No Work for PPC?

Message boards : Number crunching : No Work for PPC?

To post messages, you must log in.

AuthorMessage
Gorbag

Send message
Joined: 6 May 07
Posts: 4
Credit: 570,524
RAC: 0
Message 58297 - Posted: 31 Dec 2008, 14:40:18 UTC

Since upgrading my G5 to BOINC 6.5.0, I\'ve been getting \"no work\" messages (but there is work for other platforms). Is this a version number thing or is there really nothing for G5s to do?
ID: 58297 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator
Project administrator

Send message
Joined: 22 Aug 06
Posts: 3542
Credit: 0
RAC: 0
Message 58306 - Posted: 31 Dec 2008, 16:42:21 UTC

I understand there are some scheduling issues with the new BOINC version. However, it would seem more likely that in this case, there is currently no work available for the PPC Macs. Work for this platform will be limited going forward. ...in fact I thought I had heard that BOINC was not going to support power PC going forward either.

Anyway, I\'d suggest you add other projects that have work for the PPC consistently. Rosetta will sometimes have work and sometimes not. So, if you remain attached to Rosetta as well, you can still help out when work is available. And BOINC will automatically request work from the other project(s) if there isn\'t any from Rosetta available for PPC.
Rosetta Moderator: Mod.Sense
ID: 58306 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
mikey
Avatar

Send message
Joined: 5 Jan 06
Posts: 1449
Credit: 5,776,642
RAC: 0
Message 58324 - Posted: 1 Jan 2009, 13:00:12 UTC - in response to Message 58306.  

I understand there are some scheduling issues with the new BOINC version. However, it would seem more likely that in this case, there is currently no work available for the PPC Macs. Work for this platform will be limited going forward. ...in fact I thought I had heard that BOINC was not going to support power PC going forward either.

Anyway, I\'d suggest you add other projects that have work for the PPC consistently. Rosetta will sometimes have work and sometimes not. So, if you remain attached to Rosetta as well, you can still help out when work is available. And BOINC will automatically request work from the other project(s) if there isn\'t any from Rosetta available for PPC.


So what is the reasoning behind this? IS the cpu on the PPC that different so as to give funky results? Does it take too long? Is it too hard to make the units? What really is the problem, inquiring minds would like to know, please.
ID: 58324 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
AMD_is_logical

Send message
Joined: 20 Dec 05
Posts: 299
Credit: 31,460,681
RAC: 0
Message 58337 - Posted: 1 Jan 2009, 22:06:09 UTC - in response to Message 58324.  

So what is the reasoning behind this? IS the cpu on the PPC that different so as to give funky results? Does it take too long? Is it too hard to make the units? What really is the problem, inquiring minds would like to know, please.


If you look at a chart of BOINC crunching power by host, such as http://boincstats.com/stats/host_cpu_stats.php?pr=bo&st=0 you\'ll see that the PPC is just a tiny fraction compared to x86 processors. And new, faster x86 processors are coming out, while the PPC numbers are shrinking as old machines are replaced with x86 ones.

On top of that, Rosetta always ran very slowly on the PPC. Some other projects worked much better with the PPC.

So it just isn\'t worth the effort to port minirosetta to the PPC, or to make sure each new version still works.
ID: 58337 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Paul D. Buck

Send message
Joined: 17 Sep 05
Posts: 815
Credit: 1,812,737
RAC: 0
Message 58347 - Posted: 2 Jan 2009, 4:57:55 UTC - in response to Message 58324.  

So what is the reasoning behind this? IS the cpu on the PPC that different so as to give funky results? Does it take too long? Is it too hard to make the units? What really is the problem, inquiring minds would like to know, please.


The Power PC is a radically different processor architecture. As such, along with the fact that it is hosting a different OS makes it a different world to stand up an application. So, with fewer of these machines around day to day as people upgrade, well, it can become a logical choice to drop them from the mix of computers to support.

My G5 when I was running it could on some projects outrun the Intel processors of nearly 4 years later and I did not even have the all around \"bet\" of the G5 processors.

