[mikkyo]

run your boinc client with the command
./boinc_3.1.9_powerpc-apple-darwin -run_cpu_benchmarks
to see how fast your machine is.
Then you can do something like this:

G5 Dual 2.0 GHz 1.5GB Ram OS X 10.3.4
2004-07-07 16:36:03 [---] Suspending computation and network activity -
running CPU benchmarks
2004-07-07 16:37:04 [---] Benchmark results:
2004-07-07 16:37:04 [---] Number of CPUs: 2
2004-07-07 16:37:04 [---] 2002 double precision MIPS (Whetstone) per CPU
2004-07-07 16:37:04 [---] 3625 integer MIPS (Dhrystone) per CPU
2004-07-07 16:37:04 [---] Finished CPU benchmarks

G5 1.8GHz 1Gb Ram OS X 10.3.4
2004-07-07 16:49:00 [---] Suspending computation and network activity -
running CPU benchmarks
2004-07-07 16:50:01 [---] Benchmark results:
2004-07-07 16:50:01 [---] Number of CPUs: 1
2004-07-07 16:50:01 [---] 959 double precision MIPS (Whetstone) per CPU
2004-07-07 16:50:01 [---] 3218 integer MIPS (Dhrystone) per CPU
2004-07-07 16:50:01 [---] Finished CPU benchmarks

[Shaktai]

My power house iBook 600, 256 mb RAM, OS 10.3.4:
2004-07-05 23:53:43 [---] Benchmark results:
2004-07-05 23:53:43 [---] Number of CPUs: 1
2004-07-05 23:53:43 [---] 563 double precision MIPS (Whetstone) per CPU
2004-07-05 23:53:43 [---] 904 integer MIPS (Dhrystone) per CPU
(based on mikkyo's optimized G3 client)

mikkyo, your 1.8 ghz whetstone seems a bit low?
ALL: note that on the dual, those numbers are per CPU, so total machine power is double that.

An Athlon 3000 (stock E-machines with no tweaking)
--- - 2004-07-07 18:21:58 - Number of CPUs: 1
--- - 2004-07-07 18:21:58 - 2671 double precision MIPS (Whetstone) per CPU
--- - 2004-07-07 18:21:58 - 4757 integer MIPS (Dhrystone) per CPU

Benchmarks are based on the BOINC core client only, and do not neccesarily reflect actual project performance, since each project varies.

[bborofka]

1GHz G4 PowerBook 768MB RAM, OS 10.3.4
2004-07-07 19:03:27 [---] Benchmark results:
2004-07-07 19:03:27 [---] Number of CPUs: 1
2004-07-07 19:03:27 [---] 476 double precision MIPS (Whetstone) per CPU
2004-07-07 19:03:27 [---] 1412 integer MIPS (Dhrystone) per CPU

My 1.6GHz P4 system at work gets nearly double the Whetstone and much higher Dhrystone benchmarks. Will the Mac client receive any optimizations that could speed these up? Altivec?

[Shaktai]

Originally posted by bborofka:
My 1.6GHz P4 system at work gets nearly double the Whetstone and much higher Dhrystone benchmarks. Will the Mac client receive any optimizations that could speed these up? Altivec?

Have you checked out mikkyo's optimized clients at:
http://members.dslextreme.com/~reade...eam/boinc.html

My ibook benchmark is based on that client. Bus speed might also impact the benchmarks.

As far as Altivec optimization, that will have to take place at the project level, and will vary project by project. BOINC is just a common framework that the actual application programs run within. It's primary advantage is that it handles the "operational overhead" allowing project developers to focus on their specific application and science.

[Lateralus]

Dual G4s (7455Bs) @ 1.33GHz/1GB of RAM/10.3.4

2004-07-07 22:40:38 [---] Benchmark results:
2004-07-07 22:40:38 [---] Number of CPUs: 2
2004-07-07 22:40:38 [---] 932 double precision MIPS (Whetstone) per CPU
2004-07-07 22:40:38 [---] 1919 integer MIPS (Dhrystone) per CPU

Based on Mikkyo's optimized client as well.

