Sorry, folks, but I deserve an easy to install, reliable application. Otherwise I will not use any so called "power application" available, because I am not interested in a "power crash".
Electricity is exspensive in Germany today. The last power application for SR3 was very easy to install. If the project is interested in more crunching power, the programmers should use the same approach again.Give it a trial!
Intel Q9300 Quadcore, 2500 Mhz, 4096 MB RAM, GeForce 9800 GT, Vista Ultimate 64-bit, Ubuntu 10.10 64 bit
Sorry, folks, but I deserve an easy to install, reliable application. Otherwise I will not use any so called "power application" available, because I am not interested in a "power crash".
Electricity is exspensive in Germany today. The last power application for SR3 was very easy to install. If the project is interested in more crunching power, the programmers should use the same approach again.Give it a trial!
It looks like the app_info.xml that Bernd posted here should be fine. All you have to do is copy that and paste it in before you put it in the Einstein folder. More than likely Bernd will update the zip file to have the newer version in it, so you can also wait for that...
Ok- I'm new to the advanced apps, but got to it with SETI, and am looking forward to it here.
I'm running a 64 bit Vista Ultimate, Intel 6300, so I'm good for the app.
I'll guess both items go in the Programdata->Boinc->projects->einstein.phys.uwm.edu. Is that right?
Ralph
Easiest way is:
* Set Einstein@Home in BOINC Manager to "No new work". Let it finish, upload, and report current tasks. You can abort any unfinished work not yet started in advance.
Now, if you want to be friendly to your wingmen, then you might prefer not to abort WUs in your pipeline.
In this case, you can alternatively:
* stop BOINC
* extract the zip into the sub-dir of BOINC as specified above.
* open app_info.xml that you just downloaded with a text editor, and replace its contents with the text Bernd provided here.
* start BOINC again
:EDIT:
BOINC will then use the new power user app for newly dowloaded WUs, no work needs to be aborted manually or will crash.
As always, please continue to watch this thread for news about the app and please report any problems here (including problems with the screensaver etc.).
Note there is a known problem with BOINC when you are using BOINC's CPU throttling feature (when you configure BOINC in the web interface to use less than 100% of idle CPU cycles). So for this test (well, for E@H in general really), please do not use CPU throttling.
BTW, what I've seen so far from this app performance-wise is really promising.
BOINC will then use the new power up immediately also for half-finished work (even tho BOINC manager displays a different version label...it's lying).
I don't think so, Bikeman. Unlike the grandfathering ap_infos often provided previously, this one actually points to two different executables for the 604 and 605 cases. Bernd's comment seemed clearly to indicate that this was his intention--to use the older executable for work already branded.
Let's get this hashed out--if you are wrong then people may get a false impression of no improvement as they work their way through previously downloaded work. If I am wrong there are other consequences.
BOINC will then use the new power up immediately also for half-finished work (even tho BOINC manager displays a different version label...it's lying).
I don't think so, Bikeman. Unlike the grandfathering ap_infos often provided previously, this one actually points to two different executables for the 604 and 605 cases. Bernd's comment seemed clearly to indicate that this was his intention--to use the older executable for work already branded.
Let's get this hashed out--if you are wrong then people may get a false impression of no improvement as they work their way through previously downloaded work. If I am wrong there are other consequences.
You are absolutely right!! Sorry, I must have been too tired already :-), thanks!
I used the 'polite' .xml, and restarted BOINC.
It came up with a message saying to finish my work, remove the .xml, and restart. I had some 6.04's unstarted in the list.
I just renamed the .xml, and restarted. It just gave me some 6.04's to download.
I have four 6.05 pure results returned and validated from a Conroe-class quad I call Stoll4 (Q6600 mildly overclocked near 2.8 GHz).
As it has previously run a couple of hundred results from this frequency band covering well over a cycle, there is a decent reference both as to the natural systematic and random (including wiggle as part of random) variation.
My initial guess is that on this host for this type of work the new ap makes the host about 17% more productive, or that it takes about 85% as long as before.
While some of the random variation seen on my host may relate to application environment, workload, and OS, I'd caution against drawing strong quantitative conclusions from small samples. In particular, it seems the remaining noise content after controlling for known fixed effect is much larger for me than it was on S5R3.
While the particular amount is in considerable observational doubt, there is no doubt in my mind that the posted 6.05 ap is substantially faster than was the 6.04 on my Conroe-derived Windows XP SP3 hosts. I have three, and all are indicating substantial benefit by preliminary indications.
Edit: if you reply to this post, please be sure to delete the image link from any quoted material. It is only 13 kilobytes, but folks don't need to see it multiple times.
