Welcome to the MacNN Forums.

If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below.

You are here: MacNN Forums > Community > Team MacNN > New Altivec-enhanced Seti worker in need of testing

New Altivec-enhanced Seti worker in need of testing (Page 10)
Thread Tools
Drash
Fresh-Faced Recruit
Join Date: Aug 2001
Location: UK
Status: Offline
Dec 7, 2005, 11:03 PM
 
Originally Posted by Todd Madson
Yikes. Now I have 35 blocks waiting to upload. When will it end?
I've got 59 (and 10+ to download ???)
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 8, 2005, 01:40 PM
 
Not any time soon, the way it looks. I checked the server status page a number of times today, and it seems that WUs are currently being dished out four times faster than the results waiting for validation are accumulating. With all the validators currently off, this is indicative of four times slower result uploads than WU downloads, no matter what the technical news say. The current situation will result in quite a log jam if they don't fix the upload problem very soon...
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 8, 2005, 01:43 PM
 
Originally Posted by mikkyo
Current superbench Beta clients shouldn't work in the GUI apps.
You can try, but I doubt you'll have much success.
Mikkyo: I have been running your 'possibly GUI-Compatible' CLI 5.2.13 Client that you have posted since you wrote the above inside stock Menubar 5.2.13 for a good number of hours now, and all seems well (I'm using the G5-version on a DP 2.5 GHz). Thanks for posting these!

What's the difference between the CLI client and the one used in the GUI-apps?

Curiously,

Ron

PS: An interesting bit for E.T - the correct amount of RAM is now reported. Thanks for bringing this to light!
( Last edited by halimedia; Dec 8, 2005 at 01:58 PM. )
     
BTBlomberg
Forum Regular
Join Date: Sep 2005
Location: Chicago Suburbs
Status: Offline
Dec 8, 2005, 01:56 PM
 
It's going to create a problem once they get it fixed as the Finished Work Units that are backlogged will take a while to all get through. I know I have had several times over the last few days in which my clients were waiting on downloading WUs. Problem is the fdaster your machine at cruching the greater your backlog of WUs will be and the more likely you will have idle time for SETI. Sounds like they are working on a jury rigged fix for now, but the solution may take some time. Expect the backlog to grow before it gets better.
     
mikkyo
Senior User
Join Date: Feb 2002
Location: Silly Valley, Ca
Status: Offline
Dec 8, 2005, 05:54 PM
 
Originally Posted by halimedia
What's the difference between the CLI client and the one used in the GUI-apps?

PS: An interesting bit for E.T - the correct amount of RAM is now reported. Thanks for bringing this to light!
The clients are linked with all the GUI libs, making them bigger and slower.

Yes, the RAM is reported correctly in those clients now.
Hopefully that will help some folks.
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 8, 2005, 06:15 PM
 
Originally Posted by BTBlomberg
Problem is the fdaster your machine at cruching the greater your backlog of WUs will be and the more likely you will have idle time for SETI.
Absolutely! The three dual-cpu boxes in my zoo have a combined upload-backlog of over 120 results, while the single-cpu machines are more or less able to keep up.

I think what the folks at SETI should do ist to limit the number of allowed connections on the download server for the uploads to catch up - which they can now easily do, as they have different machines handling up- and downloads. It looks like many of my results get partially uploaded and then the connection is dropped. This wastes lots of bandwidth, I'm quite sure.

But the # of connections the web servers are able to handle is probably just one part of the equation. They likely also have problems with their NFS shares again, as they did previously.

Oh, well. All we can do is wait for the storm to clear - an maybe join another project in the meantime. Speaking of which (and waaay off-topic, I know): has the OS X CPDN Sulphur client improved? I tried early versions, and they repeatedly tanked out at around 2-3 %, no matter what hardware they ran on (PowerBook or dual-G5) - so I abandoned the project, although I think it's very worth while.

Cheers,

Ron
     
mikkyo
Senior User
Join Date: Feb 2002
Location: Silly Valley, Ca
Status: Offline
Dec 8, 2005, 06:54 PM
 
I hope you all have attached to Einstein, so that when they take the Seti servers offline, like they always do to catch up, you can help our Einstein team in the interim.
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 13, 2005, 07:11 AM
 
Things are rolling again! Happy crunching, everyone!
     
