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 > Boinc benchmarks

Boinc benchmarks (Page 2)
Thread Tools
falofolio
Fresh-Faced Recruit
Join Date: Jul 2004
Status: Offline
Reply With Quote
Jul 20, 2004, 11:25 PM
 
Originally posted by mikkyo:
Once I am able to build the Seti worker again....
Ill put out a super optimized version of it.
Unfortunately, everytime I check out the source for a new Seti version it no longer builds.
Yet another reason to not like Seti I guess.
Yes, okay, that's reasonable. I can understand that. However, I think, my question was badly phrased, because that was not what I meant to get an answer on. I'll rephrase my question here and correct me when I'm wrong here:

In the situation right now, for those who want that, we can now use the (Super-)Optimized version of BOINC in the SETI-project. However, BOINC isn't THE program that does the calculations for SETI, because it downloads the actual classic SETI@home program to do the job. So, in this context, what will the (Super-)Optimized version give me more than the standard version, when it is not the one which is doing the actual SETI calculations? Will a (Super-)Optimized version still speed up my Mac to analyse the data of one work unti --- to use the old term --- for SETI?
     
fillmeup
Fresh-Faced Recruit
Join Date: Jul 2004
Location: Oregon
Status: Offline
Reply With Quote
Jul 21, 2004, 12:38 AM
 
Many thanks mikkyo!

Machine Model: iMac
CPU Type: PowerPC G4 (2.1)
Number Of CPUs: 1
CPU Speed: 700 MHz
L2 Cache (per CPU): 256 KB
Memory: 256 MB
Bus Speed: 100 MHz

Old Benchmark:
2004-07-16 00:35:57 [---] Benchmark results:
2004-07-16 00:35:58 [---] Number of CPUs: 1
2004-07-16 00:35:58 [---] 477 double precision MIPS (Whetstone) per CPU
2004-07-16 00:35:58 [---] 1031 integer MIPS (Dhrystone) per CPU
2004-07-16 00:35:58 [---] Finished CPU benchmark

New Benchmark:
2004-07-20 21:20:46 [---] Benchmark results:
2004-07-20 21:20:46 [---] Number of CPUs: 1
2004-07-20 21:20:46 [---] 969 double precision MIPS (Whetstone) per CPU
2004-07-20 21:20:46 [---] 1090 integer MIPS (Dhrystone) per CPU
2004-07-20 21:20:46 [---] Finished CPU benchmark

Didn't know which one to use. After a little research I picked the G4 7450.
     
Lateralus
Moderator Emeritus
Join Date: Sep 2001
Location: Arizona
Status: Offline
Reply With Quote
Jul 21, 2004, 12:52 AM
 
Alrighty people:

7400/7410 - Use this if you have an original G4 PowerBook (400MHz or 500MHz). Or if you have a Yikes, Sawtooth, Gigabit Ethernet, G4 Cube, or a 466MHz or 533MHz/Dual Digital Audio G4 Power Mac. Also use this if you're using a 400-600MHz G4 processor upgrade in a B&W or Beige Power Mac G3.

7450 - Use this if your machine is not included above. Meaning your machine has one of the following; 7450, 7455 (A and B), 7457, 7447, 7447A, 7445, and to cover all bases, the upcoming 7448.

I think thats about it.
( Last edited by Lateralus; Jul 21, 2004 at 01:54 AM. )
I like chicken
I like liver
Meow Mix, Meow Mix
Please de-liv-er
     
mikkyo  (op)
Senior User
Join Date: Feb 2002
Location: Silly Valley, Ca
Status: Offline
Reply With Quote
Jul 21, 2004, 03:26 AM
 
If you have 10.3.x run in Terminal the command
hostinfo
and you will see what CPU you have.

The optimized boinc clients help you in a couple of ways.
The client runs the worker, so the faster the client is, the faster the overall operation of the boinc/worker client relationship.
The less waiting the worker has to do for the client, the sooner it can get back to work.
Also, the boinc client may perform some functions that the worker uses.
I'm not real familiar with the whole boinc/client/worker architecture, but the core client might provide some of the crunch services the worker uses.
If this is the case, the faster client does translate directly into work getting done faster.
Why would they include benchmarks in the client if the worker does all the work?
It would make far more sense to benchmark the worker in that case.

The real gains will come with optimized clients and workers.
Unless the worker just hands data to the client and tells it to perform this or that operation on it, but that isn't the case at least with Seti and Predictor.
The Seti worker has several functions that can benefit from optimization.

Now that I'm getting the seti worker to build again, we can see how significant the gains can be.
Stay tuned.

