Folding Forum

how do i force folding@home to download the smallest work units available and do only those for cpu and gpu? smallest only, no big ones taking days or week+

Not possible .. sorry

Neil-B wrote:Not possible .. sorry

why can it not be made possible?

At this moment in time this is not how fah works ... it is not programmed that way and there are not settings to allow this to happen ... A variety of discussions and possible developments are being had to try and assign work more appropriately to GPUs but at this time it isn't a feature that is available.

It may change in the future (no guarantee though) and if it does then until such time as it happens it is unlikely to be announced or given an eta as yet again that is not how FaH tends to work.

In the deep past, there was an option* to request small Work Units in the sense that the modem would download a small amount of data. This is not the same as taking a small amount of time.

While the option still exists, it is unlikely that any researchers are still catering to folders with dial up modems, nor is it any guaranty that it would result in short run times if they still existed.

To do what you wish, someone would have to run all the WUs, then re-release the ones that took the least time, even though they were already completed.

*You enter the parameter max-packet-size and set the value to SMALL.

Just wondering, what's your use-case so that we can see when we can do about it

FAMAS wrote:how do i force folding@home to download the smallest work units available and do only those for cpu and gpu? smallest only, no big ones taking days or week+

Thats absolutely something I would wish to be implemented.

PantherX wrote:Just wondering, what's your use-case so that we can see when we can do about it

I do not know why OP needs this functionality I can just talk for me.
I am not folding 24/7 but would like to use the time I am working on my machine to make my small contribution. So therefore it is nice to get a WU that earns me 10k Points, but since I am just a CPU folder on my Notebook, such WU takes me 10 hours to be completed. When I know I have just 1-2 Hours in a train or at home or whatever I would like to assign to a small Unit I can finish without beeing forced to interrupt.

This is one of my 3 wishes I would love to see in the FAC controller

Before COVID-19, CPU WUs would be calculated to meet the timeout by folding 8 hours a day. I am aware that COVID-19 Projects have shortened the timeout due to time constraints but hopefully, Projects will be back to normal soon.

While not on CPU, I do know that GPUs will be moving towards a much better WU allocation system that is being developed. In short, large WUs will be for powerful GPUs while small WUs will be for low-end GPUs. Other aspects needs to be considered too (part-time folders, etc.) but hopefully, it can be addressed sooner rather than later.

Personally, I won't fold during commute as folding is very CPU intensive which causes the battery to drain very quickly. YMMV but last I tried, the battery lasted for 20 minutes.

PantherX wrote:Before COVID-19, CPU WUs would be calculated to meet the timeout by folding 8 hours a day. I am aware that COVID-19 Projects have shortened the timeout due to time constraints but hopefully, Projects will be back to normal soon.

IF the avaialble projects use this formula and are estimated to take 16 hrs of continuous processing on "typical" hardware, that predicts the deadline should be set to 2 days. If your hardware is 33% as fast as "typical" you'll have to leave FAH processing 24x7. If your hardware is faster than that, you'll be able to interrupt processing for appropriate periods of time.

If a project is estimated to take 7 days and 8 hrs per day (or 56hrs continuous), the deadline can be set to 7 days.

If a project can be completed in 2 hrs on "typical" hardware, should we set the deadline to 0.25 days?

If your hardware is appreciably shower that "typical" hardware, it's likely that you'll still have issues, but as I said eariler, FAH needs to record measurable progress before it can reliably inform you of that fact.

Some of my kits have slow hardware. Even my GT710 is fast enought to complete work by the deadline and even a CPU:1 slot is able to complete reasonable assignments. Maybe you need to fold more hours per day.

The aforementioned project to realign assignments with GPU benchmarked data will be very helpful but that didn't become a written suggestion until recently and it takes a while to assemble benchmark data for all GPUs.

I would like control over the work unit size because I use a MacBook Pro to crunch on these work units, but I can't leave it running 24/7 without having things overheat, and having the battery swell. I run it a few hours for a few days a week. I keep getting work units that will take an i7 more than 24 hours to complete, and not enough time to do it given the limited number of hours and days I run FAH. Either give us an option for smaller work units or more time to do them.

Snap from another thread:

gunarre says
"There has been an advanced option called "max-packet-size" which could be set to "small", "normal" or "big""

I don't know if that still works and solves your problem, but it's worth a shot. If you head over to FAHControl, go to Configure, then Expert, you could try to add this to the client options.

Another option may be to limit the power the FAH client uses, so that your laptop doesn't overheat. Having your laptop overheat probably isn't very good for its lifespan.
Most regular laptops don't deal well with GPU folding. The client really stresses the hardware, and laptop cooling generally isn't up to that. Gaming laptops may be better equiped. There is no way in the client to limit the stress it puts on the GPU, it's either on (power slider on medium or full) or off (power slider on light). Some graphics drivers have a way to limit the power on the GPU though, but that doesn't sound something like Apple would built in.
You can manually set the number of threads it uses on the CPU (in FAHControl), so you can at least tune that, separate from the power slider setting.

ScottDore wrote:I can't leave it running 24/7 without having things overheat, and having the battery swell.

That shouldn't be happening. I wouldn't use - not to mention fold - on any device where the battery is swelling. It could lead to a "fiery but mostly peaceful" end to your house and your life. If you are folding on an old MacBook Pro with a removable battery, you could remove and recycle the battery, and get a new one. You can use the machine without a battery - just plugged in - if you don't want to invest in a new battery. Unibody MacBook Pros have pretty good thermal performance, as the whole case is a heatsink, though their fans might be running loudly on full power.

An overheating laptop might benefit from being cleaned and refurbished (dusting the heat sinks, re-pasting the thermal pads etc.) but isn't an easy job. If you post the first 200 or so lines of your FoldingAtHome log file, or tell us exactly which model your machine is, we might be able to tell you if it's worth the power usage to keep folding on it.

PaulTV wrote:Snap from another thread:

gunarre says
"There has been an advanced option called "max-packet-size" which could be set to "small", "normal" or "big""

I don't know if that still works and solves your problem, but it's worth a shot. If you head over to FAHControl, go to Configure, then Expert, you could try to add this to the client options.

As noted earlier in this thread, the option is still there but it's about the download size of the Work Unit, which isn't linearly related to the processing time it takes to complete.

Actually the limit is supposed to be the upload size as in most cases the upload is larger than the download. Most home links area also asymmetric, slower upload rates than download.

I have been folding on a MacBook Pro much of the time since I purchased it in 2015. It is a mid-2012 model with the 2.9 GHz i7 processor. Most WUs complete in under 18 hours, and I pause others as needed, I did replace the battery a few months ago, it was swollen from being 5+ years old, not necessarily from folding. Cleaning out some accumulated dust did help with temperatures a bit.

Several times I have received very long WU that timeout after several days of work. Could you at least give more time to finish a WU instead of trashing work and sending another WU? The first WU then needs to be restarted from the beginning by someone else and wait again. In boinc the work times for seti were long, maybe a month. When I processed malaria in boinc the times were short because they were trying to use the data proactively. Folding data are not that time sensitive, so why not just give lots of time to finish a WU?