Todd Madson
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status: Offline
Dec 13, 2005, 01:35 PM
 
Indeed, we are rolling. Onwards!
     
MacStef
Fresh-Faced Recruit
Join Date: Jun 2005
Status: Offline
Dec 14, 2005, 05:39 PM
 
since installing this new S@H my Einstein@home has given nothing but client errors. Is it me or is something a miss.

G4 DP 1GHz 1.5 GB Ram
Mac OSX 10.3.9
Boinc Manager 4.44 (Info actually says Version 4.43)
Einstein 4.82
not sure of Seti version...no WUs have downloaded, but pulled for your link at beginning of this thread.

http://einstein.phys.uwm.edu/workunit.php?wuid=2854690
http://einstein.phys.uwm.edu/workunit.php?wuid=2854365

Is something missing? Did I install something wrong? Don't really know other then this has screwed up my einstein@home.

A little help will be great.
thx
MacStef

PS...with the above machine, what do you recommend for my; Boinc, Seti and Einstein?
( Last edited by MacStef; Dec 14, 2005 at 11:46 PM. )
     
Andrew F
Guest
Status:
Dec 14, 2005, 06:32 PM
 
Originally Posted by halimedia
...Speaking of which (and waaay off-topic, I know): has the OS X CPDN Sulphur client improved? I tried early versions, and they repeatedly tanked out at around 2-3 %, no matter what hardware they ran on (PowerBook or dual-G5) - so I abandoned the project, although I think it's very worth while. ~Ron
Yes, Ron, the v4.22 CPDN sulphur model came out last week. I had many problems on my dual Macs with previous CPDN, but this one is working well for me now. They are huge WU, so in 5 days of processing, I'm only at 5.031% complete, but I'm getting some trickle credit. (Note, this wasn't full machine time, I'm sharing BOINC time with SETI, Einstein, SZTAKI, and Predictor.) Be sure to install the Mac fortran library before attaching to the project, this doesn't download automatically and the processing won't start correctly without it.

The CPDN team has acquired a dual G5, so we can expect OSX clients to be developed.

Also, I encouraged the CPDN staff to add an "apps" page to the website like most other projects have, so users can reference the latest version and release dates. Be sure to check on that from time to time.
     
MacStef
Fresh-Faced Recruit
Join Date: Jun 2005
Status: Offline
Dec 15, 2005, 05:03 AM
 
Have now completed a third Einstein WU and again Client Error. Would like to help out but this Boinc/Seti/Einstein/everything else I have loaded on my Mac since Seti switched over has been nothing but a pain in the arse. I am deleting all this BS and will dedicate to doing something more constructive...like down loading porn....LoL.
Anyway, if anyone has some useful information that may help figure out what the hell is going on with my Boinc, I would be greatly appreciative, till then I am shutting down over from now on.
Thanks.
MacStef
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 15, 2005, 12:34 PM
 
MacStef: Just a thought - temporarily suspend one of the other projects at a time to see which one might be affecting Einstein. I'm currently crunching SETI (alpha-5) and CPDN, and they don't seem to get in each other's way...

HTH,

Ron
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 15, 2005, 12:41 PM
 
Andrew F: Thanks for the feedback regarding CPDN! The models I have started during the SETI-outage have now passed 5 %. All seems well, so far. Good to hear that they're taking their Mac-development more seriously

Impatiently waiting for my Quad to arrive (tap, tap, tap....)
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 18, 2005, 05:39 AM
 
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 18, 2005, 10:32 AM
 
A question to all G5 crunchers: I'm interested in the difference low-latency RAM makes (CL 2.5 or 3 vs. CL 4) when crunching SETI. If you know what your RAM's latency is, could you post that information, along with amount of RAM, G5 specs and a link to your machine's summary on the SETI pages? I'm interested in comparing WU completion times. If enough folks post their results, I'll write up a comparison...