[Lateralus]

And with some overclocking:

Dual G4s (7455Bs) @ 1.40GHz/1GB of RAM/10.3.4

2004-07-07 22:56:49 [---] Benchmark results:
2004-07-07 22:56:49 [---] Number of CPUs: 2
2004-07-07 22:56:49 [---] 981 double precision MIPS (Whetstone) per CPU
2004-07-07 22:56:49 [---] 2024 integer MIPS (Dhrystone) per CPU

Dual G4s (7455Bs) @ 1.47GHz/1GB of RAM/10.3.4

2004-07-07 23:05:21 [---] Benchmark results:
2004-07-07 23:05:21 [---] Number of CPUs: 2
2004-07-07 23:05:21 [---] 1025 double precision MIPS (Whetstone) per CPU
2004-07-07 23:05:21 [---] 2121 integer MIPS (Dhrystone) per CPU

[bborofka]

Originally posted by Shaktai:
Have you checked out mikkyo's optimized clients at:
http://members.dslextreme.com/~reade...eam/boinc.html

My ibook benchmark is based on that client. Bus speed might also impact the benchmarks.

As far as Altivec optimization, that will have to take place at the project level, and will vary project by project. BOINC is just a common framework that the actual application programs run within. It's primary advantage is that it handles the "operational overhead" allowing project developers to focus on their specific application and science.

Thanks for the link, I didn't realize he made clients for Panther. I'm curious, what's in the difference in the code between the G4 and G3 if there's no Altivec enhancements?

Using the 3.19 client gives my much better Whetstone numbers:

2004-07-07 21:05:16 [---] 706 double precision MIPS (Whetstone) per CPU
2004-07-07 21:05:16 [---] 1408 integer MIPS (Dhrystone) per CPU

My Dhrystone numbers dropped just a few and I guess that's attributed to chance.

[mikkyo]

Cache sizes are different between the processors, as well as floating point behavior.
The clients are compile-time optimized, more so than in the default debug builds that have been released.


p.s. For those of you with G5s, 10.3 is a must. The system level optimizations really shine on the G5.

[mikkyo]

10.3.4 Dual G5 2GHz 4Gb RAM
2004-07-09 15:50:38 [---] Running CPU benchmarks
2004-07-09 15:50:38 [---] Suspending computation and network activity
- running CPU benchmarks
2004-07-09 15:51:40 [---] Benchmark results:
2004-07-09 15:51:40 [---] Number of CPUs: 2
2004-07-09 15:51:40 [---] 2395 double precision MIPS (Whetstone) per CPU
2004-07-09 15:51:40 [---] 4560 integer MIPS (Dhrystone) per CPU
2004-07-09 15:51:40 [---] Finished CPU benchmarks

[Shaktai]

Originally posted by mikkyo:
10.3.4 Dual G5 2.5GHz 512Mb RAM
2004-07-09 15:50:38 [---] Running CPU benchmarks
2004-07-09 15:50:38 [---] Suspending computation and network activity
- running CPU benchmarks
2004-07-09 15:51:40 [---] Benchmark results:
2004-07-09 15:51:40 [---] Number of CPUs: 2
2004-07-09 15:51:40 [---] 2395 double precision MIPS (Whetstone) per CPU
2004-07-09 15:51:40 [---] 4560 integer MIPS (Dhrystone) per CPU
2004-07-09 15:51:40 [---] Finished CPU benchmarks

Droooolllll!!!!!!!
I want one.... Hey, how did you get hold of one so soon???

[mikkyo]

Not mine.
Told a friend about the team.
I hope he joins.

Edit:
Okay. A confused friend...
He has had the machine for a month, it is a 2.0 and has 4Gb of RAM.
I fixed the other post.

[Shaktai]

Originally posted by mikkyo:
Edit:
Okay. A confused friend...
He has had the machine for a month, it is a 2.0 and has 4Gb of RAM.
I fixed the other post.

Okay, he can still join. That way I won't be the only confused one around here.
So, if that is a 2.0, and we increase those numbers by 25% to represent a 2.5....

Now I really, really want one.

[Lateralus]

How much of a difference does memory make?

[bborofka]

3.20 is out for all platforms. I get lower Whetstone numbers with it than mikkyo's 3.19 build.

3.19
2004-07-12 14:23:48 [---] 704 double precision MIPS (Whetstone) per CPU
2004-07-12 14:23:48 [---] 1402 integer MIPS (Dhrystone) per CPU

3.20
2004-07-12 14:25:07 [---] 478 double precision MIPS (Whetstone) per CPU
2004-07-12 14:25:07 [---] 1413 integer MIPS (Dhrystone) per CPU

Dhrystone seems to fluctuate a lot regardless of the build.

[Shaktai]

Originally posted by bborofka:
3.20 is out for all platforms. I get lower Whetstone numbers with it than mikkyo's 3.19 build.

