU only - work sent out usually every week, but limited
GPUGrid.net - Medical - GPU only - spotty work sent out
Rosetta@home - Medical - CPU only - Work availability is inconsistent
SiDock@home - Medical - CPU only - Work is usually always available
WCG - Medical + other - CPU mostly, some GPU - Project hasn't been that reliable since IBM sold (or gave) it away

 

[Endgame124]

If I wasn’t clear before, Folding at home is still medical research distributing computing, but it uses its own client instead of Boinc. I personally split my systems between F@H and a variety of astronomy Boinc apps.

 

[Exascaletech]

DC is just one big cycle of funding...research...use volunteers...no money/funding/desire...no research...less volunteers. Sad that if DC is really helping and I believe it is as we're just essentially a mini super computer for them, a better way of doing business would be in everyone's best interest. How that happens I don't know. In theory private enterprises should be more efficient than the federal government but they have not really done much with it compared to what it should/can be for very little cost to the researchers. We are doing the vast majority of their IT, hardware/software costs and research. They analyze the data. It doesn't get much better than that in the real world where volunteers take on the capital costs of running your program and donate their time and still its a reach to find a real program with real results that's not on the edge of failure. Must we do everything for them?

 

[ericlp]

Must we do everything for them?

LOL... Yup, but.... There is ton's of $$$ and volunteer time pointed at the backend... Can you imagine what it takes to sell the IDEA of DC to big corporations that will feed this data and split it up so you an I can download and work on it, and reassemble it again to make sense? And do this all reliable so that somehow anyone with a fairly decent PC anywhere on the planet can contribute? The backend of giving you credit and keeping track of all that for the masses would be an achievement in of itself. We are just installing some software, and letting our hardware work on the WU's. There is a ton of work just getting to that stage. After all the work is done ... one has to ask themselves, would they have just been better off renting a super computer to get the work done vs DC. Must "we" do everything for "them" doesn't give the people that are making all this magic happen on millions of PC's and making it run on thousands of different configurations, weather you got a Mac, PC...Running on Linux, Windows, Mac OS.. and what type of hardware M1, Intel, AMD, Invidia...etc...etc.... getting all that to take the full penitential of ones setup is well, a feat in of itself.

I'm thinking once they get a reliable quantum PC working, then DC will probably go away, until quantum tech comes to the masses (witch it will eventually). Quantum will make classical hardware obsolete.... Where it would take us 50 years to do work on a project and a Quantum PC will do it in less than a day.

I'd hate to be the network administrator volunteering to keep all the data backed up, and all the servers online to pump out and collect all the data efficiently as possible. I but that guy is NOT thinking about "WE"... He's thinking about how much more money he will need in the next 6 months to upgrade hardware to keep it all going...

 

[Exascaletech]

LOL... Yup, but.... There is ton's of $$$ and volunteer time pointed at the backend... Can you imagine what it takes to sell the IDEA of DC to big corporations that will feed this data and split it up so you an I can download and work on it, and reassemble it again to make sense? And do this all reliable so that somehow anyone with a fairly decent PC anywhere on the planet can contribute? The backend of giving you credit and keeping track of all that for the masses would be an achievement in of itself. We are just installing some software, and letting our hardware work on the WU's. There is a ton of work just getting to that stage. After all the work is done ... one has to ask themselves, would they have just been better off renting a super computer to get the work done vs DC. Must "we" do everything for "them" doesn't give the people that are making all this magic happen on millions of PC's and making it run on thousands of different configurations, weather you got a Mac, PC...Running on Linux, Windows, Mac OS.. and what type of hardware M1, Intel, AMD, Invidia...etc...etc.... getting all that to take the full penitential of ones setup is well, a feat in of itself.

I'm thinking once they get a reliable quantum PC working, then DC will probably go away, until quantum tech comes to the masses (witch it will eventually). Quantum will make classical hardware obsolete.... Where it would take us 50 years to do work on a project and a Quantum PC will do it in less than a day.

I'd hate to be the network administrator volunteering to keep all the data backed up, and all the servers online to pump out and collect all the data efficiently as possible. I but that guy is NOT thinking about "WE"... He's thinking about how much more money he will need in the next 6 months to upgrade hardware to keep it all going...