I hope that helps some.
     
falofolio
Fresh-Faced Recruit
Join Date: Jul 2004
Status: Offline
Reply With Quote
Jul 21, 2004, 08:01 AM
 
Originally posted by mikkyo:
Now that I'm getting the seti worker to build again, we can see how significant the gains can be.
Stay tuned.

I hope that helps some.

Your story makes sense and I will certainly stay tuned on this matter. Many thanks for your explanation. And I'm greatful towards you that you took the time and effort to produce these optimzed versions.

To be frank, now that I have experienced what an optimized and especially an super-optimized version can do, I cannot wait to see a full (super-)optimized version which contains also an (super-)optimized version of the worker (in my case, SETI 3.08).

Next to that, I am starting to wonder what kind of super Mac you will get, when it would have been possible to optimize the whole Mac OS X for a particular type of processer. It would be heaven, I assume.

Thanks again!
     
reader50
Administrator
Join Date: Jun 2000
Location: California
Status: Offline
Reply With Quote
Jul 21, 2004, 11:12 AM
 
Originally posted by mikkyo:
If you have 10.3.x run in Terminal the command
hostinfo
and you will see what CPU you have.
...
This works in 10.2.8 as well.
     
jarling
Forum Regular
Join Date: Mar 2002
Location: 42 minutes from the other side of the world
Status: Offline
Reply With Quote
Jul 23, 2004, 10:42 AM
 
Long time no post, but three cheers for Mikkyo on the optimized code. It has gotten some prime publicity

http://www.aspenleaf.com/distributed...rms.html#boinc

Maybe it will draw some people to the team
     
Shaktai
Mac Elite
Join Date: Jan 2002
Location: Mile High City
Status: Offline
Reply With Quote
Jul 23, 2004, 03:52 PM
 
Originally posted by jarling:
Long time no post, but three cheers for Mikkyo on the optimized code. It has gotten some prime publicity

http://www.aspenleaf.com/distributed...rms.html#boinc

Maybe it will draw some people to the team
Seti also links to it from their Download page, and so does Predictor. I guess that is pretty much the best kudo, when the project folks don't find any fault with it. It has already attracted a couple of folks. Now if only the BOINC projects will all get back up and running, then we can have some real fun.
     
mikkyo  (op)
Senior User
Join Date: Feb 2002
Location: Silly Valley, Ca
Status: Offline
Reply With Quote
Jul 23, 2004, 04:14 PM
 
Now that I know this works...
Tiger Preview 933MHz G4 256Mb RAM Boinc 3.20 Super optimized
2004-07-19 18:15:56 [---] Benchmark results:
2004-07-19 18:15:56 [---] Number of CPUs: 1
2004-07-19 18:15:56 [---] 6272 double precision MIPS (Whetstone) per CPU
2004-07-19 18:15:56 [---] 2854 integer MIPS (Dhrystone) per CPU
2004-07-19 18:15:56 [---] Finished CPU benchmarks
     
Lateralus
Moderator Emeritus
Join Date: Sep 2001
Location: Arizona
Status: Offline
Reply With Quote
Jul 23, 2004, 04:24 PM
 
Originally posted by mikkyo:
Now that I know this works...
Tiger Preview 933MHz G4 256Mb RAM Boinc 3.20 Super optimized
2004-07-19 18:15:56 [---] Benchmark results:
2004-07-19 18:15:56 [---] Number of CPUs: 1
2004-07-19 18:15:56 [---] 6272 double precision MIPS (Whetstone) per CPU
2004-07-19 18:15:56 [---] 2854 integer MIPS (Dhrystone) per CPU
2004-07-19 18:15:56 [---] Finished CPU benchmarks
I like chicken
I like liver
Meow Mix, Meow Mix
Please de-liv-er
     
falofolio
Fresh-Faced Recruit
Join Date: Jul 2004
Status: Offline
Reply With Quote
Jul 24, 2004, 05:50 AM
 
Originally posted by mikkyo:
Now that I know this works...
Tiger Preview 933MHz G4 256Mb RAM Boinc 3.20 Super optimized
2004-07-19 18:15:56 [---] Benchmark results:
2004-07-19 18:15:56 [---] Number of CPUs: 1
2004-07-19 18:15:56 [---] 6272 double precision MIPS (Whetstone) per CPU
2004-07-19 18:15:56 [---] 2854 integer MIPS (Dhrystone) per CPU
2004-07-19 18:15:56 [---] Finished CPU benchmarks



What does this mean? So, if you will have Tiger installed in the future on a G4, you will get these incredible figures with the super optimized version??? Even with this "small" amount of RAM?
( Last edited by falofolio; Jul 24, 2004 at 09:39 AM. )
     
drweaser
Fresh-Faced Recruit
Join Date: Jul 2004
Status: Offline
Reply With Quote
Jul 24, 2004, 09:30 AM
 
I am getting this message a lot with the Super Optimized G5 seti worker....

