|
|
Enhanced Optimized (Page 13)
|
|
|
|
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status:
Offline
|
|
There's no doubt, those CoreDuo (and Core2's) are one heckofa chip.
I bet the new laptops with the Core2s do nicely as well.
Next project: trying to get my G52.5 to get an RAC of 1200 or above.
I have an idea how to accomplish it but ack, if I'd waited a year.
Oh well.
|
|
|
|
|
|
|
|
|
Junior Member
Join Date: Jun 2005
Status:
Offline
|
|
at the risk of sounding over eager :-) is version 7 ppc availabe for release yet alex ?
regards
adream
|
63. (1) (b) "music" includes sounds wholly or predominantly characterised by the emission of a succession of repetitive beats
|
|
|
|
|
|
|
|
Forum Regular
Join Date: Aug 2005
Location: Cupertino, CA
Status:
Offline
|
|
Originally Posted by adream
at the risk of sounding over eager :-) is version 7 ppc availabe for release yet alex ?
Well, there were a couple other things that I want to pursue, but v7.1 is already enough of an improvement over v6 that I might as well put this out now and release v7.2 later...
v7.1-g4-nographics
v7.1-g5-nographics
Versions with graphics are a long way off, because some of the optimizations I made won't play nice with graphics and I need to figure out how to deal with this. Source will be posted later at the link given in the ZIP file, if people are interested. Don't compile it for i386, though...it won't work yet.
Todd, hopefully this will help you get your G5 over 1200 RAC.
|
|
|
|
|
|
|
|
|
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status:
Offline
|
|
|
|
|
|
|
|
|
|
|
Forum Regular
Join Date: Oct 2005
Location: Las Vegas, NV
Status:
Offline
|
|
Originally Posted by Todd Madson
Wheeee! Thank you!
I'm totally bummed. I haven't missed a new ap release since when Java Lizard's was top dog. A shiny new v7.1, and no PPC for me to run it on now.
G4 is still dead. Apple Genius said HD, I disagreed, but installed a new one I had in a box.
Still dead, won't post. 2 CPUs dead in 1 week? Don't think so.
Me thinks logic board. Hope not...too expensive!
Have fun and I look forward to seeing some G4 times from the gang.
Maybe there is still hope for a cheap fix for my G4
Cheers!
|
|
|
|
|
|
|
|
|
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status:
Offline
|
|
Here is the url for my machine:
Results for computer
And here's one of the 7.1 returned units:
Result
4466 seconds versus an average time before of 5400 to 7050 seconds!
Stay tuned, there's only one unit processed by 7.1 in, more later.
|
|
|
|
|
|
|
|
|
Junior Member
Join Date: Jun 2005
Status:
Offline
|
|
Many thanks Alex for another fine client,
its running here on a couple of g4s and my dual 2.7 g5, all faster and no errors yet
i really appreciate your efforts to enhance the clients
regards
adream
|
63. (1) (b) "music" includes sounds wholly or predominantly characterised by the emission of a succession of repetitive beats
|
|
|
|
|
|
|
|
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status:
Offline
|
|
And:
Results for computer
This is my wife's iMac G4.
Her last work unit with v6 is Result
that was in 14,209.48 seconds.
Latest with v7 is Result
completed in 10,654.37 seconds
More later as more results come in. This is highly impressive improvement!
|
|
|
|
|
|
|
|
|
Senior User
Join Date: Jun 2006
Location: Dublin, CA
Status:
Offline
|
|
Hi Alex,
You probably missed it because this thread scrolled to a new page. Can you please go back a page and search for my userid? There is an error message that is showing up on the intel version of your app. I would like to know if that is normal.
TIA!
|
|
|
|
|
|
|
|
|
Forum Regular
Join Date: Aug 2005
Location: Cupertino, CA
Status:
Offline
|
|
Originally Posted by Todd Madson
And:
Results for computer
This is my wife's iMac G4.
Her last work unit with v6 is Result
that was in 14,209.48 seconds.
Latest with v7 is Result
completed in 10,654.37 seconds
More later as more results come in. This is highly impressive improvement!
Don't compare the times on two work units unless they're within +/- 0.5 of each other in terms of claimed credit. Crunch times vary a lot more than they used to in SETI Enhanced.
(
Last edited by alexkan; Oct 27, 2006 at 10:48 AM.
