double folding?

[vrsick00]

I saw a few people folding times 2, how do you do this?

 

[Retro Rex]

Please restate your question.

 

[IROCthe5.7L]

Are you referring to running two instances of F@H?

 

[BillLeeLee]

If it's the two instances thing, they're probably either using a multiprocessor system, or a hyperthreaded Pentium 4.

 

[vrsick00]

yea the 2 instances jobber, having to seprate F@H's open and folding. thanks for the answers

 

[SoL]

What operating system are you using?

 

[vrsick00]

Xp pro

 

[budec]

Yea, you probably only want to run two folding clients at once if

1) You have a multiprocesser system
2) You have a P4 with hyperthreading
3) You only can connect to the internet a few times a week

If you connect to the internet atleast once every day or two and have a single processer machine there isn't much reason to 'double fold' as it'll just slow down performance.

 

[Tigerbiten]

The way I do it is......

Make two folders for your F@H.
I call them F@H1 & F@H2.
Download the "No-nonsense" text-only console client into both folders.
It will work with one graphical & one console but not with two graphical clients.
Then start the first client and in the advanced config section leave it as machine id 1.
Then start the second client and in the advanced config section set it as machine id 2.
The machine id gets tacked on the end of the user id so that Stanford can tell the clients apart.

Only three reasons for doing this.
1. Running a dual CPU box.
2. Running a HT CPU box.
3. Trouble downloading workunits. ie, Your on a dial up.
If you do this make sure your box is fast enough to finnish the work units in time.

Luck.........

 

[JonasSteele]

Hey,

I am still fairly new, but have XP and a P4 with HT and I believe if you leave HT on and run 2 instances of folding you will see about a 15% gain in productivity

You can put F@H1 and F@H2 at 100% on each, 50% on each means both will run at %25 total power, use 100% and each will use %50 total power and don't forget:

-local "for two instances"
-service "if you're using the task scheduler."
-advmethods "better chance of getting gromacs"
-forceSSE

And a good XP Program for setting up 2 services to start automatically and be out of sight out of mind is:

http://www.firedaemon.com/

If any of this is wrong or inacuracte I am sure the great team here will tell us.



JonasSteel

992 Points and Climing

 

[SoL]

The -local switch is only required if you are running the graphical client and a console client.

 

[gnewbury]

It's always a good idea to run a timeless WU along with the number of instances equal to the number of CPU's, either real or virtual. I'll have a network outage or the servers will be down and the regular WU stop and the timeless start crunching.

 

[JonasSteele]

The -local switch is only required if you are running the graphical client and a console client.

Thanks SoL for the insight, and since I am running two consoles, that's one less switch needed!

It's always a good idea to run a timeless WU along with the number of instances equal to the number of CPU's, either real or virtual.

Timeless WU? Please explain/elaborate? I am currently running 2 console instances on 2 P4's via Firedaemon as services, and though I have never Knock on Wood had network issues since going highspeed, I'd like to understand more about these Timeless? Wu's. Thanks in advance!


JonasSteel

1127 Points and Climing

 

[Tigerbiten]

Because Stanford want's the protien's back so they can workout the next lot of work units most have a deadline.
If you take longer than that deadline to get the WU back and it gets reissued and you don't get a points for your work.
If you set a very slow box which wont make the deadline for F@H to fold G@H then it will be issued work unit with no time limit to them.
This also means that you can set up a fast F@H box with one copy of G@H to run in the background.
This means when Stanford's servers are getting hammered and up take an hour or so to get your next WU the G@H units takes up the slack.

Luck.........

 

[ezee]

The -local switch is only required if you are running the graphical client and a console client.

Not so.The local switch is required whenever you have 2 or more instances in separate folders. Each must be set up with it's own machine id (from 1-8). This is so whether running two consoles or a console and a graphical client together.