2004-07-24 06:32:31 [SETI@home] Unrecoverable error for result 13ja04aa.29764.12688.372156.108_0 (process exited with code 131 (0x83))

Any clues?
     
Shaktai
Mac Elite
Join Date: Jan 2002
Location: Mile High City
Status: Offline
Reply With Quote
Jul 24, 2004, 01:10 PM
 
Originally posted by drweaser:
I am getting this message a lot with the Super Optimized G5 seti worker....

2004-07-24 06:32:31 [SETI@home] Unrecoverable error for result 13ja04aa.29764.12688.372156.108_0 (process exited with code 131 (0x83))

Any clues?
Are you also running the "super optimized client" or just the standard client? I was getting the same problem with the G3 version, but found that the client and worker versions were mismatched. Don't know if that was the problem or not.
     
reader50
Administrator
Join Date: Jun 2000
Location: California
Status: Offline
Reply With Quote
Jul 24, 2004, 01:36 PM
 
Originally posted by falofolio:


What does this mean? So, if you will have Tiger installed in the future on a G4, you will get these incredible figures with the super optimized version??? Even with this "small" amount of RAM?
Tiger Preview (10.4) comes with a beta version of gcc 3.5 which has an auto-vectorize flag. Translation, it finds suitable sections of code in the program and converts them to use AltiVec. This is not as effective as if the programmer coded for AltiVec directly, but it can produce a huge improvement from a normal program.

So the big speedup is not from Tiger itself, but from the compiler included. Unfortunately, the compiled binary does not run on 10.3.x or earlier, we would need gcc 3.5 on 10.3.x in order to build a compatible binary.
( Last edited by reader50; Jul 24, 2004 at 05:22 PM. )
     
drweaser
Fresh-Faced Recruit
Join Date: Jul 2004
Status: Offline
Reply With Quote
Jul 24, 2004, 01:48 PM
 
Originally posted by Shaktai:
Are you also running the "super optimized client" or just the standard client? I was getting the same problem with the G3 version, but found that the client and worker versions were mismatched. Don't know if that was the problem or not.


I had the super optimized client. Anywho.....screwed up some other stuff and don't have any seti wu's now. But predictor just kicked out a whole bunch on new WU's @ 1:45 EST.
     
falofolio
Fresh-Faced Recruit
Join Date: Jul 2004
Status: Offline
Reply With Quote
Jul 24, 2004, 03:24 PM
 
Originally posted by reader50:
..., it finds suitable sections of code in the program and converts them to use AltiVec. This is not as effective as if the programmer coded for AltiVec directly, ...
Okay, I'm not an expert in this, but why is it difficult to optimize the code of BOINC for AltiVec? Or, is it simply a matter of time?


... Unfortunately, the compiled binary does not run on 1.3 or earlier, we would need gcc 3.5 on 10.3.x in order to build a compatible binary.
What do you mean with version 1.3 here?
     
reader50
Administrator
Join Date: Jun 2000
Location: California
Status: Offline
Reply With Quote
Jul 24, 2004, 05:21 PM
 
Originally posted by falofolio:
Okay, I'm not an expert in this, but why is it difficult to optimize the code of BOINC for AltiVec? Or, is it simply a matter of time?
...
It's a matter of motivation. The project people do not optimize for any platform. Since BOINC is open source, anyone reasonably skilled with C++ could do an AltiVec optimization manually. I've looked at the code briefly, but my C++ skills are not good enough to do that yet.

...
What do you mean with version 1.3 here?
That was a typo, I meant 10.3.x, and have fixed the original post.

The 10.4 binary should run on earlier versions of OSX, the fact that it does not suggests that it links to later library versions in 10.4. Some careful digging through the source code or more careful use of compile flags might work around that, producing a binary that will run on any recent OSX version. Mikkyo might have a better idea if that is practical.
     
falofolio
Fresh-Faced Recruit
Join Date: Jul 2004
Status: Offline
Reply With Quote
Jul 24, 2004, 05:35 PM
 
Originally posted by reader50:
It's a matter of motivation. The project people do not optimize for any platform. Since BOINC is open source, anyone reasonably skilled with C++ could do an AltiVec optimization manually. I've looked at the code briefly, but my C++ skills are not good enough to do that yet.

Mikkyo might have a better idea if that is practical.

If I am correct, the (Super-)Optimized versions are not AltiVec-native. Why did Mikkyo decide not to make the Super-Optimized and/or Optimized versions AltiVec-native? I assume, it takes more time to (super-)optimize a version than make it AltiVec-native!?!

