DC Vault 2

mmmmmdonuts said:
NFS@home I believe has super short run times. They are memory hogs but take only an hour or two if I remember correctly to test them on. I honestly believe hyperthreading in BOINC is going to depend on each individual project depending on whether or not the project is FP or integer intensive. It does not surprise me that you see the same with PSP run times because that is almost like Prime 95.
Click to expand...

Cheers, I may give the dreaded NFS@home a test when I have some time - there will probably be some babysitting involved to get accurate results. Thinking about it, I've not seen any speed up on any BOINC projects when I've allowed them to wind down prior to non-BOINC work. Sorry for stating the bleedingly obvious re PSP :eek::D
 
I guess there’s unorthodox behaviour in all spheres of DCing. Whilst waiting for the wife to finish watching the bloody X Factor, I started looking at what the top yoyo crunchers were using for hardware and came across something odd.

It turns out that my previous statement that Muon may be the best sub project to produce yoyo@home points was not entirely accurate. There’s a work type, ‘mp’, within ECM that seems to be a real earner to the extent that a high scorer from another team appears to be aborting every non-‘mp’ unit in order to target this one exclusively.

There may be an innocent yet unfathomable explanation but the approach seems like it’s just not cricket to me as it serves to farm off the lesser units for others to work on. It leaves me a bit miffed yet in turn allows me to better understand Razor’s ardent exchange with the same individual recently.

Ho hum, I need to stop looking at stats and get something useful done.
 
phoenicis said:
I guess there’s unorthodox behaviour in all spheres of DCing. Whilst waiting for the wife to finish watching the bloody X Factor, I started looking at what the top yoyo crunchers were using for hardware and came across something odd.

It turns out that my previous statement that Muon may be the best sub project to produce yoyo@home points was not entirely accurate. There’s a work type, ‘mp’, within ECM that seems to be a real earner to the extent that a high scorer from another team appears to be aborting every non-‘mp’ unit in order to target this one exclusively.

There may be an innocent yet unfathomable explanation but the approach seems like it’s just not cricket to me as it serves to farm off the lesser units for others to work on. It leaves me a bit miffed yet in turn allows me to better understand Razor’s ardent exchange with the same individual recently.

Ho hum, I need to stop looking at stats and get something useful done.
Click to expand...

I know who you are talking about and I'm not surprised at all.

Edit: He might be beta testing these units. I'm not sure.
 
Last edited:
We are about to bust into the top 20 of yoyo! Outstanding crunching!!
 
Anyone know how to change the percentages on what projects to run in BOINC? Say I wanted to run 2 projects and didn't want a 50/50 split anyone know how to change it to 75/25 or 80/20?
 
mmmmmdonuts said:
Anyone know how to change the percentages on what projects to run in BOINC? Say I wanted to run 2 projects and didn't want a 50/50 split anyone know how to change it to 75/25 or 80/20?
Click to expand...
In your project preferences:
If you participate in multiple BOINC projects, this is the proportion of your resources used by yoyo@home
Resource share : 100
 
Don't know what happened to the Knights Who Say Ni (dropped to 9th so quickly, glitch possibly??) in the DC-Vault but we moved up to 4th place. :)
 
So, I woke up this morning and my computer room was only 50°F!! I'd say it's about time for me to turn on some GPU folding power... Should help heat the whole of my upstairs lol.

Going by Sazan's list 3 pages back the "highest" AMD available project on the BOINC list is Collatz so I'll be throwing a 4870 and a 6970 on that project for awhile.

Are any of the non BOINC projects GPU enabled and if so are they worth doing on AMD GPUs?
 
Gimps is supposed to be pretty good, search through this thread, Phoenics had some links to the Gimps setup for GPU's. I dont know how to setup AMD as all mine are Nvidia
 
Gimps works well. Its not to hard to setup and on page 7 of this thread Razor posts the executable file link. (There maybe an updated version on the website)

A few caveats though. My 5850 consumed about 50-75% of a X6 core. I needed to also run two different instances to peg the GPU at 100%, otherwise it was around 80%. Also, It does involve manual receiving WU and sending results though.
 
Razor_FX_II said:
phoenicis said:
There may be an innocent yet unfathomable explanation but the approach seems like it’s just not cricket to me as it serves to farm off the lesser units for others to work on. It leaves me a bit miffed yet in turn allows me to better understand Razor’s ardent exchange with the same individual recently.
Click to expand...
http://www.rechenkraft.net/phpBB/viewtopic.php?f=57&t=11914
Click to expand...
And now this ...
http://www.team-ninja.com/vbulletin/showthread.php?t=48889
 
Poem@Home News:
http://boinc.fzk.de/poem/
"GPU app infos
The GPU parts of POEM++ are mostly finished, we are now working on packaging everything to send it out. However there were many small issues due to the fact, that the OpenCL libraries cannot be statically linked. We think these are sorted out now. This pushed us back by about 1 week in our schedule. We are now estimating Thursday in one week (24.11) as the release date of the first OpenCL linux app. Thanks for waiting."
 
