F@H: few questions

[Overkiller]

On my 2.8@ 3.2 ghz rig i have the following (2 processes) after about 18 hrs:

Current Work Unit
-----------------
Name: p638_L939_K12M_ext
Download time: March 15 07:04:38
Due time: May 4 07:04:38
Progress: 36% [|||_______]
__________________________________
Current Work Unit
-----------------
Name: p639_L939_K12M_nat
Download time: March 15 06:54:00
Due time: July 28 06:54:00
Progress: 35% [|||_______]


now are these the new gormacs, the older tinkers, or what?

ALso, does it sound about right speed-wise? how fast should i crunch these 2 Wu's? Ideas?
If it helps...the first unit completed:

-147 Frames and the 2nd one did 143 frames-

Also, i only show up in the stats once this WU is done and uploaded...right?

 

[Hyperfocal]

p638_L939_K12M_ext is a 70.9 point Tinker WU with a deadline of 30 days.

p639_L939_K12M_nat is a 70.9 point Tinker WU with a deadline of 50 days.

They are both 400 frame units. They will take some time to complete. If these are two instances on a single P4, that is pretty good speed.

You should have a log file in each F@H directory called FAHlog.txt. This will have an entry for each frame written. That will give you your time per frame and from that you should be able to figure how long each WU will take.

Make sure you have the -forceASM and -advmethods arguments enabled to reduce the number of Tinker units you get. Almost everyone gets Tinkers for their first WUs.

Here is a list of all the projects and their point values.

 

[r0tt3n1]

You can look HERE for the currently running projects and determine if it is a Gromacs or Tinker. BTW, those are both Tinker wu's.......
Tinkers are large and slow, altho they are helpful to the project, us stats hounds like Gromacs. Use the -forceSSE and -advmethods flags for each cleint to help in getting Gromac wu's.
If they are your first wu's, then yes you will appear in the stats after they have finished and are returned.
IMO, after 18hrs, those units should be farther along than that. May want to stop the clients and use the flags listed above to help speed it up a bit. You should finish the Tinker wu's (some might delete them and hope to get Gromacs instead), but hopefully should finish faster, then next wu may be a Gromacs.

 

[r0tt3n1]

Dasm you's fast HF!!

 

[bgeh]

anyway here's a list of the beta proteins.....if you can't seem to find your protein's name
http://folding.stanford.edu/psummaryB.html

 

[Overkiller]

These are 2 instances on 1 P4C that is also used pretty heavily in the evenings for gaming.

On both instances i am getting 8 Min/Frame.

So that is about 2 frames per every 8 minutes, or 1 frame every 4 minutes.
Does that sound good?
-If it's 2 x 400 Frames then these 2 WU should take 4.1 Days =/

good/bad?

Now that the clients are running, how do i change the flags to add -advmethods? (i can't get firedaemon to run both instances so i dunno how to change the flags)

 

[Hyperfocal]

Add the arguments to the command line when you run the program or add them outside of the quote marks in the target box in the properties of the item in your startup folder.

Another member is getting about 6 minutes per frame with 638 on his 2.4c with one instance. So it looks like you are picking up about 30 seconds per frame using two instances. 6 minutes at 2.4 is about 4.5 minutes at 3.2.

Clear as mud?

 

[Overkiller]

so umm....sort of confused...lol

someone is getting 6 min/frame on 1 instance ONLY. i'm getting 8 min/frame x 2....

so are you saying that ideally using the above equation i should be getting 9 min/frame x 2...so i'm actually doing good? Or am i actually gaining 30 seconds per instannce so every frame i am 1 minute slower????

sorry about the noob questions

 

[Hyperfocal]

so i'm actually doing good?

For some of the slowest of the Tinker units. IIRC, I got about 30-31 hours for the p638 unit and you should get about 26.6 hours.

Note to self: stop posting after working a 14hour day. You stopped making any sense 3 hours ago

 

[Overkiller]

hey man i know the feeling

I've posted afer 20 hr work days and when i looked back upon my post i realized it made 0 sense

Thanks for all the help!

 

[Overkiller]

Originally posted by: Hyperfocal
Add the arguments to the command line when you run the program or add them outside of the quote marks in the target box in the properties of the item in your startup folder.

well i can't put both shortcuts to F@h4console.exe in my startup folder b/c they are the same file essentially. should i just rename one of the shortcuts? Would that work?

And when i run the command-line (by clicking F@h4console.exe) it just starts up. i can't add -advmethods.

thanks!

 

[MDE]

Originally posted by: Overkiller

Originally posted by: Hyperfocal
Add the arguments to the command line when you run the program or add them outside of the quote marks in the target box in the properties of the item in your startup folder.

well i can't put both shortcuts to F@h4console.exe in my startup folder b/c they are the same file essentially. should i just rename one of the shortcuts? Would that work?

And when i run the command-line (by clicking F@h4console.exe) it just starts up. i can't add -advmethods.

thanks!

To add the arguments to the shortcut, look at this for an example. You should be able to just rename one shortcut to get two copies running, like FaH1 FaH2 for example.

 

[bgeh]

well this is what it looks like on my cruncher, it's an athlon xp 1700+ palo.......
been using 2 instances due to dialup and me not being able to be awake when it needs a WU
pic
hope it helps

 

[MDE]

Originally posted by: bgeh
well this is what it looks like on my cruncher, it's an athlon xp 1700+ palo.......
been using 2 instances due to dialup and me not being able to be awake when it needs a WU
pic
hope it helps

That's an olld client, ever think about upgrading ?

 

[bgeh]

it's been a long time since i ever did F@H after the seti team call

 

[Overkiller]

thanks guys, i will try this out when i restart the clients. It looks like one of the Wu's is pulling ahead of the other one. 196 frames completed vs. 190

 

[Overkiller]

my barton 2500+ is about 90% done w/ its first WU. i think it's doing good time.

However it is running FAH4core_78.exe instead of FAH4core_65.exe

my P4 system is running 2 instances of the _65 version of the core. the barton ahs the _65 version of the core in its folder but its running the _78 version...

am i missing something? did my P4 not update it's core or...

thanks guys!

 

[Hyperfocal]

FAHcore_78.exe is the Gromacs core.

FAHcore_65.exe is the Tinker core.

There is also a new FAHcore_79.exe which will debut soon. This is a different, new core featuring SSE2, but the other two cores will continue to be used.

It depends on the types of calculations needed. There are a lot of new Tinker units because they have important science that needs the Tinker core right now. (Something stats freaks like me tend to forget from time to time.)

 

[MDE]

Any word on if the new core will be faster? Also, couldn't they update the Tinker core to use SSE or at least optimize it a bit more?

 

[bgeh]

well here's some update from another project about the F@H tinker core
What will happen to the new GAH core?
We are continuing work on the new GAH core (M3). M3 is built off of Tinker 4.x and is poised to replace both the work that the GAH core performs as well as the Tinker core. The science in these two cores has been approaching each other for some time and so it's natural to combine them. We have no ETA for this release and it would appear as a new FAH Tinker core.
from here
http://www.stanford.edu/group/pandegroup/genome/faq.html