That's because it still has a lot of debug code. However, it is supposed to fix some of the unit and and download errors.

mikkyo's compile is definitely faster, and not just on benchmarks. From what I can tell, my iBook 600 is at least 10% faster in daily production with the special G3 compile then without. (about 20-25% on floating point) I would love to know how much faster the G4 & G5 compiles are.

Maybe mikkyo can find some time to re-compile 3.20 on of these days soon.

[mikkyo]

I can't even get the current source to build.

Oh and yes RAM seems to help a lot, especially on G5s.
I imagine faster DDR RAM is a bit better, just like in the PC world.

[mikkyo]

iBook 600 OS X 10.3.4 384Mb RAM boinc 3.20 optimized
2004-07-13 23:25:24 [---] Suspending computation and network activity
- running CPU benchmarks
2004-07-13 23:26:25 [---] Benchmark results:
2004-07-13 23:26:25 [---] Number of CPUs: 1
2004-07-13 23:26:25 [---] 563 double precision MIPS (Whetstone) per CPU
2004-07-13 23:26:25 [---] 913 integer MIPS (Dhrystone) per CPU
2004-07-13 23:26:25 [---] Finished CPU benchmarks

[Shaktai]

Originally posted by mikkyo:
iBook 600 OS X 10.3.4 384Mb RAM boinc 3.20 optimized
2004-07-13 23:25:24 [---] Suspending computation and network activity
- running CPU benchmarks
2004-07-13 23:26:25 [---] Benchmark results:
2004-07-13 23:26:25 [---] Number of CPUs: 1
2004-07-13 23:26:25 [---] 563 double precision MIPS (Whetstone) per CPU
2004-07-13 23:26:25 [---] 913 integer MIPS (Dhrystone) per CPU
2004-07-13 23:26:25 [---] Finished CPU benchmarks

That is a big improvement over the generic version. Better then I saw with 3.19 optimized.

My iBook 600 but with 256mb Ram running the non-optimized version. Your version gives a very nice FP boost over the genric below.
2004-07-13 04:15:00 [---] 369 double precision MIPS (Whetstone) per CPU
2004-07-13 04:15:00 [---] 891 integer MIPS (Dhrystone) per CPU

[bborofka]

Originally posted by mikkyo:
iBook 600 OS X 10.3.4 384Mb RAM boinc 3.20 optimized
2004-07-13 23:25:24 [---] Suspending computation and network activity
- running CPU benchmarks
2004-07-13 23:26:25 [---] Benchmark results:
2004-07-13 23:26:25 [---] Number of CPUs: 1
2004-07-13 23:26:25 [---] 563 double precision MIPS (Whetstone) per CPU
2004-07-13 23:26:25 [---] 913 integer MIPS (Dhrystone) per CPU
2004-07-13 23:26:25 [---] Finished CPU benchmarks

Could you post that on your site?

[mikkyo]

Dual G4 800 1.5GB RAM Mac OS X 10.3.4
2004-07-14 09:02:33 [---] Suspending computation and network activity
- running CPU benchmarks
2004-07-14 09:03:34 [---] Benchmark results:
2004-07-14 09:03:34 [---] Number of CPUs: 2
2004-07-14 09:03:34 [---] 570 double precision MIPS (Whetstone) per CPU
2004-07-14 09:03:34 [---] 1226 integer MIPS (Dhrystone) per CPU
2004-07-14 09:03:34 [---] Finished CPU benchmarks

[DrBoar]

G4/1200 MHz
543 double precision MIPS (Whetstone) per CPU
1663 integer MIPS (Dhrystone) per CPU

AMD K6/233
257 double precision MIPS (Whetstone) per CPU
410 integer MIPS (Dhrystone) per CPU

G3/350
Client did not give a score

[mikkyo]

G4 867MHz 768Mb RAM OS X 10.3.4 Boinc 3.20 G4 version
2004-07-14 15:02:49 [---] Suspending computation and network activity
- running CPU benchmarks
2004-07-14 15:03:50 [---] Benchmark results:
2004-07-14 15:03:50 [---] Number of CPUs: 1
2004-07-14 15:03:50 [---] 583 double precision MIPS (Whetstone) per CPU
2004-07-14 15:03:50 [---] 1277 integer MIPS (Dhrystone) per CPU
2004-07-14 15:03:50 [---] Finished CPU benchmarks

[Shaktai]

Here is mine with the new G3 compile. Very sweet!
iBook 600 - 256mb RAM
2004-07-14 21:10:18 [---] Number of CPUs: 1
2004-07-14 21:10:18 [---] 573 double precision MIPS (Whetstone) per CPU
2004-07-14 21:10:18 [---] 926 integer MIPS (Dhrystone) per CPU

[mikkyo]

There you go.
If the G3 can keep up with the G4, you know the code could be better.
No altivec use whatsoever.
Too bad the Predictor worker isn't open source.
I bet it would like Altivec a lot.

