• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Folding multiple machines, same WU

T

Teecee

Gawd
2FA
Joined
May 31, 2005
Messages
948
I have started folding again but this time I have a dual core box and on more than one machine. When I use EM3 to monitor all of folding computers, it shows the same protein being worked on by all three machines/CPUs. Is that correct?
 
If they all happened to pull the same WU, that is correct. You can manually check each box to verify if you want.
 
There are many different iterations of the same protein (which is what EMIII shows you)

For example I have a dual core machine doing this

Core 1: Project: 3040 (Run 13, Clone 301, Gen 7) which is part of protein p3040_supervillin-03

2) Project: 3040 (Run 13, Clone 308, Gen 7) which is also part of protein p3040_supervillin-03

Notice that while everything else is different the Clone #s are different which means they are different iterations and thus different actual WU. EMIII only shows the p3040_supervillin-03 part so it looks like they are the same when in fact they are more than likely actually not the same WU. I believe (and this is just from reading the stanford forums) that they actually assign the same WU multiple times to make sure that the results that they get are the same. While this could happen it would seem extremly rare that you would actually get the same WU exactly at the same time
 
Also, what if I want to fold on my GPU, can I do that along with my CPU folding? Lets say I have a dual core and folding with machine ID 1 and 2. Should I fold with machine ID 3 on the GPU folding or is GPU folding completely separate from the CPU and I should start at machine ID 1?
 
I noticed this one time when I'm setting up a folding farm in a classroom (Don't worry, i'm the tech admin there so I got the permission:D ), that when I booted all within 20 mins, they all got the same protein. I didn't check about the details like clone # but I think they all will differ.
 
Teecee said:
How do I determine which Clone # each CPU or machine is on?
Click to expand...

Teecee said:
Also, what if I want to fold on my GPU, can I do that along with my CPU folding? Lets say I have a dual core and folding with machine ID 1 and 2. Should I fold with machine ID 3 on the GPU folding or is GPU folding completely separate from the CPU and I should start at machine ID 1?
Click to expand...

You can determine which Clone # etc of the protein by looking at the Fahlog.txt in the directory that you have F@H installed in.

If you want to fold on your GPU you will need to basically stop folding on one core as the GPU client pretty well requires a core to itself to run efficently. So you would use this arrangement

Core 1 - Standard F@H Client
Core 2 - Assigned to the GPU client

I have never run the GPU client as I dont have any ATI video cards but this is my understanding of the situation. I am sure I will be corrected if I am wrong. Might want to wait and/or ask Tigerbiten as he has run the GPU client extensively
 
EMIII shows the (run, clone, gen) in the single & dual protien veiw, not 3 or above in veiw.

Otherwise look in your FAHlog.txt file.
You should have a line similar to "[12:03:28] Project: 2652 (Run 0, Clone 213, Gen 3)" at the start of each protien.

The only time you will get the same protien with the same (run, clone, gen) is if you get an Early Unit End and the client dumps all the work done.
Then you will download the same workunit so you can rerun it and double check its a bad workunit and not your hardware taking a dump.
This happens two times.

As for running a GPU client.
It need a whole CPU core to run at full speed.
So on a dual core box either ..........
Run one CPU client & a GPU client.
or .............
Run the SMP client.
It depends on your CPU & Vid card which gives you more points.

Luck ............. :D
 
Also is there a way to schedule this to run? Say I want it to turn on at 7pm and turn off at 6 am? Also what would give me more bang for my buck? SMP with e6600 or one instance and GPU client on ATI X1900?
 
If you're not running it 24/7 you might want to be careful about using the SMP client as the deadlines can be very tight sometimes. I'd say try it out, and see if the deadlines are met in time running during that specified time-frame... If not, I'd say give the GPU client a go.. it has the highest FLOPs rating atm out of any platform. SMP will give you more PPD, but only if you can run it enough to finish the work in time.
 
You have to remember as well that Stanford does not rely on folding a protein once. Each unit is run several times to make sure the data comes back the same way.

From the science end redundancy and repeatability make perfect sense.
 
BillR said:
You have to remember as well that Stanford does not rely on folding a protein once. Each unit is run several times to make sure the data comes back the same way.

From the science end redundancy and repeatability make perfect sense.
Click to expand...

http://forum.folding-community.org/fpost170041.html#170041 Of course, many similar simulations of the same protein are done, but I don't think the units are usually run more than once.
 
taqueso said:
http://forum.folding-community.org/fpost170041.html#170041 Of course, many similar simulations of the same protein are done, but I don't think the units are usually run more than once.
Click to expand...


I can see where it would be easy to misread what was said in that post. If you take a moment and look at the folding FAQ:

http://folding.stanford.edu/faq.html#run.same

You will see a more exact answer.