(Okay, okay, again, I'm not an expert. Therefore, it's probably more simply put than done.)
     
reader50
Administrator
Join Date: Jun 2000
Location: California
Status: Offline
Reply With Quote
Jul 24, 2004, 05:44 PM
 
Mikkyo has avoided any changes to the source code also. The optimized versions come from setting compiler flags to produce faster final code, such as by unrolling loops and leaving out debug code.

Super optimized versions involve a lot more of that, with more careful setting of compiler directives, most likely on a file-by-file basis.

The net result is an application that runs faster, but generally requires more RAM to run in.

As to automatically adding AltiVec, only the latest gcc version (3.5) supplied with Tiger has that ability. If we can get gcc 3.5 to work under 10.3.x or earlier, it would produce almost as much of a boost.
     
fillmeup
Fresh-Faced Recruit
Join Date: Jul 2004
Location: Oregon
Status: Offline
Reply With Quote
Jul 29, 2004, 06:08 PM
 
I have a couple of questions about SETI Worker.

1. Installation. I put app_info.xml and setiathome-3.8.powerpc-apple-darwin in my projects/setiathome.berkeley.edu/ folder with my downloaded work units and setiathome_3.08_powerpc-apple-darwin. I ran chmod a+rx on the app_info.xml, setiathome-3.8.powerpc-apple-darwin, and the boinc 3.20 client. I left the existing setiathome_3.08_powerpc-apple-darwin file in the folder. When I start the client I see a new line that reads "[SETI@home] Found app_info.xml; using anonymous platform". I ran bench marks and came out the same as when I installed the super optimized 3.20 client without SETI Worker. Did I set this up right?

2. What does SETI Worker do? I wondered if the purpose of SETI Worker was to increase bench mark measured speed or had some other use?

G4 - 10.3.4
Processor type: ppc7450 (PowerPC 7450)
Processor active: 0
Primary memory available: 256.00 megabytes.
Default processor set: 44 tasks, 106 threads, 1 processors
Load average: 1.73, Mach factor: 0.14

Installed:
G4 7450 Super optimized compile
G4 7450 SETI Worker

I think my lack or RAM is limiting further speed increases?
     
mikkyo  (op)
Senior User
Join Date: Feb 2002
Location: Silly Valley, Ca
Status: Offline
Reply With Quote
Jul 29, 2004, 08:15 PM
 
Yes you did it right.
The Boinc client does the benchmarking.
The Seti worker does the seti crunching.
The benchmark will always be about the same with the same Boinc Client.
The "using anonymous platform" means that seti detected you have a customized worker and is going to send you work appropriate for it.

Now you just crunch away and wait 3 days to 3 weeks to see your stats climb or the seti server go down.

Hopefully, you also attached to Predictor so you can keep crunching when Seti is unavailable.
Just set up an account at the web site and then
./boinc_3.20_powerpc_apple_darwin -attach_project http://predictor.scripps.edu
     
Kage1
Fresh-Faced Recruit
Join Date: Jul 2004
Location: Earth (I Think)
Status: Offline
Reply With Quote
Jul 31, 2004, 01:09 AM
 
Beige G3/333, OS 10.3.4, 640 Megs, Boinc 3.20

2004-07-31 00:05:24 [---] Suspending computation and network activity - running CPU benchmarks
2004-07-31 00:06:26 [---] Benchmark results:
2004-07-31 00:06:26 [---] Number of CPUs: 1
2004-07-31 00:06:26 [---] 133 double precision MIPS (Whetstone) per CPU
2004-07-31 00:06:26 [---] 227 integer MIPS (Dhrystone) per CPU
2004-07-31 00:06:26 [---] Finished CPU benchmarks
2004-07-31 00:06:27 [---] Resuming computation and network activity

Kage_
     
falofolio
Fresh-Faced Recruit
Join Date: Jul 2004
Status: Offline
Reply With Quote
Aug 6, 2004, 02:57 AM
 
Originally posted by mikkyo:
The Seti worker does the seti crunching.
[...]
The benchmark will always be about the same with the same Boinc Client.

Now you just crunch away and wait 3 days to 3 weeks to see your stats climb or the seti server go down.
[...]

Since more of a week I've been trying to get new SETI-data to crunch with the optimized version of the SETI-worker. However, in vain. Okay, I know that there were a lot of problems lately at SETI, but since the last couple of days it seems according to the messages on their site that almost everything is okay, again. However, still I cannot get new data with the optimized version.

Okay, I am now suggesting that it might be the optimized version of the SETI-worker, but to be frank I just don't know why I cannot get new data. Does anybody have had the same problem since last week or more?

