LHC@home

We always suggest to people that it is a good idea to mix LHC with other projects. This project rarely has work, and our boxes need to be on 24/7 if we are going to get work when it comes, yet we do not want them to be idle in the meantime. The solution is to have them doing other projects.

Many small projects are springing up - and almost any would welcome our support. Some of us (me included) will also have favourites amongst the large projects.

Please post suggestions below. Please don't knock other people's suggestions - your are free to say why you think your suggestions are best, but please don't tell other people why you think theirs are wrong. I don't want this thread to turn into a flame war.

River~~

---

As thread creator, I get to go first :)

The three large projects I'd each rate as no 1 (each for different reasons) are Einstein (most interesting science), Rosetta (project most friendly to participants) and CPDN (project needed most urgently by humankind). These are personal views only, alternative views are welcome.

One of these that will particularly welcome more curnchers is Rosetta - for what they *really* want to do they need a cohort of volunteers of the size of SETI, or bigger, and they are less than 1/10th of the way there yet.

They aim to be the best project at keeping their participants informed, and in my opinion succeed -- David Baker (science) and David Kim (software) post on the Rosetta boards even more often than Bruce Allen and Ben Owen do at Einstein, and they even encourage their post grad students to tell us what they are doing.

I'd rate Einstein second on the project-participant communication stakes, by the way - I am certainly not knocking them as they rate no 1 for me on their science (General Relativity is my own field so I am biased)

Another large project, if you have a reasonably fast box, is CPDN - they are trying to work out the reality and the extent of global warming and need all the crunch power they can get. They ask for a minimum of 1.6GHz to join, and in setting project shares I'd suggest you need to give them at least 0.8GHz, so on a 2GHz box, if you have them at all you'd want to give them at least a 40% share in my opinion.

The big advantage of CPDN to mix with LHC is that CPDN work lasts so long (months for a single WU) that it really does not matter when LHC has work as far as CPDN is concerned. Put just CPDN and LHC on the same box, give CPDN only 1/3rd of the resource, and in fact CPDN will get over 90% of the time anyway.

R~~

As thread creator, I get to go first :)

The three large projects I'd each rate as no 1 (each for different reasons) are Einstein (most interesting science), Rosetta (project most friendly to participants)


I joined "Rosetta" yesterday. They sent me work on the spot ! :)


and CPDN (project needed most urgently by humankind).


Unfortunately, CPDN and the other BBC projects need powerful computers and are not compatible with Macintosh machines ... :(((

...

R~~

---

Anne

© anne schmidt

Weblog - Now Public

Unfortunately, CPDN and the other BBC projects need powerful computers and are not compatible with Macintosh machines ... :(((

too true. I have a stack of slow 686's and know the feeling. R~~

LHC and Climate Prediction on a 1 GHz computer
LHC and SETI on a dual 2 GHz Xeon workstation
LHC and SETI on a dual 595 MHz P3 workstation
LHC and Einstein on a 1.7 GHz workstation
LHC and SETI on a 400 MHz computer
LHC and Einstein on a dual 2 GHz Xeon workstation

---

I'm going for SPINHEGE, QMC, LHC, Einstein.

Spin is currently out of work. They're moving to new hardware.
QMC is in Beta but has work, even for slow hosts.
LHC.... if you read this, you know all you need...
Einstein has been my BOINC backup for a long time, my goal is 100.000 credits there, then I'll be off for the other three projects on the list.

Cheers,

Sysfried

I personally like to help out a few of the smaller projects when LHC is as usual outa work.

RCN is one of my favorites but is currently out of work until tommorow
RS is another but windows only sadly so only gets my part time cruncher there.
Einstein is also a great project and slower machines get basically all short workunits so my 2Ghz system can do a short one in 1.5-2hrs
Malaria is another of my favorites but my machine for that decided the workunits were too hard and started overheating but about 5hrs on my 700Mhz

---

Another thing CPDN is real good at is testing out any cooling and overclocking... I've always run some memtest and prime95 but even after prim95 has run for a hour or two. CPDN will occasionally find that little glitch that causes you to back down another MHz or two... none of the other project will put the load on your box that CPDN will. For a dedicated cruncher I look on that as a good thing. Not all will agree ;-) I also happen to agree it seems like a very relevant project.

Rosetta, QMC, Einstein and 'Skeeter Control are other favorites.

---

- da shu @ HeliOS,
"Free software” is a matter of liberty, not price. To understand the concept, you should think of “free” as in “free speech”, not as in “free beer".

