Folding@Home

[ichwar]

I don't get this. someone says I have to solve problems. someone else says the computer does all the work. I think solving these kinds of problems is work. so, which one is it, if I download this program, do I have to do anything other than start it running and walk away from the computer for a few hours?

 

[PhantomChick]

The program does all the work. You download the program, and then it runs in the background (like an antivirus software) while you are at the computer or while you step away from it.

What it does is connect to the Folding@Home server to download a "Work Unit", then it spends a few days performing calculations, and uploads it back to the server when it's done, and downloads a new one. (If you use the GPU client instead of the CPU client, it's apparently MUCH faster!)

I use the CPU client and it doesn't slow my computer down any while I watch streaming videos, play flash games, or work with basic software... although I do disable it for when I'm playing more requirement-heavy games.

 

[Livewire]

The program does all the work. You download the program, and then it runs in the background (like an antivirus software) while you are at the computer or while you step away from it.

What it does is connect to the Folding@Home server to download a "Work Unit", then it spends a few days performing calculations, and uploads it back to the server when it's done, and downloads a new one. (If you use the GPU client instead of the CPU client, it's apparently MUCH faster!)

I use the CPU client and it doesn't slow my computer down any while I watch streaming videos, play flash games, or work with basic software... although I do disable it for when I'm playing more requirement-heavy games.

I found a bit more information to add onto that last part - for Windows, theres 2 main clients: CPU and GPU.

CPU's just as important although PhantomChick is 100% correct - the GPU client is able to complete its work MUCH faster than the CPU client (my CPU client is reporting 16 hours left and it started it 11 hours ago; GPU's reporting 2 hours left, started it 2 hours ago), but the CPU client is better at some of the calculations Folding@Home has to have it do.

F@H countered that by making the GPU client work on different projects than the cpu client (for instance, my cpu client's working on project 4445, gpu's on 5754 - no idea what the differences are, but that's a biiiiiig number jump, and I've never seen my GPU do anything outside of the 5000 series).

None of this is really "need-to-know" information I guess, but both clients are just as useful for F@H - if you don't mind using a beta client, go for GPU. If you want it to never fail unless it's F@H's fault, go for CPU




That being said, GPU is the -only- client of the two that has difficulties detecting with it's using too many system resources. CPU client runs in idle priority - if I were to go and start a really resource intensive game, the CPU client would probably stop altogether until I was done gaming

The GPU client seems to do the same but only if the game goes fullscreen, and for me at least it can't detect that worth a dang (RCT3 lags noticibly if I don't shut it off first). Might be Vista, might not be, but it was tagged as a Beta client so I'm not surprised it's still got some bugs




For now though the only -real- work you have is downloading it, setting it up (literally "fill in the blank" stuff), and remembering -not- to close it




Oh, minor note: during config it may ask for a Passkey. This is not required, it's just a way to keep your work separate from someone else who has the same name (there's apparently two Livewire's, but I'm the one using a Passkey )

 

[ichwar]

ok. I've got just 1 slow pentium 4 proc. but I'll consider it!

 

[Livewire]

ok. I've got just 1 slow pentium 4 proc. but I'll consider it!

Define slow, cause there are ways to request "deadline-less" work units, which if it's something like a 300-600mhz you might need to do.

If it's not even that slow you'll be fine - the current deadline for a work unit I picked up 16 hours ago is i 62 days 8 hours 42 minutes. They'd prefer it be done within 42 days, but the official one is 2 months to finish it



Shouldn't be a problem on anything 1ghz or better, which I think p4's met

 

[Smith6612]

I don't get this. someone says I have to solve problems. someone else says the computer does all the work. I think solving these kinds of problems is work. so, which one is it, if I download this program, do I have to do anything other than start it running and walk away from the computer for a few hours?

Pretty much, Don't let the computer go to sleep (only switch off the monitor), make sure it has ventilation and let it rip. Keep it connected to the internet if possible as well. If you run the GPU client, just take note that the client will max out your video card's processing power, which means that if your GPU runs warm, it's going to start to get nice and toasty, so keep the chips cold

The program does all the work. You download the program, and then it runs in the background (like an antivirus software) while you are at the computer or while you step away from it.

What it does is connect to the Folding@Home server to download a "Work Unit", then it spends a few days performing calculations, and uploads it back to the server when it's done, and downloads a new one. (If you use the GPU client instead of the CPU client, it's apparently MUCH faster!)

I use the CPU client and it doesn't slow my computer down any while I watch streaming videos, play flash games, or work with basic software... although I do disable it for when I'm playing more requirement-heavy games.