I'll start:

Model: PM G5 DP 2.5 GHz
RAM: 5 GB PC3200U-30330 (that makes it CL3 - 3.0-3-3.0) (4 x 1 GB, 4 x 256 MB)
OS: 10.4.3
Energy Saver: highest (Processor Performance)
BOINC: 5.2.13 Menubar Superbench beta
SETI: alpha-5
URL: http://setiweb.ssl.berkeley.edu/show...hostid=1116838

TIA for any replies!


Edit: I just realized that you can use the System Profiler (at least in Tiger - don't know about Panther) to get the memory timings. So, even if you don't know from purchasing, System Profiler will know the timings. I changed my info above accordingly.

Edit 2: Added my Energy Saver prefs after I realized what a huge difference the setting on a new iMac G5 makes (40 % increase in WU time when going from highest to automatic!!!).
( Last edited by halimedia; Dec 18, 2005 at 12:07 PM. )
     
E.T from tellus
Fresh-Faced Recruit
Join Date: Nov 2005
Location: Finland, Tuusula
Status: Offline
Dec 18, 2005, 12:36 PM
 
Memory performance is major bottlenecks in G5.
Because of OS X nature to handle processes (circulate between processors) Memory bandwith is critical.
In most cases Altivec FPU can crunch faster than system feeds data Low latency memory increases performance almost linear by numbers. Fine tuning chunking algorithms can give amazing performance boost.
Anyway my quad G5
http://setiathome.berkeley.edu/show_...hostid=1781916
Model: PM G5 QP 2.5 Ghz
RAM: 8 x 1GB PC4200U-40440 Valueram
OS:10.4.3
Enegysaver: Highest
BOINC: 5.2.13 CLI
Seti: alpha-5
( Last edited by E.T from tellus; Dec 18, 2005 at 12:44 PM. )
     
RandC
Fresh-Faced Recruit
Join Date: Dec 2005
Status: Offline
Dec 18, 2005, 01:53 PM
 
Just received QUAD and have seti@home and einstein@home running. Originally had processor set to AUTOMATIC with alpha-4 and was running seti WU in about 3000 seconds. Yesterday changed to alpha-5 and HIGHEST and now running seti WU in about 2500 seconds. None of the alpha-5 processed WU's have credit yet though.

Model: PM G5 QUAD 2.5 GHz
RAM: 5 GB PC2-4200E-444 (2 x 512 MB, 4 x 1GB all EEC)
OS: 10.4.3
Energy Saver: highest (Processor Performance)
BOINC: 5.2.13 Manager Superbench beta
SETI: alpha-5
URL: http://setiathome.berkeley.edu/show_...hostid=1935172

Rand
( Last edited by RandC; Dec 18, 2005 at 04:24 PM. )
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 18, 2005, 02:25 PM
 
Thanks for the input! I will start compiling the info...

E.T: I noticed you also have an Xserve running SETI, and so do I. Yours is always a lot faster than mine, and I'm curious why. Could you post the info for it, if you don't mind?

Here's mine:

Model: Xserve G5 DP 2.0 GHz
RAM: 3 GB (2 x 512 MB PC3200U-30330; 4 x 512 MB PC3200U-30440 - all ECC)
OS: Server 10.3.9
Energy Saver: n/a (always highest, I assume)
BOINC: 4.44 CLI Superbench
SETI: alpha-4
URL: http://setiweb.ssl.berkeley.edu/show...hostid=1414507
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 18, 2005, 02:38 PM
 
Originally Posted by E.T from tellus
Low latency memory increases performance almost linear by numbers. Fine tuning chunking algorithms can give amazing performance boost.
I've recently read a number of comparative G5 RAM tests. While low-latency RAM seems to score better in benchmark tests, it appears to make little to no difference in real-world tests. Here are two examples:

http://www.barefeats.com/quad02.html
(Scroll down to DOES THE QUAD-CORE's MEMORY SPEED HELP?)

http://www.barefeats.com/memory.html

Does anyone have any more detailed info on this (G5-specific or not)?