On one of the other boards I once asked which projects are best suited to which processors.

Does anyone have any kind of data on that?

Some CPUs seem to do better with certain workloads.


It may be that my 2.53 celeron is better at Rosetta than LHC or vice versa.

It would be nice if BOINC would look through the data on the results on all the different machines and give a project ranking for effectiveness by each type of CPU.

Or, simply allow us to click on a link from our cpu to a page listing all the available projects ranked in order by the relative effectiveness of the particular CPU.

---

I am currently running Malaria Control and HashClash when they have got work.
I've given up on CPDN with great sadness,it has been my main project since they went BOINC, and had to detatch because my pcs were overheating and only resulting in computation errors.
I've also given up on QMC because their WUs were huge to download on dial-up.
I have also been working on Folding@home for over a year but suspended it over the summer, because otherwise my pc room would end up being an oven and impossible to work in!
Now with the cooler weather settling in at last, I am going to see again which projects to restart, reattach to, etc.
So, really I stick to small, fast WUs that allow my computers to get the job done in decent time without too many problems... It's more a hardware limitation than anything else!

---

WinXP Home, P4 3.2GHz HT, 512Mb RAM./Win2K, P4 1.8GHz, 512MB RAM

Join World Wide S.E.T.I.

Atotos, I see you're also attached to Rosetta. I don't know about the workunit download times for Spain, but the processing times are reasonable on a P4 2.8 (about 3 hours).

I wanted to run the Worldgrid cancer project at work, but was surprised by the 720meg ram requirement (that machine only has 512mg - but it's kept busy with the 5 projects I have loaded on it - with 3 generally not having any work).

Philip, as for the request about a table of which project would go well with what cpu, one of the things to remember is it's not all about the cpu. I hit a hard ram issue with Worldgrid cancer and a disk issue with CPDN (it was temporary). There's also the fact that seti is starting to put up some large variable sized workunits out there. Also in regards to workunit deadlines, do you keep the machine running 24/7 or turn it on every couple of days. How many projects and and how do you want to share the time on the cpu? I know your question is a about a single processor celeron, but what about the multi cpu machine (I'm using 4 work pcs that have HT and boinc counts these as 2 cpu machines).

I'm asking for solid answers about video cards on these variable like questions for MS FlightSim X on a different message board (If anyone knows if I can get a decent fps increase going from a Radeon 7000, 64meg video card to a 256meg video card for FSX, please let me know :) ).



William

---

....
Philip, as for the request about a table of which project would go well with what cpu, one of the things to remember is it's not all about the cpu.



I hit a hard ram issue with Worldgrid cancer and a disk issue with CPDN (it was temporary). There's also the fact that seti is starting to put up some large variable sized workunits out there.


Also in regards to workunit deadlines, do you keep the machine running 24/7 or turn it on every couple of days. How many projects and and how do you want to share the time on the cpu?


I know your question is a about a single processor celeron, but what about the multi cpu machine (I'm using 4 work pcs that have HT and boinc counts these as 2 cpu machines).
....
William

I run my single machine 24/7 with boinc engaged 100%.
How is that relevant?
Even if it affects the table somewhat, the table would still be a useful indicator of which machine.

Yes, memory, disk etc will affect the table, but they will only broaden the gaussian distribution about the mean.
We should still be able to rank projects by processor (or processor by project).

Multi core and multi chip machines should have their own row or column in the table.

---

I run my single machine 24/7 with boinc engaged 100%.
How is that relevant?
Even if it affects the table somewhat, the table would still be a useful indicator of which machine.

Yes, memory, disk etc will affect the table, but they will only broaden the gaussian distribution about the mean.
We should still be able to rank projects by processor (or processor by project).

Multi core and multi chip machines should have their own row or column in the table.

Philip,


There are some projects that warn what average duration of a workunit is, I think einstein is fairly consistant. Seti is getting into longer ones. Here are 3 from my seti on the same computer:

cpu time granted
credit
17,755.98 27.82
33,742.97 62.33
313.33 0.11

There have been references that LHC has been putting out smaller units lately too. I'm sure there's a math forumla for all of this somewhere. Someone would have to take the time and look at some work units and compare all the computers for each unit and kind of work out how long they take for a given processor. But, then again, how do you know if the system is being overclocked? Or, and this is probably the exception, if someone actually "lowered" their bus speed because the cpu is running too hot and the noisey fans bothers them (it's not me, but my colleague - I've talked to him about it and it just goes in one ear and out the other).