Pretty damned shocked by the reaction to your question Razor. I’m truly sorry that my post has played a part in causing aggravation for you.

Bob
 
phoenicis said:
Pretty damned shocked by the reaction to your question Razor. I’m truly sorry that my post has played a part in causing aggravation for you.

Bob
Click to expand...
Please don't be sorry, you have no reason to be sorry to me or anyone else.
No aggravation here, please don't misinterpret me that any of this has to do with you.
I appreciated the heads up on the issue and continue to have the up-most respect for you, and you as my teammate.
My nature is to stand up for whats right and not to back down from the things that are wrong. Good or bad it's just the way I am.

In conclusion, Beyond PM'd me a truce and hopefully that is the end of it.
 
Cheers, I'm glad it's ended well over a virtual pint.

p.s. You shouldn't have that much respect for me, the wife sometimes wins the argument and gets me to watch the X Factor with her ;)
 
phoenicis said:
Cheers, I'm glad it's ended well over a virtual pint.

p.s. You shouldn't have that much respect for me, the wife sometimes wins the argument and gets me to watch the X Factor with her ;)
Click to expand...
LOL! thats funny, I'm watching last nights DVRd X-Factor with my wife and 8 year old boy right now, just taking a break for me to check stats.
Now I need to check my house for hidden cameras.:D
 
Razor_FX_II said:
Poem@Home News:
http://boinc.fzk.de/poem/
"GPU app infos
The GPU parts of POEM++ are mostly finished, we are now working on packaging everything to send it out. However there were many small issues due to the fact, that the OpenCL libraries cannot be statically linked. We think these are sorted out now. This pushed us back by about 1 week in our schedule. We are now estimating Thursday in one week (24.11) as the release date of the first OpenCL linux app. Thanks for waiting."
Click to expand...

Interesting. Perhaps we should jump on this with a bunch of GPUs to see if we can gain some ground (assuming that the GPU client outperforms the CPU client). I could move over my GPU from DistrRTgen for a while.
 
metallicafan said:
Interesting. Perhaps we should jump on this with a bunch of GPUs to see if we can gain some ground (assuming that the GPU client outperforms the CPU client). I could move over my GPU from DistrRTgen for a while.
Click to expand...
I believe the first GPU app will be for Linux only so I won't be able to help until a Windows version is ready.

Updates from Poem GPU:
Edit (24.11): There is still a small holdup with the client. We hope to get it out tomorrow.
Edit (25.11): Sorry, there will be another delay (hopefully only until Monday). The current problem is getting both AMD and nvidia cards running with the same binary. If we don't find a solution until monday the first release will be using multiple binaries.
 
Ahh .. i missed that detail. I guess I will have to wait for a Windows version as well. :(
 
If you are having problems with ATI 11.10 and 11.11 running Collatz GPU, GIMPS GPU or other projects - un-install those versions and
Delete:
C:\Windows\system32\amdocl64.dll
C:\Windows\system32\amdoclcl64.dll
C:\Windows\syswow64\amdocl.dll
C:\Windows\syswow64\amdoclcl.dll
(similar files for 32-bit?)
Reboot and install ATI Driver 11.9 32-bit or ATI Driver 11.9 64-bit
 
Is there a list of BOINC project special hardware requirements anywhere guys? My rigs typically have a half GB of memory per CPU thread which didn't serve me well last weekend. I had 3 rigs permaset to pick up Lattice work and by the time I realised some had been grabbed they'd already gone nuts. Each work unit required almost 1 GB and failed units abounded.

I know MJ12 needs serious bandwidth/data allowance and have read that NFS@home has high memory requirements although I'm not sure how high. I thought I'd ask if any other tank traps await prior to formulating my favourite Christmas shopping list ie from me to me.

metallicafan said:
I guess I'll help over in the GPUGrid camp for a while.
Click to expand...

Welcome back to the camp :) At our current ppd we may gain 4 places over the coming week.
 
Last edited:
phoenicis said:
Is there a list of BOINC project special hardware requirements anywhere guys? My rigs typically have a half MB of memory per CPU thread which didn't serve me well last weekend. I had 3 rigs permaset to pick up Lattice work and by the time I realised some had been grabbed they'd already gone nuts. Each work unit required almost 1 MB and failed units abounded.

