Welcome to MilkyWay@home

Os-X Time comparisons PPC vs. Intel

Message boards : Number crunching : Os-X Time comparisons PPC vs. Intel
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · Next

AuthorMessage
Profile Paul D. Buck

Send message
Joined: 12 Apr 08
Posts: 621
Credit: 161,934,067
RAC: 0
Message 3369 - Posted: 25 Apr 2008, 15:18:34 UTC - in response to Message 3361.  


the practical solution is already there... and seems to be working well
I don't understand your point.


That's because you didn't even checked the link and had look at those crunching times...


If it is to step the OS down to Tiger that is not practical when the system was delivered with Leopard.


Again.. the same as above... check the link and look at the OS (Leopard)...

When you post a link with no explanation, yes, I missed the obvious.

The question remains, why the discrepancy on my, and other's (Alliance Francophone) systems and not on his. I have sent him a PM to ask if he/she can shed any light on the problem.

*I* still say investigation is warranted and will keep looking for an answer.
ID: 3369 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
[AF>Le_Pommier] McRoger

Send message
Joined: 31 Mar 08
Posts: 23
Credit: 721,836
RAC: 0
Message 3378 - Posted: 26 Apr 2008, 13:00:32 UTC - in response to Message 3358.  
Last modified: 26 Apr 2008, 13:01:19 UTC

Unfortunately, I'm only able to notice the fact, not to propose any practical solution.

Well, I did some research and am also trying to see if I can get some advice from other sources ...


the practical solution is already there... and seems to be working well



I see nice computing time indeed, that is a result.

My Mac Pro Quad 2.66Ghz takes 11:29 / WU under Leopard. It should take around 4-5 minutes (as under Ubuntu). It has 4 Gigs of ram, thus memory should not be an issue.

Can you give more information about how to achieve these results ?

Thanks!
ID: 3378 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
[AF>Le_Pommier] McRoger

Send message
Joined: 31 Mar 08
Posts: 23
Credit: 721,836
RAC: 0
Message 3379 - Posted: 26 Apr 2008, 13:05:39 UTC - in response to Message 3378.  
Last modified: 26 Apr 2008, 13:38:16 UTC

[quote]Unfortunately, I'm only able to notice the fact, not to propose any practical solution.

Well, I did some research and am also trying to see if I can get some advice from other sources ...


the practical solution is already there... and seems to be working well


On the other side, while this users Mac Pro is performing well under Leopard, his iMac is not doing that well under Tiger:

http://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=1160

My Mini was going more than twice as fast (2.00 Ghz, T7200) under Ubuntu than his iMac 2.33Ghz (T7600) under Tiger.

Thus, strange, is he using a beta/own decompiled/recompiled version on the Mac Pro (this is how optimized clients appeared first on SETI) ?

Edit: add my Mini info + question
ID: 3379 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Crunch3r
Volunteer developer
Avatar

Send message
Joined: 17 Feb 08
Posts: 363
Credit: 258,227,990
RAC: 0
Message 3380 - Posted: 26 Apr 2008, 19:04:50 UTC - in response to Message 3378.  



I see nice computing time indeed, that is a result.


i do see them too :)


My Mac Pro Quad 2.66Ghz takes 11:29 / WU under Leopard. It should take around 4-5 minutes (as under Ubuntu). It has 4 Gigs of ram, thus memory should not be an issue.


Memory is not the issue.


Can you give more information about how to achieve these results ?


Yes i can.




Join Support science! Joinc Team BOINC United now!
ID: 3380 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
stefsaber

Send message
Joined: 2 Apr 08
Posts: 32
Credit: 1,017,362
RAC: 0
Message 3381 - Posted: 26 Apr 2008, 19:21:50 UTC

My Leopard Mac Book Pro was averaging around 850 seconds per WU. Interesting to see how the other platforms can crunch through WU's quickly...
ID: 3381 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Crunch3r
Volunteer developer
Avatar

Send message
Joined: 17 Feb 08
Posts: 363
Credit: 258,227,990
RAC: 0
Message 3382 - Posted: 26 Apr 2008, 19:30:32 UTC - in response to Message 3369.  


When you post a link with no explanation, yes, I missed the obvious.


Joining a project without properly reading the board and what development was done is not an excuse.


The question remains, why the discrepancy on my, and other's ...


If you would have read some threads on the board you've probably would have know by now... anyhow, if you're not willing then i can't help you.


*I* still say investigation is warranted and will keep looking for an answer.


