DF: Like Walking in MUD!

[Insidious]

ugh........

This new protien is sssssssssssssssssssslllllllllllllllllllllllllllllllllllllllllloooooooooooooooooooooowwwwwwwwwwwwwwwwww

(129AA)

This machine went from ~210,000/day to ~124,000/day

 

[Robor]

Is that good or bad for the DF TeAm? By that I mean do we have any fast teams chasing us that might be slowed by the slow protein?

 

[FoBoT]

everybody has to crunch the same protein, so it isn't necessarily good or bad

if a team has a lot of members that move from project to project, then a slow protein is a good time for them to go to other projects, since less Ghz on DF will not result in such a quick change in the delta from one team to another, however this phenomena tends to benefit teams on the top that are trying to stay ahead of the pack, for a team trying to move up, it can make it harder

for example, the last protein was 105AA and relatively fast, one of the fastest ever done and took about 1 month to complete, so compared to the structures/stats accumulated during a slower protein (like the new 129AA one), 1 Ghz of cpu on the old protein for 1 month got you more "points" (structures) than putting the same 1 Ghz cpu on the new protein for 1 month will

since the structures generated for each protein are NOT weighted, it is to a team's advantage to move every available boxen onto DF when a fast protein is working and then when a slow protein is up, let those "loaner" boxen return to their home projects, or even move some DF boxen over to the other projects to "pay back" for the Ghz used during the "Fast" protein

the non-weighting of the structures really sucks for new comers, they see the huge numbers that long time participants have racked up on the fast proteins and see how slooooooow their boxen are doing the slooow protein and get discouraged

oh well, we get what howard needs crunched

 

[FoBoT]

for any DF'ers that want to benchmark thier boxen, there are two easy ways to do it

for winders users, dfGUI is a great tool that includes a button to create a text file containing benchmark info, see the end of this post for two examples, one from the last/fast/105AA protein and one from the new/slow/129AA protein

for any boxen with perl installed, Dyyryath's Quickie Benchmark Script for Linux Users is very handy
--------------------------------------------
these are both from my laptop

old fast protein

------------------------------------------------------------
Distributed Folding Windows dfGUI v1.8 Benchmark

Sample Size : 19670 structures over 10396 seconds.
Protein Size: 105AA

Structures Per Second: 1.89
Structures Per Minute: 113.5
Structures Per Hour : 6811
Structures Per Day : 163475

OS : Windows XP MHz: 1063
CPU: Intel(R) Pentium(R) III Mobile CPU 1066MHz
Client Switches: -df -qt -i f -rt -s 10000
------------------------------------------------------------

new slow protein

------------------------------------------------------------
Distributed Folding Windows dfGUI v1.8 Benchmark

Sample Size : 10755 structures over 11504 seconds.
Protein Size: 129AA

Structures Per Second: 0.93
Structures Per Minute: 56.1
Structures Per Hour : 3366
Structures Per Day : 80775

OS : Windows XP MHz: 1063
CPU: Intel(R) Pentium(R) III Mobile CPU 1066MHz
Client Switches: -df -qt -rt -s 3000
------------------------------------------------------------

you can see the new protein takes about twice as long to X number of structures

 

[Insidious]

Good or Bad?

Well, we lost most of our folders last time the protien size grew...........

So I guess that means its BAD!