I know MJ12 needs serious bandwidth/data allowance and have read that NFS@home has high memory requirements although I'm not sure how high. I thought I'd ask if any other tank traps await prior to formulating my favourite Christmas shopping list ie from me to me.



Welcome back to the camp :) At our current ppd we may gain 4 places over the coming week.
Click to expand...

I think you mean GB of memory in your statements? Another project that has a fairly high memory requirement is Cosmology@Home. I believe its around half a GB as well. Some of the sub projects in yoyo, WCG and RNA World have a fairly high memory requirement but you can just opt out of those projects.
 
You're quite right! Corrected.

Any idea how much NFS needs? Do you think a GB per thread will be sufficient for this and anything else lurking?
 
I cant comment on all of the projects but typically the NFS units will be fine on 500MB/thread. I don't remember any taking up much more than that but I guess its been a little while since I have crunched that project.

With Cosmology@Home I would plan on using 400-500MB/thread at a minimum. Often by the time the WU was nearing completion many of the WUs I was crunching a couple weeks ago were using 750MB/thread.

At any rate - There are very few projects you would have to worry about RAM issues if you had 1GB/thread! ;) Maybe Lattice, or CEP2 in WCG. Thats all I can think of at the moment.

Edit: Building my new 2500K rig today. Looks like I might need it fend off mmmmmdonuts in WCG!
 
It would seem that there's not that many memory demanding projects out there that I haven't already encountered. I'm not sure increasing the capacity is worthwhile for a couple of difficult projects that can't be worked around but I probably won't be able to resist anyway. My flimsy excuse will likely be that it also adds a bit future proofing.

Thanks for the helpful replies fellas.
 
With the Yoyo contest over do we have any other targets in mind? I was thinking about switching back to muon but the gap between us and the next team is scary -_-.
 
Briliu said:
With the Yoyo contest over do we have any other targets in mind? I was thinking about switching back to muon but the gap between us and the next team is scary -_-.
Click to expand...

Can't speak for everyone else, but we got the WCG Christmas challenge going on till the 24th if you would like to help there.

If not there we can always use help on 17 or Bust, or PSP PRP if you don't mind setting them up. 17 or Bust just uses Prime 95 and is super easy to get going, just make sure you use the latest Prime 95 client (v26.6) not the 26.3. PSP PRP is not that easy to get going but we can gain a huge amount of points for passing the next team. I am willing to help you if you want to try either one of these.

If you want a BOINC project I suggest we go for Cosmology@Home. The only issue with this project is it takes quite a bit of RAM per thread but I have found it to be a very stable WU.

After the WCG challenge is over I am most likely focusing on the projects I have listed if that helps you at all.
 
I'm BOINCing it up on Cosmology@home for the next couple of months.
 
DooKey said:
I'm BOINCing it up on Cosmology@home for the next couple of months.
Click to expand...
The help in Cosmology is welcomed. I find that each work unit typically uses less than 756KByte.
Running 4 work units in Windows on a QuadCore with 4MB RAM works fine.
 
Thanks for the tips Donuts! I'm leaning towards Cosmology now.

Do you guys think it'd play nice at 7 threads on an 2600K with 8gb of ram?

Also, if i were to pause it mid solution does it free up the ram? I typically pause it to play an hour or two of games then resume the "solve".
 
Briliu said:
Do you guys think it'd play nice at 7 threads on an 2600K with 8gb of ram?

Also, if i were to pause it mid solution does it free up the ram? I typically pause it to play an hour or two of games then resume the "solve".
Click to expand...
Cosmology with 7 threads in 8 GBytes RAM should work fine and as long as you do not check "Leave applications in memory while suspended" it will unload the work units from memory - in this way it is like shutting down Boinc and you may loose up to 30 minutes of work depending on when last checkpoint was saved.
 
Razor_FX_II said:
The help in Cosmology is welcomed. I find that each work unit typically uses less than 756KByte.
Running 4 work units in Windows on a QuadCore with 4MB RAM works fine.
Click to expand...

Damn. 4MB of RAM? Too rich for me. :D

I'll add some horsepower Cosmology after the Christmas Race is over.
 
I'm sticking with phoenicis for being the [H]ard-est DC'er Commando for the last several months.
I believe he deserves my vote and recognition.
 
TheMacGuy gets my vote. He's dedicated almost 4 years to WCG. I recruited him not long after I joined the team in 2008. He is a coworker of mine and has elected not to participate in this forum.

WCG is the second most important distributed computing effort behind FAH. Getting to the #1 spot is a awesome achievement and should be recognized IMHO.
 
You must log in or register to reply here.
Back
Top