Now i'm waiting for what that investigation will come up with as a clue* ...

[* read all posts on the board you're posting on... and all threads of it!!!]






Join Support science! Joinc Team BOINC United now!
ID: 3382 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
stefsaber

Send message
Joined: 2 Apr 08
Posts: 32
Credit: 1,017,362
RAC: 0
Message 3383 - Posted: 26 Apr 2008, 19:53:15 UTC - in response to Message 3382.  

[* read all posts on the board you're posting on... and all threads of it!!!]


With the size of some boards that could take quite a while ;)

From what I gather it appears that the OS and whether or not one is running an optimized or beta app that really makes the big difference in times for WU's as opposed to clock drive speed.

Since Leopard is one of platforms that doesn't have a beta/optimized app we are going to have slower speeds for a machine of the same hardware as a general rule of thumb.

Good to see you with this project Crunch3r, I know your contribution to other BOINC projects as been great! :)
ID: 3383 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Crunch3r
Volunteer developer
Avatar

Send message
Joined: 17 Feb 08
Posts: 363
Credit: 258,227,990
RAC: 0
Message 3384 - Posted: 26 Apr 2008, 20:07:17 UTC - in response to Message 3383.  
Last modified: 26 Apr 2008, 20:08:58 UTC


With the size of some boards that could take quite a while ;)


Yeha, it might but it'S usually bether than jumping in and start spamming without having a clue ... (not you)


From what I gather it appears that the OS and whether or not one is running an optimized or beta app that really makes the big difference in times for WU's as opposed to clock drive speed.


Yes it does.


Since Leopard is one of platforms that doesn't have a beta/optimized app we are going to have slower speeds for a machine of the same hardware as a general rule of thumb.


Well yes and no... there are some factors involved in this.


Good to see you with this project Crunch3r, I know your contribution to other BOINC projects as been great! :)


Yes, you and me know that ... however some think i'm a security risk ... now that's why the MACs have to suffer that hard... OOPPPS did i really post that thought ? Hmmm ask the adminstration they can tell you much more ...

Or sign up at my message board as long as this post is still here ;)

Join Support science! Joinc Team BOINC United now!
ID: 3384 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
stefsaber

Send message
Joined: 2 Apr 08
Posts: 32
Credit: 1,017,362
RAC: 0
Message 3385 - Posted: 26 Apr 2008, 20:54:38 UTC - in response to Message 3384.  


Yes, you and me know that ... however some think i'm a security risk ... now that's why the MACs have to suffer that hard... OOPPPS did i really post that thought ? Hmmm ask the adminstration they can tell you much more ...

Or sign up at my message board as long as this post is still here ;)


Probably thoughts of jealous paranoia on some, pity that Macs have to suffer as a result. But probably not the thread or board to get into it.

ID: 3385 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Crunch3r
Volunteer developer
Avatar

Send message
Joined: 17 Feb 08
Posts: 363
Credit: 258,227,990
RAC: 0
Message 3386 - Posted: 26 Apr 2008, 20:58:42 UTC - in response to Message 3385.  
Last modified: 26 Apr 2008, 21:00:15 UTC


Yes, you and me know that ... however some think i'm a security risk ... now that's why the MACs have to suffer that hard... OOPPPS did i really post that thought ? Hmmm ask the adminstration they can tell you much more ...

Or sign up at my message board as long as this post is still here ;)


Probably thoughts of jealous paranoia on some, pity that Macs have to suffer as a result. But probably not the thread or board to get into it.


If your interested to know more about why the macs got punished... sign up over there and i'll tell you all the story...

Anyone welcome... but not here ;)... let's leave the poor devs. out of this ;)
Thank you.

Join Support science! Joinc Team BOINC United now!
ID: 3386 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
zombie67 [MM]
Avatar

Send message
Joined: 29 Aug 07
Posts: 115
Credit: 501,600,397
RAC: 5,019
Message 3389 - Posted: 27 Apr 2008, 1:31:38 UTC - in response to Message 3379.  

On the other side, while this users Mac Pro is performing well under Leopard, his iMac is not doing that well under Tiger:

http://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=1160

My Mini was going more than twice as fast (2.00 Ghz, T7200) under Ubuntu than his iMac 2.33Ghz (T7600) under Tiger.

Thus, strange, is he using a beta/own decompiled/recompiled version on the Mac Pro (this is how optimized clients appeared first on SETI) ?

Edit: add my Mini info + question


My iMac is still running the stock app.

Check this out.

