• 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.

F@H Cluster

E

Enigma

[H]ard|Gawd
Joined
Oct 15, 2003
Messages
1,031
Is it possible to do a folding @ home cluster, with say; 10 q6600 machines??? and have all the machines work o the same unit??? just curious I found a good deal locally on q6600 cpus..... is this even feasable and if so could you linux gurus point me to some documentation?

Thanks in advance
enigma
 
No.

At the moment the FAH SMP client is limited to 4 client cores so with anything over a quad cpu you need to run more clients.
Saying that the client does not scale well so running the 4 client cores on 2 cpu cores gets you the best points.
So with 10x Q6600's your best PpD will be had from running 20x SMP clients, locking each client to one half of each quad core.

Luck ................ :D
 
the SMP client doesn't go over 4 processes, and those aren't setup to be clustered (that I've ever seen)....generally doing 10 Q6600s will either be 10 single SMP clients or 20 SMP clients using either virtual machines or Affinity Changer

edit: should've hit refresh :p
 
could you not send the same unit to each machine and have it work on part of the unit and sent it back to the main control machine and have it re-compose the data as if it is one machine....... doing this would only utilize the 4 cores on each board individually
 
Enigma said:
could you not send the same unit to each machine and have it work on part of the unit and sent it back to the main control machine and have it re-compose the data as if it is one machine....... doing this would only utilize the 4 cores on each board individually
Click to expand...

Stanford has already done that stage when it designed the work units.
Because of that, each work unit cannot be broken down further.

Also there is too much data passing between the 4 cores to make a cluster of 4x single cored computers worth while.
You can even see this with a Quad cored CPU.
If you keep all 4 client cores in the 2 CPU cores that share the same cache it ~150-200 PpD (~10%) faster than if you split them between 2 CPU cores that have different caches.

Stanford keeps a record of what it sends where.
The way it identifies your clients is with both the UserID and the MachineID.
So for each client your running the combination of UserID + MachineID must be unique.
Otherwise you run the risk of some of your clients folding for zero points.
Be warned, It has happened.

Luck .................. :D
 
I need a different machine ID for every machine I'm running? even if its an entirely different box? my dorm has 4, all machine number 1?
 
Patman said:
I need a different machine ID for every machine I'm running? even if its an entirely different box? my dorm has 4, all machine number 1?
Click to expand...

You don't need different machine id for different computers unless you are running more than one clients on the said computer. All set to machine id 1 is fine.
 
Patman said:
I need a different machine ID for every machine I'm running? even if its an entirely different box? my dorm has 4, all machine number 1?
Click to expand...

No.
Its the combo of UserID + MachineID that must be unique.

Normally each computer you set up has a different UserID so it does not matter what you set the MachineID to.
Its when you clone drives or copy the FAH folder between drives that you have to be carefull.
Now there's a risk that all the computers will end up with the same UserID.
Also running two or more clients on a single computer means that they will have the same UserID.
Now you will need to make sure they have different MachineID's.

For example ....
Say you had five FAH clients.
They can have ..... UserID's A, B, C, D, E and MachineID's 1, 1, 1, 1, 1. <- Normal setup.
Or .......UserID's A, A, A, A, A and MachineID's 1, 2, 3, 4, 5.<- Usefull if your sneaker netting.
Or even something like .... A, A, B, C, D and 1, 2, 1, 1, 1. <- 1 dual & 3 single clients.

You can find both your UserID and MachineID at the top of the FAHlog.txt file.
One of mine looks like ......
[09:44:24] - Ask before connecting: No
[09:44:24] - User name: Tigerbiten (Team 33)
[09:44:24] - User ID: 122E0E363EB70018
[09:44:24] - Machine ID: 1

Luck ............ :D
 
You must log in or register to reply here.
Back
Top