Some interesting points for sure. I reflect on the best in-depth article I have ever seen about DC about BOINC in retrospect.

Maybe it just makes more sense to me as I was there in the beginning of DC and I like many others didn't just simply install some software and sit back but bought/ran 5,10,15 computers that were 100% dedicated to DC as it was thought that this would really help find a cure. The vast majority thought that. I'm still looking for any real progress. That said, I'm crunching as we speak but researchers will always show some distant progress as part of the funding. That's the game.

It's the old build it and they will come and no one wants to pay. My point forever was that if DC could show just 1real science practical example of quality of life due to DC that the masses would join. None of that has ever happened as far as I know. F@H was the most promising project when I started but only incremental success ever happened and many thought some of these finds were simply nothing but enticements to keep up the interest.

As long as DC has been around, almost no one even knows about it still. Even Linux users rarely know. And when you do get someone to join, to most, after a year with no "finds", they quit. I think DC could be so much better, but sadly the project forums enthusiasm has been very poor.

 

[Skillz]

It's just not that cut and dry with any type of simulations. Even no results are results in the scientific community. It's also not so much as going to print out a step-by-step guide on "this is how you cure cancer" with these simulations. What they do is allow scientists to better understand how they work, what they do so they can further their understanding of whatever it is they are researching. Which makes it nearly impossible to write papers that volunteers would be able to understand and pinpoint that yes, that batch of work was the key to solving the mystery.

However, their are some projects where it is more cut and dry. Such as discovering large prime numbers. Pretty straight forward there. Discovering types of stars called neutron stars (pulsars) for example.

 

[Exascaletech]

Yes, you are correct that some projects are much simpler to declare success. I base my opinion of research success in the medical field of MD friends and my gf who is a PharmD and worked for a major in research for 5 years. Not the level at which DC is at but understands the concept and 100% understands pharmacology/chemistry.

The point remains that in my 20 on/off years in it, I can't point to 1 single "here look at this moment", I really really wish I could but I can't. If I or anyone could, the masses would join, points or not and we could move ahead at warp speed and better define what works and what doesn't, funding wouldn't be an issue and it could easily be a thing of national pride. That would be my wish.

 

[mmonnin03]

Just have to actually look

Science - GPUGRID

www.gpugrid.net

 

[ericlp]

Well, back in my day, I ran around 10 computers all running DC off of a floppy disk drives. lol Yes, the days of SETI where DC was much stronger I think.

I got really offended when seti would beg for donations in email .... I was like, ok I buy all this hardware, set it all up... and pay for the power to crunch it and spend countless hours of my time babysitting equipment keeping it running and updating software ...etc...etc... and they want me to donate $$$ ??? But now that I think about it, I probably should have donated more to them and I laugh about it when I reflect back on it.

However millions of people participated in the search not one alien was found, so that is even more disappointing. However, I think if we did make contact, just like if one day, AI searching though all the DC data somehow unravels and comes up with some sort of cure.... Wouldn't you think, that would be thanks to us all for doing the work? Even if the work that you did do, didn't produce anything?

Sometimes like batman, he is doing all the work, without much recognition. I guess the unsung hero. I suppose it's just a moral thing to do.

 

[Exascaletech]

SETI, yeah I remember the start of that and a big push was from Art Bell whom I met at his house. Wasn't interested in finding aliens, my neighbors are easily located. Have you read "BOINC in review", if not, you going back so far will appreciate the time warp in the read. Doing a lot of investing, I'm so used to the 99% hype/marketing in almost everything so when I see a project tout x million hours of computing time and then follow on with.........nothing as far as proof or results and having been in this game as long as I have and seeing DC dwindle as much as it has it makes one pause. Still think just one good find and the flood gates will open but we never got there. Just think what 1 national evening news broadcast telling of a find and how everyone could help. We could do the last 20 years of crunching in 2 months, funding/support issues would be over. I can get excited about that. This is something that shouldn't be a red/blue, old/young, issue that everyone should agree on and almost everyone has a device that could help. Glad you can keep the moral right thing attitude, I do that with Int'l charity work but finding it an uphill battle to build a super machine dedicated even with my stock profit fun money.