Sorry, folks, but I deserve
)
Sorry, folks, but I deserve an easy to install, reliable application. Otherwise I will not use any so called "power application" available, because I am not interested in a "power crash".
Electricity is exspensive in Germany today. The last power application for SR3 was very easy to install. If the project is interested in more crunching power, the programmers should use the same approach again.Give it a trial!
Intel Q9300 Quadcore, 2500 Mhz, 4096 MB RAM, GeForce 9800 GT, Vista Ultimate 64-bit, Ubuntu 10.10 64 bit
RE: Sorry, folks, but I
)
It looks like the app_info.xml that Bernd posted here should be fine. All you have to do is copy that and paste it in before you put it in the Einstein folder. More than likely Bernd will update the zip file to have the newer version in it, so you can also wait for that...
RE: Ok- I'm new to the
)
Easiest way is:
* Set Einstein@Home in BOINC Manager to "No new work". Let it finish, upload, and report current tasks. You can abort any unfinished work not yet started in advance.
* Stop BOINC
* Then unpack the contents of http://einstein.phys.uwm.edu/power_apps/einstein_S5R4_6.05_windows_intelx86.zip
(two executables and one app_info.xml file) into the einstein.phys.uwm.edu sub-dir of your BOINC installation (exact location differs with BOINC version).
* Start boinc again.
Now, if you want to be friendly to your wingmen, then you might prefer not to abort WUs in your pipeline.
In this case, you can alternatively:
* stop BOINC
* extract the zip into the sub-dir of BOINC as specified above.
* open app_info.xml that you just downloaded with a text editor, and replace its contents with the text Bernd provided here.
* start BOINC again
:EDIT:
BOINC will then use the new power user app for newly dowloaded WUs, no work needs to be aborted manually or will crash.
As always, please continue to watch this thread for news about the app and please report any problems here (including problems with the screensaver etc.).
Note there is a known problem with BOINC when you are using BOINC's CPU throttling feature (when you configure BOINC in the web interface to use less than 100% of idle CPU cycles). So for this test (well, for E@H in general really), please do not use CPU throttling.
BTW, what I've seen so far from this app performance-wise is really promising.
CU
Bikeman
RE: BOINC will then use the
)
I don't think so, Bikeman. Unlike the grandfathering ap_infos often provided previously, this one actually points to two different executables for the 604 and 605 cases. Bernd's comment seemed clearly to indicate that this was his intention--to use the older executable for work already branded.
Let's get this hashed out--if you are wrong then people may get a false impression of no improvement as they work their way through previously downloaded work. If I am wrong there are other consequences.
RE: RE: BOINC will then
)
You are absolutely right!! Sorry, I must have been too tired already :-), thanks!
CU
Bikeman
Wow- that sounds really easy.
)
Wow- that sounds really easy. Thanks!
I'll be back when it's done something.
Bikeman Thanks for
)
Bikeman
Thanks for confirming I got it right earlier.
ATM I am off crunching on another project for a few weeks. But, I will come back to Einstein and really get stuck in with this new client.
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
I'm unsure what
)
I'm unsure what happened.
I used the 'polite' .xml, and restarted BOINC.
It came up with a message saying to finish my work, remove the .xml, and restart. I had some 6.04's unstarted in the list.
I just renamed the .xml, and restarted. It just gave me some 6.04's to download.
OK- update: I'm not
)
OK- update:
I'm not downloading data.
I seem to be downloading the 6.04 programs.
Is that right?
I have four 6.05 pure results
)
I have four 6.05 pure results returned and validated from a Conroe-class quad I call Stoll4 (Q6600 mildly overclocked near 2.8 GHz).
As it has previously run a couple of hundred results from this frequency band covering well over a cycle, there is a decent reference both as to the natural systematic and random (including wiggle as part of random) variation.
My initial guess is that on this host for this type of work the new ap makes the host about 17% more productive, or that it takes about 85% as long as before.
While some of the random variation seen on my host may relate to application environment, workload, and OS, I'd caution against drawing strong quantitative conclusions from small samples. In particular, it seems the remaining noise content after controlling for known fixed effect is much larger for me than it was on S5R3.
While the particular amount is in considerable observational doubt, there is no doubt in my mind that the posted 6.05 ap is substantially faster than was the 6.04 on my Conroe-derived Windows XP SP3 hosts. I have three, and all are indicating substantial benefit by preliminary indications.
Edit: if you reply to this post, please be sure to delete the image link from any quoted material. It is only 13 kilobytes, but folks don't need to see it multiple times.