)
|
|
|
|
|
|
|
|
|
Forum Regular
Join Date: Aug 2005
Location: Cupertino, CA
Status:
Offline
|
|
Originally Posted by zombie67
You probably missed it because this thread scrolled to a new page. Can you please go back a page and search for my userid? There is an error message that is showing up on the intel version of your app. I would like to know if that is normal.
For the most part, I wouldn't worry about problems unless they're causing you to crash, misclaim credit or fail to validate, since these are the only sorts of effects that my changes are likely to have. In your case, I'd check the file permissions on your SETI binaries...it looks like an operation to modify the app's icon is failing because of insufficient permissions.
|
|
|
|
|
|
|
|
|
Senior User
Join Date: Jun 2006
Location: Dublin, CA
Status:
Offline
|
|
Originally Posted by alexkan
For the most part, I wouldn't worry about problems unless they're causing you to crash, misclaim credit or fail to validate, since these are the only sorts of effects that my changes are likely to have. In your case, I'd check the file permissions on your SETI binaries...it looks like an operation to modify the app's icon is failing because of insufficient permissions.
Thanks for getting back to me!
Well, trying to fix it was a disaster. I tried to make the owner & permissions match my PPC version, for the seti application. Then BOINC wouldn't even start. I changed them back, but no good. Then I tried reinstalling your optomized app, with the default permissions, but that didn't work either. It required actually reinstalling BOINC, as if that makes any sense.
The only thing I can think of is that I was using a 5.6.something for BOINC, and that was causing the ownership issues. Anyway, after installing 5.4.9, everything is back on track. I will take a look at the results later and see if the error messages are gone.
Thanks again for such a great app!
|
|
|
|
|
|
|
|
|
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status:
Offline
|
|
Wow. Weirdness.
Okay, I got back from work and check this out:
Results for computer
Take a look at some of the numbers. Interesting reduction.
But look at the workunits:
Result
indicates "No heartbeat from core client for 31 sec - exiting"
So does:
Result
And:
Result
And:
Result
Is something broken or is this normal?
Here's a page of my numbers with the old client.
Versus:
Results for computer
Mostly 6000-8000 seconds with the older client.
(
Last edited by Todd Madson; Oct 27, 2006 at 06:58 PM.
)
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Aug 2005
Location: Golden Valley, AZ
Status:
Offline
|
|
Nothing like that on my system so far.
|
|
|
|
|
|
|
|
|
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status:
Offline
|
|
I rebooted the thing, it had been up for over a week and I'm guessing it just needed
a punt to the head. I'll post if more weirdness occurs.
|
|
|
|
|
|
|
|
|
Not a Member
|
|
Originally Posted by alexkan
Well, there were a couple other things that I want to pursue, but v7.1 is Source will be posted later at the link given in the ZIP file, if people are interested.
Well i'm interested
Good opportunity to learn some of the stuff you do ;-)
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Jun 2006
Location: Chicago
Status:
Offline
|
|
I've blown about 30 WUs this morning after upgrading to v7.1-G4. The error messages are all pretty much like this:
Sat Oct 28 08:30:48 2006|SETI@home|Finished download of 02my03aa.22748.9330.698572.3.213
Sat Oct 28 08:30:48 2006|SETI@home|Throughput 165675 bytes/sec
Sat Oct 28 08:30:49 2006||request_reschedule_cpus: files downloaded
Sat Oct 28 08:30:49 2006|SETI@home|Starting result 02my03aa.22748.9330.698572.3.213_1 using setiathome_enhanced version 513
Sat Oct 28 08:30:50 2006|SETI@home|Unrecoverable error for result 02my03aa.22748.9330.698572.3.213_1 (process exited with code 2 (0x2))
Sat Oct 28 08:30:50 2006||request_reschedule_cpus: process exited
Sat Oct 28 08:30:50 2006|SETI@home|Computation for result 02my03aa.22748.9330.698572.3.213_1 finished
Sat Oct 28 08:39:58 2006||request_reschedule_cpus: project op
This was with a 1GHz iLamp running 10.4.7. v6 had been running just fine.
Anyone have any suggestions?