Cheers,

Ron
( Last edited by halimedia; Dec 18, 2005 at 03:40 PM. )
     
E.T from tellus
Fresh-Faced Recruit
Join Date: Nov 2005
Location: Finland, Tuusula
Status: Offline
Dec 18, 2005, 03:55 PM
 
In our case (crunching seti) memory bandwith is issue, normally computation is case now it is memory.
one of my other computer
http://setiathome.berkeley.edu/show_...hostid=1789945
is Dual 2.7Ghz G5 one WU took avg 4 100sec
and my 2.5Ghz quad needs only 2 400sec for same work

So what is changed between those computers
1) Quad has double size L1 cache 1024mb
2) Quad has faster memory bus 533mhz DDR2 vs 400mhz DDR
3) Dual 2.7Ghz G5 has faster processor -> in theory one WU should be crunched faster
Scientific computing has little bit different needs from platform than office use
In ideal case dedicating processor/seti process gives best results -> OS X circulating processes between processors, memory bandwith gives little bit penalty
Our experts has done amazing job to optimize seti crunchers, they have found limits of platform.
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 19, 2005, 07:20 AM
 
Originally Posted by E.T from tellus
Our experts has done amazing job to optimize seti crunchers, they have found limits of platform.
That's exactly why I'm interested in gathering real-world data regarding memory latency. I have a hunch that low-latency RAM will make a difference when crunching SETI with the optimized workers, but the proof is in the pudding. Any chance of getting your Xserve info (see my post above)? And details on the DP 2.7 GHz would be great, too!

Cheers,

Ron
( Last edited by halimedia; Dec 19, 2005 at 07:29 AM. )
     
E.T from tellus
Fresh-Faced Recruit
Join Date: Nov 2005
Location: Finland, Tuusula
Status: Offline
Dec 19, 2005, 10:30 AM
 
Here is
Model: Xserve G5
CPU: 2 x 2.3GHz
RAM: 8 x 1Gb DDR ECC PC 3200U-30330
OS: OS X 10.4.3 server
Energy Saver: n/a (highest)
BOINC: 5.2.13 CLI Superbench
SETI: alpha-5
http://setiathome.berkeley.edu/show_...hostid=1436815

Model: Power Mac G5
Cpu: 2 x 2.7Ghz
RAM: 8 x 1Gb DDR PC 3200U-30330
OS: OS X 10.4.3
Energy Saver: highest
BOINC: 5.2.13 CLI Superbench
SETI: alpha-5
( Last edited by E.T from tellus; Dec 19, 2005 at 10:38 AM. )
     
TiloProbst
Junior Member
Join Date: Jul 2005
Status: Offline
Dec 19, 2005, 10:02 PM
 
hey people .. sry for that noob question, but how do I install that alpha 5 version from http://writhe.org.uk/seti@home/ ?

unzipping gives me an .xml file and some seti@home-G5-a5 file. where do I put those?

currently I am running the boinc4 application from http://members.dslextreme.com/~reade...eam/boinc.html with

/Applications/Boinc4/boinc_4.44 -dir /Applications/Boinc4/ -attach_project http://setiathome.berkeley.edu/ <project id>

from the Terminal.

thx for your help, I cant wait to improve the numbers of my Dual 2GHz G5 from those crappy 240 Recent Average I achieve now.
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 20, 2005, 04:23 AM
 
Originally Posted by TiloProbst
unzipping gives me an .xml file and some seti@home-G5-a5 file. where do I put those?
First, stop boinc. Then, put both files into .../[your boinc folder]/projects/setiathome.berkeley.edu/

You may also want to move the stock worker (called setiathome_4.18_powerpc-apple-darwin) to a backup directory somewhere.

That's it! Happy crunching!

Ron
     
TiloProbst
Junior Member
Join Date: Jul 2005
Status: Offline
Dec 20, 2005, 07:59 AM
 
ok thx, do I have to change anything in my Terminal startup line?

currently it is: /Applications/Boinc4/boinc_4.44 -dir /Applications/Boinc4/ -attach_project http://setiathome.berkeley.edu/ *project_id*

besides from better credit results is there any way to prove, that the new g5 optimized cruncher is used instead of the default one?

another question: in *my boinc folder*/slots/0/ and .../1/ there seem to be old cruncher apps too (name: setiathome_4.18_powerpc-apple-darwin). do I have to replace them too? what is the purpose of those four empty folders ../slots/2/ to /slots/5/ ?