Some extra info: I'm using the internal 56k-modem in my iBook to surf on the Internet. May be you get very easily a time-out message with these kind of slow connections.
( Last edited by falofolio; Aug 6, 2004 at 03:03 AM. )
     
mikkyo  (op)
Senior User
Join Date: Feb 2002
Location: Silly Valley, Ca
Status: Offline
Reply With Quote
Aug 6, 2004, 03:45 PM
 
Is that a G4 iBook or a G3?
Are you using the Altivec Seti Worker or the Seti worker I built?
     
falofolio
Fresh-Faced Recruit
Join Date: Jul 2004
Status: Offline
Reply With Quote
Aug 7, 2004, 05:36 AM
 
Originally posted by mikkyo:
Is that a G4 iBook or a G3?
Are you using the Altivec Seti Worker or the Seti worker I built?
It's an iBook G4 (ppc7450) 1 GHz with 640Mb RAM. And, I'm using BOINC in combination with the SETI-worker which are both optimized for the 7450.

BTW, which worker version is optimized for AltiVec? And am I able to run it on my iBook? (I only see an AltiVec-native version for Tiger, but not for Panther which I'm using.)
     
mikkyo  (op)
Senior User
Join Date: Feb 2002
Location: Silly Valley, Ca
Status: Offline
Reply With Quote
Aug 7, 2004, 04:59 PM
 
See this thread
http://forums.macnn.com/showthread.p...hreadid=221778
The G5 Altivec version should work on a G4, just not as fast as a G4 specific version.
It is possible that Seti isn't handing out work for custom compiled clients right now.
What is the message you are getting exactly?

Now that XCode 1.5 is out but doesn't have GCC 3.5.
Ill probably try to get backwards compatible Tiger Preview built compiled clients working this weekend.

We should get javalizard to add altivec specific instructions to the boinc client itself as well.
( Last edited by mikkyo; Aug 8, 2004 at 06:05 PM. )
     
falofolio
Fresh-Faced Recruit
Join Date: Jul 2004
Status: Offline
Reply With Quote
Aug 7, 2004, 07:12 PM
 
Originally posted by mikkyo:
It is possible that Seti isn't handing out work for custom compiled clients right now.
What is the message you are getting exactly?
Thanks for notifying me about the other forum on the AltiVec optimized version!

In about 60% of the cases, I am getting the following messages, expect that the "Deferring..."-message at the end shows every time a different time:

2004-08-08 01:09:02 [---] Fewer active results than CPUs; requesting more work
2004-08-08 01:09:02 [SETI@home] Requesting 53112 seconds of work
2004-08-08 01:09:02 [SETI@home] Sending request to scheduler: http://setiboincdata.ssl.berkeley.edu/sah_cgi/cgi
2004-08-08 01:09:12 [SETI@home] Scheduler RPC to http://setiboincdata.ssl.berkeley.edu/sah_cgi/cgi failed
2004-08-08 01:09:12 [SETI@home] Scheduler RPC to http://setiboincdata.ssl.berkeley.edu/sah_cgi/cgi failed
2004-08-08 01:09:12 [SETI@home] No schedulers responded
2004-08-08 01:09:12 [SETI@home] No schedulers responded
2004-08-08 01:09:12 [SETI@home] Deferring communication with project for 1 hours, 44 minutes, and 11 seconds
2004-08-08 01:09:12 [SETI@home] Deferring communication with project for 1 hours, 44 minutes, and 11 seconds

However, in the rest of the cases, when the connection with scheduler has succeeded, then it sends me a message telling that there is no work available. Consequently, BOINC goes in standby.
( Last edited by falofolio; Aug 7, 2004 at 07:26 PM. )
     
mikkyo  (op)
Senior User
Join Date: Feb 2002
Location: Silly Valley, Ca
Status: Offline
Reply With Quote
Aug 7, 2004, 09:40 PM
 
No schedulers responded means that the scheduler was down or overloaded, the internet was down, or some other problem occurred talking to the scheduler.
So you got no work because the scheduler never replied.

No work available means there was no work for your type of machine (possibly no work for macs, or custom compiled machines).

The Deferring communication message is automatically what happens when the scheduler doesn't respond or there is no work. The client will idle for a minute, then 2, then 20, then a hour, then 2 hours, etc checking in at the end of each rest period until it finally hears back from the scheduler.

You should definitely set up an account with Predictor at http://predictor.scripps.edu/ and then attach the client to that project so you can at least crunch something while waiting for seti units.

Be sure you are running boinc 3.20 too as 3.19 isn't accepted anymore.

After you set up a predictor account, to attach do
./boinc_3.20_powerpc-apple-darwin -attach_project http://predictor.scripps.edu

I hate to see folks idle when they could be crunching and helping the team.
     
monsoonstorm
Fresh-Faced Recruit
Join Date: Aug 2004
Status: Offline
Reply With Quote
Aug 8, 2004, 11:50 AM
 
Cool!