QS
|
|
|
|
|
|
|
|
|
Not a Member
|
|
Originally Posted by QSilver
I've blown about 30 WUs this morning after upgrading to v7.1-G4. The error messages are all pretty much like this:
Sat Oct 28 08:30:48 2006|SETI@home|Finished download of 02my03aa.22748.9330.698572.3.213
Sat Oct 28 08:30:48 2006|SETI@home|Throughput 165675 bytes/sec
Sat Oct 28 08:30:49 2006||request_reschedule_cpus: files downloaded
Sat Oct 28 08:30:49 2006|SETI@home|Starting result 02my03aa.22748.9330.698572.3.213_1 using setiathome_enhanced version 513
Sat Oct 28 08:30:50 2006|SETI@home|Unrecoverable error for result 02my03aa.22748.9330.698572.3.213_1 (process exited with code 2 (0x2))
Sat Oct 28 08:30:50 2006||request_reschedule_cpus: process exited
Sat Oct 28 08:30:50 2006|SETI@home|Computation for result 02my03aa.22748.9330.698572.3.213_1 finished
Sat Oct 28 08:39:58 2006||request_reschedule_cpus: project op
This was with a 1GHz iLamp running 10.4.7. v6 had been running just fine.
Anyone have any suggestions?
QS
You should check your file permission on the executable.
Boinc cannot find the app.
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Aug 2005
Location: Golden Valley, AZ
Status:
Offline
|
|
Did you quit Boinc Manager before you changed the executable?
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Nov 2004
Location: Virginia
Status:
Offline
|
|
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Jun 2006
Location: Chicago
Status:
Offline
|
|
Originally Posted by Not a Member
You should check your file permission on the executable.
Boinc cannot find the app.
The file permissions at -rwxr-xr-x look OK.
BOINC Manager was not running when I upgraded.
The only difference that I can see compaed to the v6 file is that v6 shows up in List View as a "Unix Executable File." Somehow the v7 file thinks it's a Text File.
QS
|
|
|
|
|
|
|
|
|
Forum Regular
Join Date: Oct 2005
Location: Las Vegas, NV
Status:
Offline
|
|
Originally Posted by beadman
Alex:
Here are a couple of results from my iBook 1.33 GHz G4, OS X 10.4.7: The first one is using v6 and the second is using v7.
Result 19,864 cpusec
Result 15,964 cpusec
Thanks for the great client!!
beadman
Wow, Beadman! That's cooking! Your times for that AR on v6 are very similar to what my G4 was running. v7 appears to be a huge step forward on G4. From appearances, it's running close to a 2GHz Pent M using the fastest 5.15 xB (Pent M specific) client for x86 that Chicken has released. Pent M is very efficient (for Netburst) w/ a short pipeline & low latency so for the G4 to run this close w/ 33% less clock, 1/4th the L2 cache & @ 1/2 FSB speed of the Pent M is astounding. That should make for a nice increase in RAC!
(
Last edited by Gecko_r7; Oct 29, 2006 at 11:41 PM.
)
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Nov 2004
Location: Virginia
Status:
Offline
|
|
Thanks, Geck0 - I'm really hoping my PowerBook G4 will also show a major increase in RAC when I get back to the office on Tuesday and update the client. I missed several days on my iBook as I was in Colorado Springs all week with my MBP and the iBook was at home. I'll post some PowerBook results after I get it installed. My poor MPB's RAC is suffering badly after sitting in an airport for a couple of hours, followed by a 4 hour flight and then waiting for the luggage and an hour-long drive home...
Hows your iMac doing? I took a quick look, but didn't find it.
beadman.
(
Last edited by beadman; Oct 29, 2006 at 11:58 PM.
)
|
|
|
|
|
|
|
|
|
Forum Regular
Join Date: Oct 2005
Location: Las Vegas, NV
Status:
Offline
|
|
Hi beadman.
It's going to be fun to see where your's ends up once is get's steady time under it's belt w/ 7.1. FWIW, iMac seems to have settled @ 1540-1560 RAC & @ 150-165 on the list. Never expected it to do this well, so it's a nice surprise.
Looking forward to Alex's upcoming v7.2 on this rig.
I'm taking the G4 back to the local Apple store tomorrow and going to have them send-in for a diagnostic/repair estimate. Fingers crossed!
Cheers!
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Aug 2005
Location: Golden Valley, AZ
Status:
Offline
|
|
My little eMac is starting to crank up, running between 14000 and 17000 seconds.
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Sep 2004
Location: London
Status:
Offline
|
|
Originally Posted by QSilver
The file permissions at -rwxr-xr-x look OK.