UPDATE
ok the seti app in boinc4/slots/0/ seems to get automatically replaced with the Alpha5 app

i was bothered by the Terminal saying
"Found app_info.xml; using anonymous plattform"

but later
"Resuming computation for result (...) using setiathome version 4.18"
( Last edited by TiloProbst; Dec 20, 2005 at 08:27 AM. )
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 20, 2005, 08:58 AM
 
Originally Posted by TiloProbst
ok thx, do I have to change anything in my Terminal startup line?

currently it is: /Applications/Boinc4/boinc_4.44 -dir /Applications/Boinc4/ -attach_project http://setiathome.berkeley.edu/ *project_id*
You don't need to issue '-attach_project' every time. You only need to do that once to attach the project to your boinc client when you set it up on a machine for the first time. The client then creates the corresponding directory in .../[your boinc folder]/projects/ and starts populating it after communicating with that project's servers.

You only need to issue the command again if you want do add an additional project (e.g. CPDN, Einstein, etc.) to a particular machine.

besides from better credit results is there any way to prove, that the new g5 optimized cruncher is used instead of the default one?
Yes, check Activity Monitor (or top in the Terminal). Instead of 'setiathome', the process should now be called 'seti@home-G5-a5' (on a G5). And - as you said - you may just be able to tell by the *slight* reduction in work unit times...

another question: in *my boinc folder*/slots/0/ and .../1/ there seem to be old cruncher apps too (name: setiathome_4.18_powerpc-apple-darwin). do I have to replace them too? what is the purpose of those four empty folders ../slots/2/ to /slots/5/ ?
The slot dirs are used to store data pertaining to your most recent computations. Leave these as they are - the BOINC client manages them for you.

i was bothered by the Terminal saying
"Found app_info.xml; using anonymous plattform"
"Resuming computation for result (...) using setiathome version 4.18"
No prob, that's what you should see when you use an 'unofficial' worker...

HTH,

Ron
     
TiloProbst
Junior Member
Join Date: Jul 2005
Status: Offline
Dec 20, 2005, 10:07 AM
 
Originally Posted by halimedia
You don't need to issue '-attach_project' every time. You only need to do that once to attach the project to your boinc client when you set it up on a machine for the first time. The client then creates the corresponding directory in .../[your boinc folder]/projects/ and starts populating it after communicating with that project's servers.

You only need to issue the command again if you want do add an additional project (e.g. CPDN, Einstein, etc.) to a particular machine.
k


Yes, check Activity Monitor (or top in the Terminal). Instead of 'setiathome', the process should now be called 'seti@home-G5-a5' (on a G5). And - as you said - you may just be able to tell by the *slight* reduction in work unit times...
yes, Activity Monitor says setiªhome-G5-a5 *g* now lets hope for better results


The slot dirs are used to store data pertaining to your most recent computations. Leave these as they are - the BOINC client manages them for you.
k

No prob, that's what you should see when you use an 'unofficial' worker...
hmm, ok, but why it still says:

[SETI@home] Starting result (...) using setiathome version 4.18 <- that is just incorrect, where does Boinc get this from?

btw, can I stay with 4.44 Boinc, or should I upgrade?
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 20, 2005, 10:23 AM
 
Originally Posted by TiloProbst
hmm, ok, but why it still says:

[SETI@home] Starting result (...) using setiathome version 4.18 <- that is just incorrect, where does Boinc get this from?
My guess is that the optimized worker is based on the 4.18 code-base released by SETI. But Rick or Alex would have to give you the specifics. All I can say is that it's normal to see this in the log...

btw, can I stay with 4.44 Boinc, or should I upgrade?
I'm having some issues with BOINC Menubar 5.2.13 Superbench (beta). I'd therefore recommend sticking with 4.44, although I don't know if there are any probs with the CLI-version of 5.2.13.