As far as the results being \"funky\", well, the specification used to define numeric processing gives people the false sense that the results produced by two processors built to that standard will be the same. Well, not exactly. Both sets of results will be \"valid\", but different. That makes it harder to compare across processor lines. For example, LHC has such requirements for precision in comparison of results that they support very few architectures ... because of this difficulty. We even saw, from time to time differences and difficulties comparing from AMD to Intel in the early days (if memory serves).

Anyway, with fewer and fewer PPC around (I just stopped using mine for BOINC and gave it away) a project may decide not to support it any longer.

Interestingly to my mind, is that the WEP-M+2 project ran almost twice as fast on the G5 as it did on my Mac Pro of 4 years some later (top of the line Mac Pro I might add) ... meaning that a 2 GHz G5 beat a 3.2 GHz Intel Xeon ... YMMV
ID: 58347 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Keck_Komputers
Avatar

Send message
Joined: 17 Sep 05
Posts: 211
Credit: 2,949,189
RAC: 558
Message 58350 - Posted: 2 Jan 2009, 8:08:06 UTC - in response to Message 58347.  

For example, LHC has such requirements for precision in comparison of results that they support very few architectures ... because of this difficulty. We even saw, from time to time differences and difficulties comparing from AMD to Intel in the early days (if memory serves).

Your memory is correct. Eventually they had to provide their own math libs and bypass the standard libs to get results that matched cross platform.
BOINC WIKI

BOINCing since 2002/12/8
ID: 58350 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
mikey
Avatar

Send message
Joined: 5 Jan 06
Posts: 1449
Credit: 5,776,642
RAC: 0
Message 58358 - Posted: 2 Jan 2009, 12:04:41 UTC - in response to Message 58337.  

So what is the reasoning behind this? IS the cpu on the PPC that different so as to give funky results? Does it take too long? Is it too hard to make the units? What really is the problem, inquiring minds would like to know, please.


If you look at a chart of BOINC crunching power by host, such as http://boincstats.com/stats/host_cpu_stats.php?pr=bo&st=0 you\'ll see that the PPC is just a tiny fraction compared to x86 processors. And new, faster x86 processors are coming out, while the PPC numbers are shrinking as old machines are replaced with x86 ones.

On top of that, Rosetta always ran very slowly on the PPC. Some other projects worked much better with the PPC.

So it just isn\'t worth the effort to port minirosetta to the PPC, or to make sure each new version still works.


That makes sense, thanks
ID: 58358 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
mikey
Avatar

Send message
Joined: 5 Jan 06
Posts: 1449
Credit: 5,776,642
RAC: 0
Message 58359 - Posted: 2 Jan 2009, 12:05:05 UTC - in response to Message 58347.  

So what is the reasoning behind this? IS the cpu on the PPC that different so as to give funky results? Does it take too long? Is it too hard to make the units? What really is the problem, inquiring minds would like to know, please.


The Power PC is a radically different processor architecture. As such, along with the fact that it is hosting a different OS makes it a different world to stand up an application. So, with fewer of these machines around day to day as people upgrade, well, it can become a logical choice to drop them from the mix of computers to support.

My G5 when I was running it could on some projects outrun the Intel processors of nearly 4 years later and I did not even have the all around \"bet\" of the G5 processors.

As far as the results being \"funky\", well, the specification used to define numeric processing gives people the false sense that the results produced by two processors built to that standard will be the same. Well, not exactly. Both sets of results will be \"valid\", but different. That makes it harder to compare across processor lines. For example, LHC has such requirements for precision in comparison of results that they support very few architectures ... because of this difficulty. We even saw, from time to time differences and difficulties comparing from AMD to Intel in the early days (if memory serves).

Anyway, with fewer and fewer PPC around (I just stopped using mine for BOINC and gave it away) a project may decide not to support it any longer.