BOINC Manager was not running when I upgraded.
QS
After you drop in the the worker, re-install BOINC over the top, it should reset the permissions. Under messages tab you should see something about new anon xml file.
You may lose any wu's that were already started, but you should not lose any more than that. If you did that already then completely remove BOINC and all its files and do a clean install and try again as above.
K.
|
|
|
|
|
|
|
|
|
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status:
Offline
|
|
My G5 is basically screaming. The RAC graph is at a 45-50% angle with no end in sight.
I'll keep you posted as to where RAC tops out. Currently it's 1144.
Look at previous pages of crunched units they were in the 6300, 6400,6500 range
for the most part, we're now down in the low four thousands for those averages
but it's hard to say this early on. Give it a week or so.
G4s are doing well too. My old G4 routinely would get RAC of 98, it's
now somewhere in the 112 range.
Wife's G4 is at RAC 239, I've got an Athlon 2800 machine that is at RAC 218
with Chicken Good installed and an Athlon 2400 at RAC 165.
(
Last edited by Todd Madson; Oct 30, 2006 at 12:09 PM.
Reason: typo)
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status:
Offline
|
|
On average, you can expect a speed increase of roughly 15% with v7.1 vs. v6ng - both on G4 and G5. Topping a RAC of 1200 with a DP 2.5 GHz G5 may be possible, but only with long runs of WUs with low CPU sec./credit values.
|
|
|
|
|
|
|
|
|
Fresh-Faced Recruit
Join Date: Nov 2005
Location: Europe
Status:
Offline
|
|
Interesting. All my 7.1 WUs contain this message, although they do validate ok:
<tt><blockquote>WU true angle range is : 0.446738
No heartbeat from core client for 31 sec - exiting
OS X optimized S@H Enhanced application by Alex Kan
Version info: OS X Altivec (G4/G5, G5-optimized v7.1-nographics) V5.13 by Alex Kan</blockquote></tt>
e.g.: <a href="http://setiweb.ssl.berkeley.edu/result.php?resultid=404881610">Workunit 97076716</a>:
<tt><blockquote><core_client_version>5.4.9</core_client_version>
<stderr_txt>
OS X optimized S@H Enhanced application by Alex Kan
Version info: OS X Altivec (G4/G5, G5-optimized v6-nographics) V5.13 by Alex Kan
Work Unit Info:
...............
WU true angle range is : 0.630636
<font color="red">No heartbeat from core client for 31 sec - exiting</font>
OS X optimized S@H Enhanced application by Alex Kan
Version info: OS X Altivec (G4/G5, G5-optimized v7.1-nographics) V5.13 by Alex Kan
Work Unit Info:
...............
WU true angle range is : 0.630636
<font color="red">No heartbeat from core client for 31 sec - exiting</font>
OS X optimized S@H Enhanced application by Alex Kan
Version info: OS X Altivec (G4/G5, G5-optimized v7.1-nographics) V5.13 by Alex Kan
Work Unit Info:
...............
WU true angle range is : 0.630636
Flopcounter: 12145345600692.083984
Spike count: 0
Pulse count: 0
Triplet count: 2
Gaussian count: 0
</stderr_txt></blockquote></tt>
And another one:
<a href="http://setiweb.ssl.berkeley.edu/result.php?resultid=405543082">Workunit 97237043</a>:
<tt><blockquote>5.4.9
OS X optimized S@H Enhanced application by Alex Kan
Version info: OS X Altivec (G4/G5, G5-optimized v7.1-nographics) V5.13 by Alex Kan
Work Unit Info:
...............
WU true angle range is : 0.949786
<font color="red">No heartbeat from core client for 31 sec - exiting</font>
OS X optimized S@H Enhanced application by Alex Kan
Version info: OS X Altivec (G4/G5, G5-optimized v7.1-nographics) V5.13 by Alex Kan
Work Unit Info:
...............
WU true angle range is : 0.949786
<font color="red">No heartbeat from core client for 31 sec - exiting</font>
OS X optimized S@H Enhanced application by Alex Kan
Version info: OS X Altivec (G4/G5, G5-optimized v7.1-nographics) V5.13 by Alex Kan
Work Unit Info:
...............
WU true angle range is : 0.949786
<font color="red">SETI@Home Informational message -9 result_overflow
NOTE: The number of results detected exceeds the storage space allocated.</font>
</blockquote></tt>
Any ideas?