 

[ezee]

Hey,

I am still fairly new, but have XP and a P4 with HT and I believe if you leave HT on and run 2 instances of folding you will see about a 15% gain in productivity

You can put F@H1 and F@H2 at 100% on each, 50% on each means both will run at %25 total power, use 100% and each will use %50 total power and don't forget:

-local "for two instances"
-service "if you're using the task scheduler."
-advmethods "better chance of getting gromacs"
-forceSSE

And a good XP Program for setting up 2 services to start automatically and be out of sight out of mind is:

http://www.firedaemon.com/

If any of this is wrong or inacuracte I am sure the great team here will tell us.



JonasSteel

992 Points and Climing

Mostly true. Exceptions. -advmethods no longer has any effect on whether you get a tinker or gromac. It does alert the assignment server that you are willing to fold new WU's that are in beta development, also that you are willing to try new or newly revised beta cores.

Before gromacs and double gromacs were released to the general folding community, -advmethods did improve chances of getting gromacs, but no longer.

Similar comments apply to -forcesse. A while back, athlon xp's did not automatically use sse instructions, but rather used the less efficient 3DNow! AMD worked with the Pande group and corrected that issue and -forcesse is no longer needed for AMD processors.

-forcesse can be useful if you have had a system crash or other improper termination (like a power outage) and either been lucky enough to not lose previously completed gromac frames or unlucky and restart from the initial work packet. In those situations, sse will be disabled on restarting that work unit and -forsesse will restore it.

Hope this is helpful.

 

[T-3]

Not so.The local switch is required whenever you have 2 or more instances in separate folders. Each must be set up with it's own machine id (from 1-8). This is so whether running two consoles or a console and a graphical client together.

I have 2 instances running on both of my duallies and I have never used the -local switch, I do use seperate folders with id 1 & 2. The only switch I am running now is the -forcesse switch.

T-3

 

[SoL]

Not so.The local switch is required whenever you have 2 or more instances in separate folders. Each must be set up with it's own machine id (from 1-8). This is so whether running two consoles or a console and a graphical client together.

I have several dual boxes that don't use the -local switch. Never have. For years. Works fine for me.

You only need to use the -local switch if you've previously had the GUI version installed... otherwise it doesn't do anything. It doesn't hurt to use it though.

 

[JonasSteele]

Not so.The local switch is required...

I have never used the -local switch....

I have several dual boxes that don't use the -local switch...

Well looks like general census says it is not needed, none the less I see no harm in leaving it there..

-advmethods no longer has any effect on whether you get a tinker or gromac. It does alert the assignment server that you are willing to fold new WU's that are in beta development, also that you are willing to try new or newly revised beta cores.

...advmethods did improve chances of getting gromacs, but no longer.

My question now is does accepting new WU's in beta development and newly revised cores via -advmethods reduce productivity and scoring, increase it, or neither?


JonasSteele

1485 Points and Climing

 

[gnewbury]

Timeless WU? Timeless? Wu's. Thanks in advance!


JonasSteel

1127 Points and Climing

Back in the day there was Genome@home, One could put it on a CPU and it would crank out WU for Stanford WITHOUT net access. Long history, story over, I retired as #1 on the team.
As shown here .
Thankfully Stanford instituted a "timeless" WU for folding. Not as timeless as Genome, but if you get it in within a few weeks or a month it still counts toward science.
When you run the client config use the GAH and no-net option.

 

[ezee]

Well looks like general census says it is not needed, none the less I see no harm in leaving it there..





My question now is does accepting new WU's in beta development and newly revised cores via -advmethods reduce productivity and scoring, increase it, or neither?


JonasSteele

1485 Points and Climing

Ok guys, I stand corrected re: -local switch and agree that only needed if using GUI and at least 1 more instance using console. Apologies if I misled.

Processing beta WU's or using beta cores is more likely to produce fewer total points as, by their very nature, there may be bugs or unstable WU's that end early or destroy themselves for zero points. That said, though, a new core may be far more efficient or the work units may have better inherent relative point value. (like dgromacs when they were in beta) On balance, then, it's a toss-up.

Further on beta, the new FAH5 is soon due for release as beta testing is going very smoothly, according to Bruce at folding-community.org, possibly within a week.