Mac OS 10.3.4, G4 DP500, 1GB RAM

Before - per CPU
331 Whet
743 Dhry

Now - per CPU
1024 Whet
833 Dhry

now If only I could get some workunits... I would love to join predictor too, but their account creation is STILL disabled

I guess I've just got a lot of idle time then....


Will edit with my powerbook G4 stats later.

MonsoonStorm
     
Shaktai
Mac Elite
Join Date: Jan 2002
Location: Mile High City
Status: Offline
Reply With Quote
Aug 8, 2004, 01:17 PM
 
Originally posted by monsoonstorm:
now If only I could get some workunits... I would love to join predictor too, but their account creation is STILL disabled

MonsoonStorm
Predictor had been stable all weekend, despite the key people taking a weekend off to relax. That means there is a good chance account creation will be re-enabled this week.
     
mikkyo  (op)
Senior User
Join Date: Feb 2002
Location: Silly Valley, Ca
Status: Offline
Reply With Quote
Aug 9, 2004, 10:02 PM
 
Ya know, until then, someone on our team could give out their userid for others to use until account creation is re-enabled...


No one can do anything other than crunch for you with just the userid.
     
mikkyo  (op)
Senior User
Join Date: Feb 2002
Location: Silly Valley, Ca
Status: Offline
Reply With Quote
Aug 10, 2004, 07:06 PM
 
Account creation is back on!!

http://predictor.scripps.edu/create_account_form.php

Join now!
     
mikkyo  (op)
Senior User
Join Date: Feb 2002
Location: Silly Valley, Ca
Status: Offline
Reply With Quote
Aug 15, 2004, 04:01 AM
 
I got the benchmark part working...

Dual 800 G4 1.5Gb RAM 10.3.5 boinc 3.20 super-altivec'd gcc3.5 compile
2004-08-15 00:54:11 [---] Benchmark results:
2004-08-15 00:54:11 [---] Number of CPUs: 2
2004-08-15 00:54:11 [---] 5397 double precision MIPS (Whetstone) per CPU
2004-08-15 00:54:11 [---] 2376 integer MIPS (Dhrystone) per CPU
2004-08-15 00:54:11 [---] Finished CPU benchmarks

I really don't know if I can get the rest of the client building/working on Tiger Preview and then running on 10.3.X.
     
Shaktai
Mac Elite
Join Date: Jan 2002
Location: Mile High City
Status: Offline
Reply With Quote
Aug 15, 2004, 05:01 AM
 
Wow!

Now those are some impressive numbers.

I've made the decision, that either this month or next month, I am getting some kind of a new Power Mac or G5 iMac (when available). Not sure what yet, since I haven't finalized what I can afford, but something.
( Last edited by Shaktai; Aug 15, 2004 at 05:10 AM. )
     
fillmeup
Fresh-Faced Recruit
Join Date: Jul 2004
Location: Oregon
Status: Offline
Reply With Quote
Aug 29, 2004, 08:39 PM
 
Help mikkyo!!!

Here's my benchmark with the new BOINC 4.05:
2004-08-29 17:25:43 [---] Running CPU benchmarks
2004-08-29 17:25:43 [---] Suspending computation and network activity - running CPU benchmarks
2004-08-29 17:26:44 [---] Benchmark results:
2004-08-29 17:26:44 [---] Number of CPUs: 1
2004-08-29 17:26:44 [---] 317 double precision MIPS (Whetstone) per CPU
2004-08-29 17:26:44 [---] 744 integer MIPS (Dhrystone) per CPU
2004-08-29 17:26:44 [---] Finished CPU benchmarks

It absolutely !!!

Please give us some hope!
We need your magical Super Optimized client right away!
(Seti Worker would be nice too.)
     
mikkyo  (op)
Senior User
Join Date: Feb 2002
Location: Silly Valley, Ca
Status: Offline
Reply With Quote
Aug 29, 2004, 10:40 PM
 
What CPU was that?
I haven't tried building the latest boinc, which project is this for?
     