Tiger, 2.16 Core Duo w/ opt app: ~478 sec

Tiger, 2.33 Core 2 Duo w/ stock app: ~865 sec

ID: 3389 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Paul D. Buck

Send message
Joined: 12 Apr 08
Posts: 621
Credit: 161,934,067
RAC: 0
Message 3390 - Posted: 27 Apr 2008, 7:29:17 UTC - in response to Message 3378.  

puting time indeed, that is a result.

My Mac Pro Quad 2.66Ghz takes 11:29 / WU under Leopard. It should take around 4-5 minutes (as under Ubuntu). It has 4 Gigs of ram, thus memory should not be an issue.

Can you give more information about how to achieve these results ?

Yes, we can develop the compiling technology so that the project can make better executables to be delivered to the participants.

The person (zombie67) who's computer that was referred to as proof that there was something somehow wrong with OS-X and / or my computer is running an optimized compile. Thusly, they are obtaining faster run times.

It is not clear to me all of the whys and wherefores and bent feelings but I can only surmise that the project asked for and received assistance in improving the algorithms used but would not accept compiled binaries because of the potential of legal liability if something went wrong or malicious code was inserted into the delivered binaries. In this, the M-Way project's sponsors reacted as would any large organization to an offer such as this regardless of the person making the offer.

It would have nothing to do with the person's trustworthiness or lack thereof, in the end, if the university cannot prove the provenance of the source code to binary transition they have opened themselves up for significant liability. Other projects make other choices ... For example, I will point to SETI@Home which is open source and allows third parties to make and distribute binaries, Einstein@Home which makes the optimized applications "in-house" and as possible tries to later "mainstream" some of the technology and CPDN which uses a third party application under license and has very strict rules about visibility of source code. Three different choices as to the delivery mechanisms and the use of optimization possibilities ... LHC@Home had issues with math libraries and cross-platform issues ... the permutations go on and on ...

And so ...

For whatever reason, the mechanism used to create the application for Linux is producing a "better" and faster executing binary. In that OS-X is unix as is Linux and they essentially use the same compiler, there should be little to no difference in end run times (most other things being equal, machine usage, background processes and so forth ...).

So, as it turns out, there is little need to read all the threads on all the forum boards if people simply answer questions in a straightforward manner and don't play "gottcha" games.

Since I have made it this far with figuring out what the problem is, and I also have spent most of a day reading up and making tries at compiling an OS-X version. Though I have familiarity with programming languages I am not a C programmer or an expert on GCC and so am re-inventing the wheel. However, I have made some progress even if it is all negative to this point, and I am hoping that tomorrow I will be able to take another whack at making a binary and from there I can begin testing to see if I can figure out the best compiler switches for OS-X binaries and to get the runtimes down.

Should I discover something that works I can report this to the project's managers and then they can make the binaries and we can move forward. If my past history with compilers is any guide, finding and setting the "right" compiler switches will make a significant difference. Exactly how we test this should I figure something out is also TBD ... For OS-X as a target the issue is complicated by the variation in the target environments, things optimized for the G4 are not well optimized for the G5 or the Intel.

But the time discrepancy is real, it likely has nothing to do with the OS in question including windows, but has to do with making the best binary possible with the tools available to the project

If someone does have some experience with Xcode you can PM me, or post here, and I will welcome your thoughts and suggestions.

If you have any other questions I doubt that I will suggest you read the UBW before I give you an obscure answer ... but, that is just me, ...
ID: 3390 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
[AF>Le_Pommier] McRoger

Send message
Joined: 31 Mar 08
Posts: 23
Credit: 721,836
RAC: 0
Message 3391 - Posted: 27 Apr 2008, 9:03:18 UTC - in response to Message 3390.  



Can you give more information about how to achieve these results ?

Yes, we can develop the compiling technology so that the project can make better executables to be delivered to the participants.

The person (zombie67) who's computer that was referred to as proof that there was something somehow wrong with OS-X and / or my computer is running an optimized compile. Thusly, they are obtaining faster run times.



Obvious there was something with an optimized app/compilation there for OS X. :-)

My question to Crunch3r was aimed at getting more than clues about this beta app.

Since we cannot get to know more, let's just wait for it to be released.
ID: 3391 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Paul D. Buck

Send message
Joined: 12 Apr 08
Posts: 621
Credit: 161,934,067
RAC: 0
Message 3394 - Posted: 27 Apr 2008, 21:29:51 UTC

Is there anyone here that has one of the earliest Mac Pros that was built on the Core Duo CPU/chipset?