(
Last edited by gulliver; Oct 30, 2006 at 06:46 PM.
)
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Sep 2004
Location: London
Status:
Offline
|
|
Originally Posted by gulliver
Any ideas?
Out of curiosity, do you have BOINC set as a/the screensaver?
Obvious I know but, something else going on, on your computer, at that time?
Alex said as long as they validate to ignore error messages.
K.
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Jun 2006
Location: Chicago
Status:
Offline
|
|
Originally Posted by Knightrider
After you drop in the the worker, re-install BOINC over the top, it should reset the permissions. Under messages tab you should see something about new anon xml file.
You may lose any wu's that were already started, but you should not lose any more than that. If you did that already then completely remove BOINC and all its files and do a clean install and try again as above.
K.
I completely removed BOINC and eveything in /Library/ApplicationSupport/BOINC Data. Even got rid of the .plist in the Preferences folder.
All of the v7 files come in with a group of 'staff'. I changed them to group 'admin', to match the v6 files. I think the problem has something to do with the fact that all of the v7 files have a Kind of 'Text document" (Get Info shows Open With "Text Edit (default)"). This is in contrast to the v6 files where the science app has a Kind = "Unix Executable File" (Get Info shows Open With "<None>"). The rest of the v6 files have Kind = "Document".
I'm back on v6 for the moment.
QS
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Sep 2004
Location: London
Status:
Offline
|
|
Originally Posted by QSilver
All of the v7 files come in with a group of 'staff'. I changed them to group 'admin', to match the v6 files. I think the problem has something to do with the fact that all of the v7 files have a Kind of 'Text document" (Get Info shows Open With "Text Edit (default)"). This is in contrast to the v6 files where the science app has a Kind = "Unix Executable File" (Get Info shows Open With "<None>"). The rest of the v6 files have Kind = "Document".
QS
That is interesting. The files are xml, Text and Unix Executable. 'Open with' is set to 'None' and if you tried to open them they would open on the default file association or you would be asked to choose an app.
The files in my SAH/project folder all have the following.
Ownership and Permissions = read & write
Details
Owner = boinc_master. access = r & w
Group = boinc_project. access = r & w
Others = read only
These may not be correct however, as I get an error message on all my wu's
MacOS Error -5000 occured in /Users/alexkan/seti/boinc/api/mac_icon.C line 107
Perhaps someone who is running V7 without errors would be kind enough to post their Ownership and Permissions settings on the files in the sah project folder.
I have noticed before that the OS can sometimes get it wrong when automatically allocating Ownership and Permissions.
K.
K.
|
|
|
|
|
|
|
|
|
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status:
Offline
|
|
Qsilver & others with permissions issues:
Why not drop everything into a directory and then use the cli to CHMOD 777 the files in
that directory? That worked for me ages ago when I had issues like yours.
CD to the directory and perform chmod 777 to all files in that directory.
Then restart the client.
It should work - I've rarely had success using the gui for that, I always use the
terminal to make it work.
(
Last edited by Todd Madson; Oct 31, 2006 at 10:42 AM.
)
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Jun 2006
Location: Chicago
Status:
Offline
|
|
Originally Posted by Todd Madson
Qsilver & others with permissions issues:
Why not drop everything into a directory and then use the cli to CHMOD 777 the files in
that directory? That worked for me ages ago when I had issues like yours.
CD to the directory and perform chmod 777 to all files in that directory.
Then restart the client.
It should work - I've rarely had success using the gui for that, I always use the
terminal to make it work.
Todd--I used Terminal to chgrp from 'staff' to 'admin'. Hadn't bothered with chmod since v7 has the same permissions that v6 had. I'll give chmod a try when I get home.
QS
|
|
|
|
|
|
|
|
|
Administrator
Join Date: Jun 2000
Location: California
Status:
Offline
|
|
If the executable is not reading as an executable, you could try the old standby:
chmod +x workername
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Jun 2006
Location: Chicago
Status:
Offline
|
|
Originally Posted by reader50
If the executable is not reading as an executable, you could try the old standby:
chmod +x workername
Neither chmod +x nor chmod 777 helped.
Is there any other way of designating a file as an executable?