I suggest you check out this thread and ask there:

http://forums.macnn.com/showthread.php?t=274161

HTH,

Ron
     
rick
Fresh-Faced Recruit
Join Date: Sep 2005
Status: Offline
Dec 20, 2005, 11:35 AM
 
Originally Posted by halimedia
First, stop boinc. Then, put both files into .../[your boinc folder]/projects/setiathome.berkeley.edu/
I've been meaning to put up a concise installation guide on the web site for a while. You posted what I consider the definitive version halfway down page 7. Can I steal this to put on the site (with appropriate mad props)?

Originally Posted by halimedia
My guess is that the optimized worker is based on the 4.18 code-base released by SETI. But Rick or Alex would have to give you the specifics. All I can say is that it's normal to see this in the log...
Yeah, our code is based on 4.18. I've just had a look at the latest official code at it still says the version is 4.18. I doubt they'd alter the functionality or analysis algorithms. The version number doesn't really mean much at this point: I doubt they'd alter the functionaity or analysis algorithms.

They might update the version number for the forthcoming SETI@home enhanced.
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 20, 2005, 11:52 AM
 
Originally Posted by rick
I've been meaning to put up a concise installation guide on the web site for a while. You posted what I consider the definitive version halfway down page 7. Can I steal this to put on the site (with appropriate mad props)?
I'd be honored! Feel free to use and change it as you see fit. Bear in mind, though, that it is strongly geared towards BOINC Menubar. You may want to provide different and tailored instructions for Manager, Menubar and CLI...

Just a thought...

Ron
     
rick
Fresh-Faced Recruit
Join Date: Sep 2005
Status: Offline
Dec 20, 2005, 02:13 PM
 
Zut alors! E.T is now second with Karl coming in a close fourth.

Nice.
     
E.T from tellus
Fresh-Faced Recruit
Join Date: Nov 2005
Location: Finland, Tuusula
Status: Offline
Dec 20, 2005, 04:25 PM
 
Thanks Rick
I think we all should thank you and Alex for amazing job. Anyway if you coming to Macworld Expo in SFO i buy you a dinner
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 21, 2005, 04:25 AM
 
E.T is number one! Congratulations! May the aliens come!
     
Todd Madson
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status: Offline
Dec 21, 2005, 04:34 PM
 
Hummina.

My G5 2.5 DP with 2.5 GB ram runs with processor prefs set to highest and I
also have nap mode disabled with the optimized client installed for the menubar
version.

It's weird how fast it seems to process work units now, sometimes so fast I
can't believe it.

I'll check my ram speed at home but it was two matched pairs of ValueSelect
and the original 256 megs of ram the machine had when I bought it.
     
E.T from tellus
Fresh-Faced Recruit
Join Date: Nov 2005
Location: Finland, Tuusula
Status: Offline
Dec 22, 2005, 09:13 AM
 
Karl you are almost there
     
Karl Schimanek
Junior Member
Join Date: Oct 2004
Location: Germany
Status: Offline
Dec 22, 2005, 04:17 PM
 
Yes
But i have to many client errors
I wonder why
I'm waiting for alpha6 to switch to the BOINC5 Manager.

Regards
Karl

P.S. I fear, SGI strikes back
     
Karl Schimanek
Junior Member
Join Date: Oct 2004
Location: Germany
Status: Offline
Dec 23, 2005, 03:17 PM
 
     
rick
Fresh-Faced Recruit
Join Date: Sep 2005
Status: Offline
Dec 23, 2005, 03:49 PM
 
Originally Posted by Karl Schimanek
SETI@home error -6 Bad workunit header
!swi.data_type || !found || !swi.nsamples
File: /Users/alexkan/Desktop/new-seti/seti_apple/client/seti_header.cpp
Line: 193
Well goddamn if that isn't an interesting error. Have you got any copies of the work units? It looks as though all the other people who got the same work unit also had client errors. Maybe SETI@home is sending out duff work units...?

I tried using SETI@home Grabber on the work units you posted but it looks as though the data server is broken or something.