Interestingly to my mind, is that the WEP-M+2 project ran almost twice as fast on the G5 as it did on my Mac Pro of 4 years some later (top of the line Mac Pro I might add) ... meaning that a 2 GHz G5 beat a 3.2 GHz Intel Xeon ... YMMV


That makes sense too, thanks
ID: 58359 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
kh

Send message
Joined: 14 Feb 09
Posts: 1
Credit: 10,642
RAC: 0
Message 59809 - Posted: 26 Feb 2009, 4:26:06 UTC

Since Rosetta@home really doesn\'t support the Mac/PPC architecture any more, I respectfully request that you update this page by removing the reference to supporting the Mac OS X PPC architecture:

http://boinc.berkeley.edu/projects.php

This is one of the first pages people interested in BOINC are likely to see and what a waste of time it could be for them trying to get work on a project that doesn\'t support their machine. The message posted in the BOINC Manager doesn\'t imply that work will never be available.

For the record, I haven\'t processed many Rosetta@home work units but my G5 sure chews through the ones I did get faster than an Intel machine on the same network (the Intel machine is faster in MHz than the PPC, too). PPC didn\'t look to be at a disadvantage at all in this project.
ID: 59809 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator
Project administrator

Send message
Joined: 22 Aug 06
Posts: 3542
Credit: 0
RAC: 0
Message 59815 - Posted: 26 Feb 2009, 15:23:35 UTC

The message posted in the BOINC Manager doesn\'t imply that work will never be available.


...that\'s because there will be work available, from time to time. It would be incorrect to say that work will never be available.

This page had more room for details and was updated:
http://boinc.berkeley.edu/wiki/List_of_projects_by_system_requirements
Rosetta Moderator: Mod.Sense
ID: 59815 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
mwgiii

Send message
Joined: 29 Sep 05
Posts: 3
Credit: 90,006
RAC: 0
Message 60256 - Posted: 21 Mar 2009, 15:16:34 UTC

Since there might be occasional work available, I will leave my PPC Mac attached to Rosetta.

While I am waiting I will keep my PPC Mac busy with World Community Grid and Superlink
ID: 60256 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Dotsch
Avatar

Send message
Joined: 12 Feb 06
Posts: 102
Credit: 139,340
RAC: 22
Message 60271 - Posted: 22 Mar 2009, 12:23:01 UTC - in response to Message 60256.  
Last modified: 22 Mar 2009, 12:23:12 UTC

Since there might be occasional work available, I will leave my PPC Mac attached to Rosetta.

While I am waiting I will keep my PPC Mac busy with World Community Grid and Superlink

I think SIMAP, POEM and Lattice are also very interesting projects wich has MacOS PPC support.
ID: 60271 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
David Riese

Send message
Joined: 9 Dec 05
Posts: 3
Credit: 40,639,962
RAC: 34,590
Message 63353 - Posted: 15 Sep 2009, 0:25:20 UTC - in response to Message 60271.  
Last modified: 15 Sep 2009, 0:34:13 UTC

Since there might be occasional work available, I will leave my PPC Mac attached to Rosetta.

While I am waiting I will keep my PPC Mac busy with World Community Grid and Superlink

I think SIMAP, POEM and Lattice are also very interesting projects wich has MacOS PPC support.


I have three multi-processor G5 PowerMacs and several G4 Macs (mostly PowerBooks). It has been months since work has been available for these units. Because of this, I\'d like to echo the request that Rosetta discontinue advertising support for the OS X/PPC platform. It would appear that the \"gradual phase-out\" is, for all practical purposes, complete. It\'s too bad - those units were doing a lot of crunching for Rosetta (roughly 15% of my total).

IMHO, SIMAP is not a particularly appealing alternative - work is available only very periodically. My PPC units are primarily running SETI and Einstein.
ID: 63353 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
nick n
Avatar

Send message
Joined: 26 Aug 07
Posts: 49
Credit: 219,102
RAC: 0
Message 63387 - Posted: 18 Sep 2009, 2:54:47 UTC

Just put them on world community grid. They have alot of nice projects and I am assuming they will support PPC\'s for a very long time since IBM is running it.
ID: 63387 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
mwgiii

Send message
Joined: 29 Sep 05
Posts: 3
Credit: 90,006
RAC: 0
Message 63681 - Posted: 14 Oct 2009, 3:08:09 UTC

My lowly ole G5 has been getting steady work from Rosetta for the last week or so. Hopefully this flow will continue.
ID: 63681 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Message boards : Number crunching : No Work for PPC?



©2019 University of Washington
http://www.bakerlab.org