If so, does it have SSE3 and SSE3 extensions?
ID: 3394 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
zombie67 [MM]
Avatar

Send message
Joined: 29 Aug 07
Posts: 115
Credit: 501,600,397
RAC: 5,019
Message 3395 - Posted: 27 Apr 2008, 22:13:47 UTC - in response to Message 3394.  

Is there anyone here that has one of the earliest Mac Pros that was built on the Core Duo CPU/chipset?

If so, does it have SSE3 and SSE3 extensions?


The first Mac Pros were all Core 2 Duo-based Xeons (aka Woodcrest).

There were no Core Duo Mac Pros.

However, there *were* Core Duo (aka Yonah) based Mac Books, and Mac Book Pros, and Mac Minis. Not sure about iMacs.

No SSE3 for Core Duo.




ID: 3395 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Dave Przybylo
Avatar

Send message
Joined: 5 Feb 08
Posts: 236
Credit: 49,648
RAC: 0
Message 3396 - Posted: 27 Apr 2008, 22:15:54 UTC - in response to Message 3395.  
Last modified: 27 Apr 2008, 22:20:05 UTC

Is there anyone here that has one of the earliest Mac Pros that was built on the Core Duo CPU/chipset?

If so, does it have SSE3 and SSE3 extensions?


The first Mac Pros were all Core 2 Duo-based Xeons (aka Woodcrest).

There were no Core Duo Mac Pros.

However, there *were* Core Duo (aka Yonah) based Mac Books, and Mac Book Pros, and Mac Minis. Not sure about iMacs.

No SSE3 for Core Duo.





Sorry to contradict you on this but SSE3 has been around for a long time. About 4 years now, and it has been used since Pentium 4's. You can view a page here that will show you all processors with SSE3 support. SSE3 Support
Dave Przybylo
MilkyWay@home Developer
Department of Computer Science
Rensselaer Polytechnic Institute
ID: 3396 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
zombie67 [MM]
Avatar

Send message
Joined: 29 Aug 07
Posts: 115
Credit: 501,600,397
RAC: 5,019
Message 3397 - Posted: 28 Apr 2008, 2:40:01 UTC - in response to Message 3396.  

Is there anyone here that has one of the earliest Mac Pros that was built on the Core Duo CPU/chipset?

If so, does it have SSE3 and SSE3 extensions?


The first Mac Pros were all Core 2 Duo-based Xeons (aka Woodcrest).

There were no Core Duo Mac Pros.

However, there *were* Core Duo (aka Yonah) based Mac Books, and Mac Book Pros, and Mac Minis. Not sure about iMacs.

No SSE3 for Core Duo.





Sorry to contradict you on this but SSE3 has been around for a long time. About 4 years now, and it has been used since Pentium 4's. You can view a page here that will show you all processors with SSE3 support. SSE3 Support


Ah! Sorry, I was thinking of SSSE3. So replace that in what I said above. The rest of what I said is still true. There were no Core Duo Mac Pros.

ID: 3397 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Crunch3r
Volunteer developer
Avatar

Send message
Joined: 17 Feb 08
Posts: 363
Credit: 258,227,990
RAC: 0
Message 3398 - Posted: 28 Apr 2008, 8:24:42 UTC - in response to Message 3397.  

OK... so here we go...

OSX optimized applcations

Have fun :)






Join Support science! Joinc Team BOINC United now!
ID: 3398 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
zombie67 [MM]
Avatar

Send message
Joined: 29 Aug 07
Posts: 115
Credit: 501,600,397
RAC: 5,019
Message 3399 - Posted: 28 Apr 2008, 13:53:40 UTC - in response to Message 3389.  

Tiger, 2.16 Core Duo w/ opt app: ~478 sec

Tiger, 2.33 Core 2 Duo w/ stock app: ~865 sec


After installing the opt app, the C2D is down from ~865 sec to ~405 sec.

ID: 3399 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Martin P.

Send message
Joined: 21 Nov 07
Posts: 52
Credit: 1,756,052
RAC: 0
Message 3400 - Posted: 28 Apr 2008, 15:34:59 UTC - in response to Message 3398.  

OK... so here we go...

OSX optimized applcations

Have fun :)


Hi Crunch3r,

thanks a lot! Do these apps also work with the Xeons in the MacPro?

ID: 3400 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · Next

Message boards : Number crunching : Os-X Time comparisons PPC vs. Intel

©2024 Astroinformatics Group