In summary, I suspect it's not our fault or yours.



On another note, I was trying to post this earlier today but the forums were broken (argh! what does work!?!):

Originally Posted by Karl Shimanek
I'm waiting for alpha6 to switch to the BOINC5 Manager.
Bad news on that front I'm afraid. I'm experiencing a few technical difficulties with my PowerBook. The lower memory slot finally spazzed out today and I get system freezes when running intensive tasks (i.e. benchmarking new SETI@home code).

I was hoping to get alpha 6 out before Christmas but it just ain't gonna happen now. I've got all of next week off work though and a couple of people have said they'll give me remote access to their G5s so I should be able to test the new code remotely. Alex is also around now so perhaps he can figure something out.

New ETA: new year. Ish.

Maybe.
Since trying to post this, the lower memory slot in my PowerBook has completely died. But the upper one seems to work and I've just tried running some new SETI@home code and it didn't crash... guess I'll see how it goes...
     
Lauger
Fresh-Faced Recruit
Join Date: Nov 2005
Status: Offline
Dec 23, 2005, 11:21 PM
 
Model: PM G5 DP 2.0 GHz
RAM: 1 GB PC3200U-30330 (4 x 256 MB)
OS: 10.4.3
Energy Saver: highest (Processor Performance)
BOINC: 5.2.5 Menubar - stock
SETI: alpha-5
URL: http://setiweb.ssl.berkeley.edu/show...ostid= 1596056
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 24, 2005, 06:43 AM
 
Originally Posted by rick
Maybe SETI@home is sending out duff work units...?
Rick, I agree - they have been sending out quite a few duds (i.e. WUs that cause client errors on all clients that recieve it) lately. For me, this started when they were recovering from the recent outage a couple of weeks ago. In some cases, the duds even choked the BOINC client, and it had to be restarted.

Here's a recent dud that ran on a machine I brought on-line just two days ago (Xserve G5 single 2 GHz):

http://setiweb.ssl.berkeley.edu/work...?wuid=42924677
( Last edited by halimedia; Dec 24, 2005 at 10:31 AM. )
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 24, 2005, 10:28 AM
 
Merry Christmas, Esa and Karl! You made it!

A RAC of >3k? Truly impressive! Kudos to Alex and Rick again - and to Apple for giving us a great final product of the PPC-lineage!!

Happy Holidays to everyone!

Cheers,

Ron
( Last edited by halimedia; Dec 24, 2005 at 11:14 AM. )
     
Todd Madson
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status: Offline
Dec 25, 2005, 11:01 PM
 
WOW.

This afternoon before Christmas dinner I decided, just for fun, to haul out my G4/400
box to see how it would do in a race against my overclocked Athlon 2400 box running
at 2.2 ghz (running Xandros Linux). I grabbed the Naparst optimized code and installed
it and then installed the graphical client on the G4 and gave each machine 1 gigabyte of
PC133 ram to make it an even race as far as ram goes.

Here's the Athlon stats:
http://setiathome.berkeley.edu/resul...hostid=1934731

The optimized client would be starting from around 2 PM CST 25Dec05

Here's the G4/400 stats:
http://setiathome.berkeley.edu/resul...hostid=2006261

VERY surprising. So far the G4 has thrown two blocks both in the
10K seconds range (shocking really, the guesstimate said to expect
around 28 hours per block!).

The Athlon has been saying 2-3 hours but looking at the stats
it's between 11-14,000 seconds.

Weird.

That Altivec Optimized G4 code is FREAKING UNBELIEVABLE.

My G5 dual is still beating the pants off of everyone in my team put
together currently but the optimization made a machine I wasn't
sure would work well very useable.

Now we will have to see how my wife's G4 1.25 ghz machine goes..
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 26, 2005, 09:35 AM
 
A question to Esa and Karl: how much heat are your Quads generating when crunching SETI? Is it like having a space heater running all the time? My DP 2.5 GHz is creating quite a bit of heat when crunching, and I'm wondering how much 'hotter' the Quad is going to be...

TIA for any feedback!