QS
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Aug 2005
Location: Golden Valley, AZ
Status:
Offline
|
|
What did you use to expand the file, I have seen stuffit screw up the worker several times when I downloaded them. BomArchiveHelper usually does the better job.
|
|
|
|
|
|
|
|
|
Junior Member
Join Date: Jun 2005
Status:
Offline
|
|
Originally Posted by arkayn
What did you use to expand the file, I have seen stuffit screw up the worker several times when I downloaded them. BomArchiveHelper usually does the better job.
i second that, stuffit is bleedin useless nowadays, often screwing with zip files
bomarchivehelper is very good, very quick and leaves unix stuff well alone
its in /system/library/coreservices/
its much much quicker at zipping and unzipping than stuffit (and it uses both processors unlike stuffit)
regards
adream
|
63. (1) (b) "music" includes sounds wholly or predominantly characterised by the emission of a succession of repetitive beats
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Jun 2006
Location: Chicago
Status:
Offline
|
|
Originally Posted by adream
i second that, stuffit is bleedin useless nowadays, often screwing with zip files
bomarchivehelper is very good, very quick and leaves unix stuff well alone
its in /system/library/coreservices/
its much much quicker at zipping and unzipping than stuffit (and it uses both processors unlike stuffit)
regards
adream
Thanks arkayn & adream!
Apparently, StuffIt was ignoring the filetypes when it unzipped the download. BOMArchiveHelper unzipped Alex's download and once I dropped everything into the BOINC Data folder and restarted BOINC Manager, v7.1 fired right up.
Thanks again!
QS
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Aug 2005
Location: Golden Valley, AZ
Status:
Offline
|
|
Glad I could figure out what the problem was and get you back up and running.
|
|
|
|
|
|
|
|
|
Manu
|
|
Hi!
Look at this workunit:
Click here
My iMac G4 running v7.1 with 800Mhz actually beat a 3 Ghz Pentium which was running an optimized worker too. oO
I don't think that's realistic (800Mhz <-> 3000 Mhz), there must be something wrong.
What do you think?
Manu
|
|
|
|
|
|
|
|
|
Forum Regular
Join Date: Oct 2005
Location: Las Vegas, NV
Status:
Offline
|
|
Originally Posted by Manu
Hi!
Look at this workunit:
Click here
My iMac G4 running v7.1 with 800Mhz actually beat a 3 Ghz Pentium which was running an optimized worker too. oO
I don't think that's realistic (800Mhz <-> 3000 Mhz), there must be something wrong.
What do you think?
Manu
Hi Manu,
Yes, your G4 did crunch this WU faster if judging by reported time alone and later model G4s w/ Alex's aps can compete very well w/ some single CPU P4 Netburst chips (Northwoods for ex.) in the mid2-3GHZ range. The P4 architectures (w/ exception of Pent M) are not near as efficient clock to clock and have much longer pipelines and more latency per clock cycle. Alex's aps also maximize the advantages that Altivec provides to G4 and G5 over the P4's use of FPU and SSE2, SSE3 etc. His optimized aps are THE FASTEST available of any architecture, PERIOD! So....it is not a level field between G4 and P4 Mhz to Mhz. G4 can provide close performance to many older single-CPU, single threaded Intel P4s @ 1/2 to 1/3 the clock rating. The latest Prescott and Cedar Mill P4 varieties e/ HyperThreading ability and larger L2 caches can do substantially more work than standard single-cpu Apple G4s however. If one throws DP G4s, or aftermarket upgraded 1.5,1.7, 2GHZ processor cards into the mix, THAT can be a different story.
In your case though, 800Mhz vs. 3.0 GHz is a little too wide a gap to overcome, esp. against a HyperThreaded model. Your P4 friend is actually crunching 2 WUs at 28K, vs. your 1 at 26K. He has HyperThreading enabled which gives him an extra "virtual"processor and allows him to crunch a 2nd WU using a second thread....though at dramatically reduced speed for each one, compared to if his CPU was totally dedicated to crunching only 1 WU. Understand? More WUs, processing at reduced speed for each, but a NET increase in work done when looking at TOTAL production for any period of time. From the cache shown he's likely running a P4 630 Prescott or 631 Cedar Mill and his RAC is @ 335. Basically he's doing @ 2X the amount of work in the time your G4 does 1. He can crunch 1 WU much quicker than your G4 if he disables HT (though at the cost of lower overall production). Depends on how you want to look at it, but real world, he's ultimately faster since he can produce more results in the same period of time.