Also, have a look at:

http://fah-web.stanford.edu/serverstat.html

You will see that there is definitely a finite number of servers, each loaded with from hundreds to thousands of the same iteration of the same protein.

I currently have 4 machines with exactly the same protein and after 5 years I find that is usually the case.

It would be wonderful if one run could be considered the only run needed, but there are too many variables involved for that to happen.

You can also search project by project in your own stats (at Stanford) to see how many of the same work units you have folded.

Luck:)




 
BillR said:
Each unit is run several times to make sure the data comes back the same way.
Click to expand...

I was responding to this, not the fact that many runs are done with the same protein. If by unit you mean something different than a work unit downloaded by a client, then I can see where the misunderstanding is. The links you gave do not have any information stating that the same work units are or are not rerun to verify the results.
 
Its my understanding that if you get the workunit in before the first deadline then each protien plus (run, clone, gen) numbers are unique.
If its later than the first deadline or you get a EUE then the workunit is re-issued.

Teecee said:
Also what would give me more bang for my buck? SMP with e6600 or one instance and GPU client on ATI X1900?
Click to expand...

Running 24/7 then ............
An e6600 @3.5Ghz with the SMP client will drop around 1500-1800 PpD.
An X1900 with the GPU client will drop around 600-800 PpD.
An e6600 @ 3.5Ghz with the CPU client will drop anything from 200 upto 800 PpD depending on which protien you can get. But 250-300 PpD would be about average.
Also you'll pull slightly less power running just the SMP client.

Luck ........... :D
 
taqueso said:
I was responding to this, not the fact that many runs are done with the same protein. If by unit you mean something different than a work unit downloaded by a client, then I can see where the misunderstanding is. The links you gave do not have any information stating that the same work units are or are not rerun to verify the results.
Click to expand...

Perhaps it's a reading thing, so I'll quote from Stanford:

"I just finished a WU and now I got another for the same protein. Is there something wrong? No, most likely everything is fine. We're studying the dynamics of a few proteins, so you're likely to get the same protein to work on multiple times. Each WU gives us additional information about the dynamics of that protein, so it is important to us. Indeed, if we did only 1 WU per protein, we would not learn very much".

If you look at the grand total of work units (yes proteins are the same as work units) that would mean Stanford coders would have had to code millions and millions of individual and different proteins to keep us all in work. If that were the case then for a given unit you would only ever see that unit folded once. I recall a number of years back V,J. stating they like to see 8 complete runs of each unit in order to insure parity.

I assume these are your stats as reported by Stanford:

http://fah-web.stanford.edu/cgi-bin/main.py?qtype=userpagedet&username=taqueso&teamnum=33&prange=0

Under Contributions by team and project:

http://fah-web.stanford.edu/cgi-bin/fahproject?p=638

You have done 20 of these exact same work units.

Hope that makes things a tad more clear

Luck
 
My turn to quote stanford, from the FAQ: "What has the project completed so far? We have been able to fold several proteins in the 5-10 microsecond time range with experimental validation of our folding kinetics." I bolded the part I want to emphasize. By your reasoning, we would have all been redoing the same few things tens of thousands of times.

You said that a protein is a work unit, I think this is dead wrong. What Stanford calls a single project studies a single protein or small family of proteins and is split up into many many "work units" that are crunched by many volunteers. I am saying that each chunk of work is not normally redone. Yes, we all have many work units from the same projects, working on the same proteins, in the same situations, but slightly different parts, timeframes, etc.

Under Contributions by team and project:

http://fah-web.stanford.edu/cgi-bin/fahproject?p=638

You have done 20 of these exact same work units.
Click to expand...

No, I have done 20 different work units from that project.

Also, I do recall from years ago reading that they did do some kind of random sampling of redos to catch cheaters, but that might have been dnet.
 
I don't mean to be rude but looks like people are steering this thread in a different direction. Can anyone answer the question about scheduling when the folding starts and stops?
 
Teecee said:
I don't mean to be rude but looks like people are steering this thread in a different direction. Can anyone answer the question about scheduling when the folding starts and stops?
Click to expand...

If you set up folding as a service, you can use the task scheduler to start/stop it.

A command like net start "service name here" will start the service, and net stop "service name here" will stop the service.

If it isn't a service, you can still use the task scheduler, and kill the process with a technique such as this one.
 
taqueso said:
If you set up folding as a service, you can use the task scheduler to start/stop it.

A command like net start "service name here" will start the service, and net stop "service name here" will stop the service.

If it isn't a service, you can still use the task scheduler, and kill the process with a technique such as this one.
Click to expand...

Thanks
 
You must log in or register to reply here.
Back
Top