[Shaktai]

Originally posted by mikkyo:
Too bad the Predictor worker isn't open source.
I bet it would like Altivec a lot.

Let's get more Mac's crunching, and maybe when the CASP 6 crunch is over, they will consider it. Come September, I will have something to replace my now gone PowerMac 800, I just don't know what it will be yet. I really can't stand doing my photo work on a PC even with a high end P4, and the iBook is just too small a screen (and slow).

[DrBoar]

604 E 200 MHz CPU running OS 10.0.4 with only 184 MB RAM

004-07-15 09:09:20 [---] Benchmark results:
2004-07-15 09:09:20 [---] Number of CPUs: 1
2004-07-15 09:09:20 [---] 125 double precision MIPS (Whetstone) per CPU
2004-07-15 09:09:20 [---] -79072182 integer MIPS (Dhrystone) per CPU

The negative number is interesting and also that it had lower whetstone score than the AMD K6/233 despite that the 604E is almost twise as fast in both F@H and SETI

A dual 1.25 G4
590 double precision MIPS (Whetstone) per CPU
1804 integer MIPS (Dhrystone) per CPU
With a score of 543 and 1663 on a upgraded sawtooth with a much slower bus this seem to indicate that bus speed is not critical at least for the benchmarking.

Pentium III 730 MHz
Measured floating point speed 860.83 million ops/sec
Measured integer speed 1537.88 million ops/sec

Pentium 4 2.66
1675 double precision MIPS (Whetstone) per CPU
3673 integer MIPS (Dhrystone) per CPU
@ the teams service

[mikkyo]

The negative number is a bit odd, maybe the variable size is different on that machine.
That would make sense then.
Did you recompile the client on the 604? You probably should.

Those other machines should give a big boost!
It will take many days for all the credit to start to stabilize, so don't worry if it seems low for awhile.

Now if I can just get Cobramac to switch over to Predictor..

[bborofka]

PowerBook G4 1GHz - 3.20 G4 optimized build
2004-07-15 10:28:26 [---] 694 double precision MIPS (Whetstone) per CPU
2004-07-15 10:28:26 [---] 1398 integer MIPS (Dhrystone) per CPU

[mikkyo]

Here is the same iBook 600 with my newly optimized client
If I only had more RAM! Oooo it likes RAM.

iBook G3 600 OS X 10.3.4 384Mb RAM boinc 3.20 optimized even more
2004-07-15 23:32:31 [---] Suspending computation and network activity
- running CPU benchmarks
2004-07-15 23:33:32 [---] Benchmark results:
2004-07-15 23:33:32 [---] Number of CPUs: 1
2004-07-15 23:33:32 [---] 1030 double precision MIPS (Whetstone) per CPU
2004-07-15 23:33:32 [---] 991 integer MIPS (Dhrystone) per CPU
2004-07-15 23:33:32 [---] Finished CPU benchmarks

[Shaktai]

Originally posted by mikkyo:
Here is the same iBook 600 with my newly optimized client
If I only had more RAM! Oooo it likes RAM.

iBook G3 600 OS X 10.3.4 384Mb RAM boinc 3.20 optimized even more
2004-07-15 23:32:31 [---] Suspending computation and network activity
- running CPU benchmarks
2004-07-15 23:33:32 [---] Benchmark results:
2004-07-15 23:33:32 [---] Number of CPUs: 1
2004-07-15 23:33:32 [---] 1030 double precision MIPS (Whetstone) per CPU
2004-07-15 23:33:32 [---] 991 integer MIPS (Dhrystone) per CPU
2004-07-15 23:33:32 [---] Finished CPU benchmarks

Wow!!!!
Might have to get me some more RAM for mine and give that client a whirl. Been wanting to add more for a long time. Maybe now is the time.

And if you can get Cobramac to switch over, I'll be all too glad to settle for third or fourth place. (I noticed you screaming up from behind me.)