The reason the GPU client is faster is because typically video cards have TONS of processors on them. My GTX280s have according to the nVidia website has 240 cores of Stream processors. Though despite all of this power, Stream processors can only calculate certain things, but they are very, very efficient and can work quickly. Also keep in mind though, to do GPU folding, I believe you need to have a CUDA enabled card from nVidia with the driver, or you need a card that supports ATi's new CUDA like platform to fold using the GPU. If you have an onboard card, don't even think about it unless it is on-board but dedicated. I will lastly add, the GPUs typically get the larger projects as well. Right now one of my video cards is crunching down data on a project with a data length of 30,000. The CPU's been working on it's third fold for today (would be LOADS faster if the app supported multithreading!). Right now I'm donating some work to another site, but sometime soon I'll switch to the x10Hosting group to give some Folding points in.

So right now, I'm running 3x GPU folds, 3x GPU apps in memory, as well as a single CPU client.

 

[Livewire]

Pretty much, Don't let the computer go to sleep (only switch off the monitor), make sure it has ventilation and let it rip. Keep it connected to the internet if possible as well. If you run the GPU client, just take note that the client will max out your video card's processing power, which means that if your GPU runs warm, it's going to start to get nice and toasty, so keep the chips cold



The reason the GPU client is faster is because typically video cards have TONS of processors on them. My GTX280s have according to the nVidia website has 240 cores of Stream processors. Though despite all of this power, Stream processors can only calculate certain things, but they are very, very efficient and can work quickly. Also keep in mind though, to do GPU folding, I believe you need to have a CUDA enabled card from nVidia with the driver, or you need a card that supports ATi's new CUDA like platform to fold using the GPU. If you have an onboard card, don't even think about it unless it is on-board but dedicated. I will lastly add, the GPUs typically get the larger projects as well. Right now one of my video cards is crunching down data on a project with a data length of 30,000. The CPU's been working on it's third fold for today (would be LOADS faster if the app supported multithreading!). Right now I'm donating some work to another site, but sometime soon I'll switch to the x10Hosting group to give some Folding points in.

So right now, I'm running 3x GPU folds, 3x GPU apps in memory, as well as a single CPU client.

Query for ya: Multicore cpu handy?

http://folding.stanford.edu/English/FAQ-SMP

Sounds like a bit of a nightmare, but requirements are a recent Dual Core or better (tri-core, quad-core, 8-core, multiple cpu's, etc). Multi-processor processing, might be something to look into.


I know I will be once I upgrade this to a quad core; can run that on 2 cores, leave the GPU alone (it'll use a third cause not everything can go on the GPU right now), and a 4th to run all the other mundane stuff. Sounds good to me

 

[Smith6612]

Query for ya: Multicore cpu handy?

http://folding.stanford.edu/English/FAQ-SMP

Sounds like a bit of a nightmare, but requirements are a recent Dual Core or better (tri-core, quad-core, 8-core, multiple cpu's, etc). Multi-processor processing, might be something to look into.


I know I will be once I upgrade this to a quad core; can run that on 2 cores, leave the GPU alone (it'll use a third cause not everything can go on the GPU right now), and a 4th to run all the other mundane stuff. Sounds good to me

Thanks for the link. I'll probably load that up onto my computer once the current workload my CPU is doing on one out of the 4 cores which are also HyperThreaded.

 

[Livewire]

With any luck I'll get my finances in order soon enough as well to figure that out; the odd part for me is there's 2 different sets of directions for two different setups.

I've no idea what the advantages/disadvantages are, just that it's listed as another High Performance client that can improve all calculations by 4x (GPU does 30x but can't do everything better, and in fact -can- do some things -worse!-).

 

[ichwar]

ok, my graphics card is several years old. it can't even run compiz or beryl on linux, or 3d emulation on google earth, so I doubt it would ever be able to run the gpu client. so I'd have to stick with the cpu. my proc is 3.4 ghz, but surprisingly slow. Maybe it's that my 1gb of ram is always full?

 

[Livewire]

ok, my graphics card is several years old. it can't even run compiz or beryl on linux, or 3d emulation on google earth, so I doubt it would ever be able to run the gpu client. so I'd have to stick with the cpu. my proc is 3.4 ghz, but surprisingly slow. Maybe it's that my 1gb of ram is always full?

Keep in mind that's the one thing Folding@Home won't like - full ram...It'll still work well on the 3.4 as far as getting the calculations done, but it can tend to be a bit of a resource hog if you let it be.


I say give it a shot and if it doesn't work out, erase it - won't hurt anything to give it a shot

 

[Smith6612]

ok, my graphics card is several years old. it can't even run compiz or beryl on linux, or 3d emulation on google earth, so I doubt it would ever be able to run the gpu client. so I'd have to stick with the cpu. my proc is 3.4 ghz, but surprisingly slow. Maybe it's that my 1gb of ram is always full?

If anything, there is a retired client that's probably still floating around that doesn't use new GPU2 capabilities with CUDA and the other thing for ATi, but is an older GPU folding program that uses DirectX. Might want to see if that is still around and still works for you to use on your older card.