Shaktai
Mac Elite
Join Date: Jan 2002
Location: Mile High City
Status: Offline
Reply With Quote
Aug 30, 2004, 02:35 AM
 
Originally posted by mikkyo:
What CPU was that?
I haven't tried building the latest boinc, which project is this for?
Seti just moved to Boinc 4.x. Climate Prediction is also on 4.x Predictor will be moving over to 4.x in another week or two. As soon as the last CASP units are submitted they are planning the switchover and some improved server hardware. They're testing internally now, so within another two weeks, everyone should be on the latest BOINC platform. For now though, anyone who moves to 4.x, won't be able to crunch on predictor, until they make the move.
     
mikkyo  (op)
Senior User
Join Date: Feb 2002
Location: Silly Valley, Ca
Status: Offline
Reply With Quote
Aug 30, 2004, 02:41 AM
 
iBook G3 600 OS X 10.3.5 384Mb RAM boinc 4.06
2004-08-29 23:32:18 [---] Benchmark results:
2004-08-29 23:32:18 [---] Number of CPUs: 1
2004-08-29 23:32:18 [---] 5705 double precision MIPS (Whetstone) per CPU
2004-08-29 23:32:18 [---] 1925 integer MIPS (Dhrystone) per CPU

Now to see if it can actually get and return units..
     
fillmeup
Fresh-Faced Recruit
Join Date: Jul 2004
Location: Oregon
Status: Offline
Reply With Quote
Aug 30, 2004, 07:14 AM
 
Sorry, I was in a panic.

Machine Model: iMac
CPU Type: PowerPC G4 (2.1)
Number Of CPUs: 1
CPU Speed: 700 MHz
L2 Cache (per CPU): 256 KB
Memory: 256 MB
Bus Speed: 100 MHz

G4 - 10.3.5
Processor type: ppc7450 (PowerPC 7450)
Processor active: 0
Primary memory available: 256.00 megabytes.
Default processor set: 44 tasks, 106 threads, 1 processors
Load average: 1.73, Mach factor: 0.14

The benchmark I mentioned above was from running Seti, but my Climate Prediction numbers were just as low. Probably the same for all BOINC 4.05.
     
javalizard
Fresh-Faced Recruit
Join Date: Mar 2001
Location: Pittsburgh, PA
Status: Offline
Reply With Quote
Aug 31, 2004, 12:23 PM
 
I have developed the altivec optimized SETI Worker from the source about a month ago. The project has been having serious problems with keeping work queued. And now the latest move to version 4.

As soon as the source is out there for the seti worker, I will update it with my altivec code and we should see a 50% speed increase on the G4. The G3 and the G4 are about the same speed except the G4 has altivec. The G5 will see a 25% speed-up. The memory architecture in the G5 is so amazing that non-altivec functions are twice as fast thus leading to the smaller performance gain by the altivec code.

I will post this in the "G5 altivec" thread. probably in a week or two.

Brad
     
Shaktai
Mac Elite
Join Date: Jan 2002
Location: Mile High City
Status: Offline
Reply With Quote
Aug 31, 2004, 03:49 PM
 
Originally posted by mikkyo:
iBook G3 600 OS X 10.3.5 384Mb RAM boinc 4.06
2004-08-29 23:32:18 [---] Benchmark results:
2004-08-29 23:32:18 [---] Number of CPUs: 1
2004-08-29 23:32:18 [---] 5705 double precision MIPS (Whetstone) per CPU
2004-08-29 23:32:18 [---] 1925 integer MIPS (Dhrystone) per CPU

Now to see if it can actually get and return units..
Ye gads! did you post the correct machine speed? That is incredible if it works.
     
Shaktai
Mac Elite
Join Date: Jan 2002
Location: Mile High City
Status: Offline
Reply With Quote
Aug 31, 2004, 03:50 PM
 
Originally posted by javalizard:
I will post this in the "G5 altivec" thread. probably in a week or two.

Brad
Thanks for the update. I am sure we will have lots of willing testers when you have it.
     
fillmeup
Fresh-Faced Recruit
Join Date: Jul 2004
Location: Oregon
Status: Offline
Reply With Quote
Aug 31, 2004, 04:13 PM
 
Originally posted by mikkyo:
iBook G3 600 OS X 10.3.5 384Mb RAM boinc 4.06
2004-08-29 23:32:18 [---] Benchmark results:
2004-08-29 23:32:18 [---] Number of CPUs: 1
2004-08-29 23:32:18 [---] 5705 double precision MIPS (Whetstone) per CPU
2004-08-29 23:32:18 [---] 1925 integer MIPS (Dhrystone) per CPU

Now to see if it can actually get and return units..
There is NO WAY that a G3 600 could possibly do those numbers. I'm running a G4 7450 700Mhz and on it's best day only did just under 1000 for Whetstone. And that was only with the Super Optimized Client.