You should still be proud though. He has almost 4x the clock, but does just under 2x the work.
Your G4 is doing great!
(
Last edited by Gecko_r7; Nov 2, 2006 at 10:37 PM.
)
|
|
|
|
|
|
|
|
|
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status:
Offline
|
|
So if we had P4 machines, turning off hyperthreading would allow for crunching of
work units with full resources to single work units at a time?
I've wondered if the dual cores are similarly affected?
What about dual processor machines? If they were told to crunch only 1 WU at a
time would it then devote both cpus to one workunit? I doubt they're programmed
to optimize this but it might be interesting.
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status:
Offline
|
|
Originally Posted by Todd Madson
So if we had P4 machines, turning off hyperthreading would allow for crunching of work units with full resources to single work units at a time?
Yes, but the RAC of that machine would be reduced in the process, altough each individual WU would be processed faster. [edit: well, not with full resources, really - some execution units would probably be idle sometimes (see below)]
I've wondered if the dual cores are similarly affected?
Yes, they are. [edit: however, HT is not equivalent to a full-fledged multi-core architecture - it just uses idle execution units more cleverly. See <a href="http://en.wikipedia.org/wiki/Hyper_Threading" target="_blank">Wikipedia</a>]
What about dual processor machines?
Same. A core is a core, whether there is only one per die or more (hey, I'm becoming a poet )
If they were told to crunch only 1 WU at a time would it then devote both cpus to one workunit?
Nope, SETI-computations are single-threaded, i.e. execution occurs in sequence in one alculation that cannot be [edit: it could theoretically be, but it isn't in the current code] split to more than one core.
I doubt they're programmed to optimize this but it might be interesting.
Academically interesting, maybe. But overall performance will most definitely be better if more than one WU is crunched in parallel than when spinning off threads of a single WU-calculation to more than one core, since multi-threading would in most cases incur a performance penalty due to the overhead involved in spinning off threads and later merging the results.
(
Last edited by halimedia; Nov 3, 2006 at 02:16 PM.
)
|
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status:
Offline
|
|
BTW, as I mentioned before, reaching a RAC of above 1200 is possible with a dual-2.5GHz G5 under v7.1. But it probably won't be sustainable - in my experience so far, RAC varies between 1220 and about 1150, averaging around 1185.
See the current results of my dual-2.5GHz G5
|
|
|
|
|
|
|
|
|
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status:
Offline
|
|
Halimedia: I see your floating point and other stuff are way higher than mine - are
you still running the Superbench Boinc client or is there a new one?
As far as RAC goes:
I suspected this - I'm running the thing and it's doing much better but the RAC seems
to waver a lot depending on other things I opt to run: Right now it's 1,169.69
which I think is amusing. If I'm running other things in the background such as the
Apache webserver, Bittorrent or other apps it tends to degrade performance when
I'm not actually there - if I just run Seti and I leave for a day or two is when the
interesting peaks begin. But I've not seen it get any higher than 1170 or so. Yet.
My current results are here: Computer summary
But this new version Alex cooked up is pretty viciously better. Me likey as they say.
|
|
|
|
|
|
|
|
|
Fresh-Faced Recruit
Join Date: Oct 2006
Status:
Offline
|
|
I can download an app from alex to speed (a bit up) but i cant find the manual explaining me how to install them.
i have active for seti
imac 24" (i386)
MacPro 2x3ghz (i386)
G5 dual 1,8
G5 dual 2,7
i want to bump up the speed ( i wanna blow the stats ;-) but need a simple installer or manual what and how to install
someone can help me with this (i like the dutch manual :-)) but english is also oke
BBbent
|
|
|
|
|
|
|
|
|
Junior Member
Join Date: Jun 2005
Status:
Offline
|
|
Originally Posted by Boudewijn
I can download an app from alex to speed (a bit up) but i cant find the manual explaining me how to install them.
i have active for seti
imac 24" (i386)
MacPro 2x3ghz (i386)
G5 dual 1,8
G5 dual 2,7
i want to bump up the speed ( i wanna blow the stats ;-) but need a simple installer or manual what and how to install
someone can help me with this (i like the dutch manual :-)) but english is also oke
BBbent
heres some intructions i wrote last night for a couple of friends, hopefully its useful
regards
adream
Step 1.