Cheers,

Ron
     
Todd Madson
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status: Offline
Dec 26, 2005, 11:41 AM
 
Halimedia: My 2.5 dual is generating heat but in the winter we keep the house
temperatureat a reasonable level anyway so it isn't so bad. It's winter here and
the 2.5 dual, a Pentium M laptop and my wife' iMac lamp style is in the same room.

When it's cold out we shut the doors in the room and it gets very cozy in here.

Last summer I just kept Nap on and it wasn't bad at all but I can see with nap
disabled this coming spring/summer could get interesting. Luckily we do have
central air.

Of course, down in the basement where it gets as low as 40 degrees fahrenheit
might be a better location for the G5 in the future...

And:
http://setiathome.berkeley.edu/resul...hostid=2006261

Here we are with more work units generated by the G4/400 -
FIVE work units done in 10,500 seconds or thereabouts.

The Athlon is still doing them around 11-14,000. So the
CPU that is 1800 mhz faster has poorer performance!
     
virex
Junior Member
Join Date: Feb 2001
Location: Reading
Status: Offline
Dec 26, 2005, 11:58 AM
 
todd, you should really consider getting the optimized client running on that g4. It's only claiming 5-6 credits

     
Todd Madson
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status: Offline
Dec 26, 2005, 12:05 PM
 
Virex: It is on there, but I just started it yesterday afternoon as opposed to the Athlon
which has been running this for a while - probably up to 14 days.

I didn't run the G4 at all on it since I didn't think it was worth it, how wrong I was...

Now I'm testing the Wife's G4/1.2 ghz to see how that goes.
( Last edited by Todd Madson; Dec 26, 2005 at 12:20 PM. )
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 26, 2005, 02:55 PM
 
Originally Posted by Todd Madson
My 2.5 dual is generating heat but in the winter we keep the house
temperatureat a reasonable level...
Todd, thanks for your feedback. I, too, have a DP 2.5 GHz G5 crunching, and am well aware of the heat it produces (and consider it somewhat problematic for use in a confined room insulated to Central/Northern European construction standards). Since I'm expecting delivery of a Quad over the next few days, I'm wondering wether I should expect the need to run A/C even during winter time - something I'd much rather avoid, needless to say. That's why I geared my question specifically to Quad owners (Esa and Karl).

Regards,

Ron
     
Karl Schimanek
Junior Member
Join Date: Oct 2004
Location: Germany
Status: Offline
Dec 26, 2005, 03:39 PM
 
Hello-hello :-)

Merry xMas@all!
The CPU temperature of each core is around 65°C when crunching SETI. On idle around 50°C. Both on maximum (energy settings).

The generated heat by crunching is nice and warm
I think around 35°C (on the back)

Regards
Karl

P.S. The SGI passed me
     
halimedia
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Dec 26, 2005, 03:45 PM
 
Originally Posted by Karl Schimanek
The CPU temperature of each core is around 65°C when crunching SETI. On idle around 50°C. Both on maximum (energy settings).

The generated heat by crunching is nice and warm
I think around 35°C (on the back)
Thanks for the info! Do you have any means to empirically compare the heat output by the fans
to a DP G5?

P.S. The SGI passed me
As a small solace, you will always beat that monster in sheer size (from a space-saving perspective) and by the fact that your Quad only has an eighth of the CPUs the SGI box has For my part, I wouldn't want to pay the electricity bill where that box is housed!! Max RAC was around 16k, IIRC? They must have been running SETI on the back-burner lately...

http://setiathome.berkeley.edu/show_...hostid=1934524

http://www.ualberta.ca/AICT/RESEARCH....may21c.02.jpg

Cheers,

Ron
( Last edited by halimedia; Dec 26, 2005 at 04:10 PM. )
     
 
Thread Tools
 
Forum Links
Forum Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Top
Privacy Policy
All times are GMT -4. The time now is 07:38 PM.
All contents of these forums © 1995-2017 MacNN. All rights reserved.
Branding + Design: www.gesamtbild.com
vBulletin v.3.8.8 © 2000-2017, Jelsoft Enterprises Ltd.,