Next comes resources, is it up to the individual projects to furnish this data? And if so, are you expecting to compare "project a" workunits to "project b" workunits (which it seems so)?

"It may be that my 2.53 celeron is better at Rosetta than LHC or vice versa."

How can you tell if the machines are really similiar or not? 4 of the 5 I run have HT (and those show up as 2 processors) - I'm probably getting a little bit more work done, but it's taking me longer (parallel vs. serial).

There might be a couple of people out there that can furnish those numbers (especially the boinc farmers - if they're keeping results statistics).

Good hunting and when you find a table, please post it to this thread.



William

---

> Phillip, I have an Intel P4 2.53 GHz @ 2.75 Ghz and started with Einstein then Seti. Both worked well until early this year when for some reason Einstein went from 8 odd hours a Wu to 39 Hours a WU. I could not find a reason so stopped Einstein. Added Rosetta and it works a treat, possibly better than Seti.
So P4 works well with Seti and Rosetta. Runs Windows.

My AMD 4800+ does not seem to care what I put on it. It is currently doing Seti, Einstein, Rosetta, QMC, Ralph and Docking. Reasonable credit on all projects.
I did have one issue with Ralph that need me to eventually reformat and reinstall but other than that all work well (currently a Boinc screensaver issue with Rosetta, I just turned it off). Runs Windows.

My AMD Opteron 275 runs Einstein, Rosetta, QMC, ClimatePrediction, Predictor, Ralph and Docking. Runs Linux.
My AMD Opteron 285 runs Einstein, Rosetta, QMC, ClimatePrediction, LHC, Ralph and Docking. Runs Linux.
Climate presents no problem to these server type chips. All other projects run without a hitch.

Bear in mind that Climate has ultra long WU's (months).
QMC can have almost 40 hour WU's but also 1 to 2 hour ones as well.
Einstein eith 40 minute ones or 5 to 6 hour ones (more usual).
Rosetta you can set the time that you crunch from 1 to 24 hours so this makes very user freindly (will depend on protein being worked on).
Ralphs are the same as Rosetta.
Predictor is not going at the moment.
Seti has both short and long with the longer ones working your computer a bit harder.
Docking all WU's are the same but take much longer on Windows (6 to 20+ hours), than on Linux (2.5 to 6 hours) as dependant on processor speed and code is Linux optimised.

This is my own feelings with what I have found works for me. As you will see from my Sig, Rosetta has been a favorite with Ralph, but recently I have tended toward Einstein, QMC and Climate with Docking a fun new alpha project.

Hope this helps.

---

Mine are in the first graphic and the team's are in the second:


It suits me. ;-)

---

Click here to join the #1 Aussie Alliance on LHC.

RS is another but windows only sadly so only gets my part time cruncher there.

RieselSieve is testing Linux/FreeBSD apps (see here). Also note that Pentium 4s do not run the RS sieve apps very efficiently (comparision of various CPU architectures). However, another app called LLR is under testing and should allow Pentium 4s to contribute more optimally to the project. An RS WU takes a little over an hour on my Core Duo 1.73GHz, running the SSE2 optimized sieve.

I also run a number of other projects as indicated in my signature. I highly recommend Rosetta for many reasons like good communication to the participants and the ability to set your own WU run time.

---

I highly recommend Rosetta for many reasons like good communication to the participants and the ability to set your own WU run time.

I agree totally.

Settable run times was a participant suggestion and the fact that we got it is proof that the communication is genuinely two way on that project.

River~~

Ehm, i would say :
With LHC, you need to run an other project (a fact known since ... always).
Do like you wish/like ; you have the choice, and... more 30 projects from alpha status to full production status ! Attach, try, test, you like it ? good ; you don't like it, detach.
Keep an eye on the new and often little projects. They will try also to give you a very good support and to have a good communication with you.

---

Do you want to get banned for 31 years, your account and credits deleted at a Boinc project ? Predictor@home is your best choice.

Hey there River~~

You seem quite up to date on all the different programs out there, maybe you could answer a couple of questions that I have ---

Has the ROSETTA program fixed their software where it will work well with others? I seem to remember a while back, they had major problems with starting and stopping in a shared environment.

I notice on your listing the program "WCG" -- what can you tell us about it, and what is the URL for the site? The same question goes for "LEIDEN", which I suspect probably has a lot to do with electrical engineering research ....

Thanks

---

If I've lived this long, I've gotta be that old

Member of eleven projects :-)
Its not the lack of projects but just a certain itch to crunch, cant explain it.
I just wish Icathome could generate more projects :-)
Regards
Masud.

---