Go to the boinc download page and download the boinc manager, either the
graphical manager or the command line manager will do. these can be found
here: Download BOINC client software There is no longer any advantage
in usng any optimised Boinc apps, just the worker clients.
Step 2.
Install the Boinc manager software and run while connected to the WAN.
You will have to attach to at least one project when you first run it.
If your participating in SETI then attach and wait for the manager
to download the seti worker and associated files.
Step 3.
Close the Boinc manager down, and download the optimised seti clients
from alex khans ftp page
http://tbp.berkeley.edu/~alexkan/seti/ obviously choosing the latest version
for your particular CPU
Step 4.
Unzip the optimised clients using BOMArchivehelper ( found in /system/library/coreservices/ ) .
Drag and drop all the contents of that folder into:
/library/application support/Boinc data/projects/setiathome.berkely.edu/
or whichever folder you chose to install the Boinc data into ( if you are using the command line client )
Make sure you copy all the files including "app_info.xml"
Step 5.
Thats it, just run boinc and watch your RAC rise !
For your reference i include a folder listing of my client folder, including a
few work units so dont get confused by them, you won't have many when you first run the client
Once you are up and running using the optimised clients its worth checking your results for a day or so to confirm that the output file shows the right alex khan version and that there are no errors.
adream folder listing:
admins-power-mac-g5:/Library/Application Support/BOINC Data/projects/setiathome.berkeley.edu admin$ ls -l
total 18888
-rw-r--r-- 1 admin admin 362077 Nov 2 13:22 08jn03aa.5247.7056.778410.3.26
-rw-r--r-- 1 admin admin 362074 Nov 2 21:28 08jn03aa.5247.9858.167318.3.135
-rw-r--r-- 1 admin staff 416 Oct 26 16:29 AUTHORS
-rw-r--r-- 1 admin staff 17988 May 27 08:04 COPYING
-rw-r--r-- 1 admin staff 793 May 27 08:02 COPYRIGHT
-rw-r--r-- 1 admin staff 181 May 27 08:04 EXTRA_NOTE
-rw-r--r-- 1 admin staff 1722 May 27 08:04 README
-rw-r--r-- 1 admin staff 358 Oct 26 16:30 app_info.xml
-rwxr-xr-x 1 admin staff 3000700 Oct 26 22:06 seti_enhanced-ppc-v7.1-g5-nographics
|
63. (1) (b) "music" includes sounds wholly or predominantly characterised by the emission of a succession of repetitive beats
|
|
|
|
|
|
|
|
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status:
Offline
|
|
Originally Posted by Todd Madson
Halimedia: I see your floating point and other stuff are way higher than mine - are you still running the Superbench Boinc client or is there a new one?
Yup, I'm still using the superbench-version of the BOINC 5.4.9 client. Not that there's any benefit in that when crunching SETI-enhanced, but it works fine, so I didn't revert to the stock client.
If I'm running other things in the background such as the
Apache webserver, Bittorrent or other apps it tends to degrade performance when
I'm not actually there - if I just run Seti and I leave for a day or two is when the
interesting peaks begin.
Yes, and if there's a lot of network-traffic, it's mainly the TCP/IP-stack gobbling up resources. Time for Apple building in NICs supporting TOE, at least in the Pro-product line and in Xserves, IMO.
But this new version Alex cooked up is pretty viciously better. Me likey as they say.
Yes, I like it too. But what I think is really wild is that my lowly 1.66 GHz CoreDuo Intel mini consistently beats a dual-2 GHz G5 Xserve when running v7.0 and v7.1, respectively. Apple definitely did the right thing switching to Intel, IMO!
(
Last edited by halimedia; Nov 4, 2006 at 05:01 AM.
)
|
|
|
|
|
|
|
|
|
Mac Elite
Join Date: Apr 2000
Location: Minneapolis, MN USA
Status:
Offline
|
|
Okay folks, this is odd: look at this:
Computer summary
Last night for fun I tried to revert to the Superbench just to get those high FP
scores (just because it's cool) but had to leave the house to do something and
put the old one back until I have more time.
All of a sudden my machine seems to have gone bonkers with tons of units
in the 2000 seconds range - my machine went from 190 in the results column
to 306 overnight! Coincidence? Probably. But I'm impressed.
RAC dropped but I expected that I guess.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Forum Rules
|
|
|
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
|
HTML code is Off
|
|
|
|
|
|
|
|
|
|
|
|