If they are correct, then I worship at your feet O' great one.
     
mikkyo  (op)
Senior User
Join Date: Feb 2002
Location: Silly Valley, Ca
Status: Offline
Reply With Quote
Aug 31, 2004, 05:46 PM
 
Well, just wait until your try the version for the 7450

The newer clients will be available soon.
     
reader50
Administrator
Join Date: Jun 2000
Location: California
Status: Offline
Reply With Quote
Sep 1, 2004, 12:03 PM
 
They are up. Thanks mikkyo.
     
fillmeup
Fresh-Faced Recruit
Join Date: Jul 2004
Location: Oregon
Status: Offline
Reply With Quote
Sep 1, 2004, 04:19 PM
 
Thank You O' BOINC client God

It's running Great!
     
drweaser
Fresh-Faced Recruit
Join Date: Jul 2004
Status: Offline
Reply With Quote
Sep 1, 2004, 05:12 PM
 
Results from the G5 version (4.06):

2004-09-01 17:06:47 [---] Running CPU benchmarks
2004-09-01 17:06:47 [---] Suspending computation and network activity - running CPU benchmarks
2004-09-01 17:07:48 [---] Benchmark results:
2004-09-01 17:07:48 [---] Number of CPUs: 2
2004-09-01 17:07:48 [---] 3854 double precision MIPS (Whetstone) per CPU
2004-09-01 17:07:48 [---] 4161 integer MIPS (Dhrystone) per CPU
2004-09-01 17:07:48 [---] Finished CPU benchmarks
2004-09-01 17:07:49 [---] Resuming computation and ne

compared to the super optimized 3.20:

2004-07-19 21:34:21 [---] Running CPU benchmarks
2004-07-19 21:34:21 [---] Suspending computation and network activity - running CPU benchmarks
2004-07-19 21:35:22 [---] Benchmark results:
2004-07-19 21:35:22 [---] Number of CPUs: 2
2004-07-19 21:35:22 [---] 3858 double precision MIPS (Whetstone) per CPU
2004-07-19 21:35:22 [---] 3816 integer MIPS (Dhrystone) per CPU
2004-07-19 21:35:22 [---] Finished CPU benchmarks
2004-07-19 21:35:23 [---] Resuming computation and network activity

Looks like a slight improvement on the Dhrystone side at this point. Can't wait to see the 10.4 versions any a few months....

drweaser

<a href="http://www.boinc.dk/index.php?page=user_statistics&project=pah&userid= 1868"><img border=0 width="280" height="70" src="http://1868.pah.sig.boinc.dk?133"></a>
     
Shaktai
Mac Elite
Join Date: Jan 2002
Location: Mile High City
Status: Offline
Reply With Quote
Sep 1, 2004, 10:11 PM
 
Here is the new 4.06 (optimized) on my iBook 600 G3 with 640mb RAM. Very sweet, this is a significant improvement over the stock 4.06 client. Actual production on climate prediction isn't quite as much, but it still helps.

2004-09-01 19:06:15 [---] Benchmark results:
2004-09-01 19:06:15 [---] Number of CPUs: 1
2004-09-01 19:06:15 [---] 1112 double precision MIPS (Whetstone) per CPU
2004-09-01 19:06:15 [---] 1001 integer MIPS (Dhrystone) per CPU
2004-09-01 19:06:15 [---] Finished CPU benchmarks
     
Lateralus
Moderator Emeritus
Join Date: Sep 2001
Location: Arizona
Status: Offline
Reply With Quote
Sep 1, 2004, 10:33 PM
 
New score for Dual 1.40GHz G4/1GB using Super-Optimized 4.06 7450 + SETI Worker:
2004-09-01 21:30:29 [---] Benchmark results:
2004-09-01 21:30:29 [---] Number of CPUs: 2
2004-09-01 21:30:29 [---] 1943 double precision MIPS (Whetstone) per CPU (Up from 1935)
2004-09-01 21:30:29 [---] 2194 integer MIPS (Dhrystone) per CPU (Down from 2200)
2004-09-01 21:30:29 [---] Finished CPU benchmarks
I like chicken
I like liver
Meow Mix, Meow Mix
Please de-liv-er
     
Lateralus
Moderator Emeritus
Join Date: Sep 2001
Location: Arizona
Status: Offline
Reply With Quote
Oct 13, 2004, 11:42 PM
 
1.1GHz PowerPC 750GX G3 - 1MB L2/640MBs using Super-Optimized G3 4.06 + SETI Worker:
2004-10-13 22:46:18 [---] Benchmark results:
2004-10-13 22:46:18 [---] Number of CPUs: 1
2004-10-13 22:46:18 [---] 2002 double precision MIPS (Whetstone) per CPU
2004-10-13 22:46:18 [---] 1797 integer MIPS (Dhrystone) per CPU
2004-10-13 22:46:18 [---] Finished CPU benchmarks

Very impressive.
( Last edited by Lateralus; Oct 13, 2004 at 11:52 PM. )
I like chicken
I like liver
Meow Mix, Meow Mix
Please de-liv-er
     
 
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 08:40 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.,