[mikkyo]

New G4 version on same Dual 800
Dual G4 800 1.5GB RAM Mac OS X 10.3.4 boinc 3.20 super optimized
2004-07-15 23:44:10 [---] Suspending computation and network activity
- running CPU benchmarks
2004-07-15 23:45:11 [---] Benchmark results:
2004-07-15 23:45:11 [---] Number of CPUs: 2
2004-07-15 23:45:11 [---] 1119 double precision MIPS (Whetstone) per CPU
2004-07-15 23:45:11 [---] 1262 integer MIPS (Dhrystone) per CPU
2004-07-15 23:45:11 [---] Finished CPU benchmarks

[Lateralus]

Athlon XP 2600+ (Thorton core). 2.1GHz, 333MHz FSB, 512MBs of DDR.

--- - 2004-07-16 22:50:18 - Benchmark results:
--- - 2004-07-16 22:50:18 - Number of CPUs: 1
--- - 2004-07-16 22:50:18 - 2536 double precision MIPS (Whetstone) per CPU
--- - 2004-07-16 22:50:18 - 4449 integer MIPS (Dhrystone) per CPU
--- - 2004-07-16 22:50:18 - Finished CPU benchmarks

[mikkyo]

The new super optimized compiles are working well and so reader50 has added them to the client page.
If you haven't picked up a 3.20 client before today, be sure to grab the newest ones.
I think you will be very happy with the difference.
I'll try and get super optimized seti workers built and tested so those folks on Seti can really see the benefits.
The new 3.20 client will not work properly with OS X 10.2.8, so a rebuild for the 10.2.8 users is not available. Everyone should upgrade to 10.3.4 anyway, it is worth it if you crunch anything.

[Shaktai]

Originally posted by mikkyo:
The new super optimized compiles are working well and so reader50 has added them to the client page.
If you haven't picked up a 3.20 client before today, be sure to grab the newest ones.
I think you will be very happy with the difference.
I'll try and get super optimized seti workers built and tested so those folks on Seti can really see the benefits.
The new 3.20 client will not work properly with OS X 10.2.8, so a rebuild for the 10.2.8 users is not available. Everyone should upgrade to 10.3.4 anyway, it is worth it if you crunch anything.

Looks like the "super optimized" G3 compile wants more memory. 256mb aren't enough, the benchmarks with 256mb were pitiful. Worse then the standard client. Guess its time to pick up that 512mb memory chip for my iBook.

2004-07-17 20:31:18 [---] Number of CPUs: 1
2004-07-17 20:31:18 [---] 234 double precision MIPS (Whetstone) per CPU
2004-07-17 20:31:18 [---] 399 integer MIPS (Dhrystone) per CPU

We need to pull the 3.19 clients including the Jaguar one. As of this next week 3.20 will be required. Apparently even the 3.19 Jaguar client has problems with some platforms.

[Lateralus]

I am quite dissapointed that my 2.1GHz Athlon is faster than my 1.40GHz G4s combined.

[mikkyo]

Did you try the newest 3.20 client on your G4?
Get the one that matches your processor type and run the benchmarks again

[Shaktai]

PowerMacMan

What mikkyo said. Don''t know if you are running the "standard" download client, or one of the "optimized clients. You have plenty of RAM, if you haven't tried it, then check out the "optimized clients" HERE

Try the "super optimized" for the 7450 first, and if that doesn't work well, try the standard without debug.

[studiodave56]

quote:
Originally posted by mikkyo:
The new super optimized compiles are working well and so reader50 has added them to the client page.
If you haven't picked up a 3.20 client before today, be sure to grab the newest ones.
I think you will be very happy with the difference.
I'll try and get super optimized seti workers built and tested so those folks on Seti can really see the benefits.
The new 3.20 client will not work properly with OS X 10.2.8, so a rebuild for the 10.2.8 users is not available. Everyone should upgrade to 10.3.4 anyway, it is worth it if you crunch anything.


Looks like the "super optimized" G3 compile wants more memory. 256mb aren't enough, the benchmarks with 256mb were pitiful. Worse then the standard client. Guess its time to pick up that 512mb memory chip for my iBook.

2004-07-17 20:31:18 [---] Number of CPUs: 1
2004-07-17 20:31:18 [---] 234 double precision MIPS (Whetstone) per CPU
2004-07-17 20:31:18 [---] 399 integer MIPS (Dhrystone) per CPU

We need to pull the 3.19 clients including the Jaguar one. As of this next week 3.20 will be required. Apparently even the 3.19 Jaguar client has problems with some platforms.
__________________

I just installed the 3.20 "super optimized" for the 7410 CPU and I get,

1022 double precision MIPS (Whetstone) per CPU
815 integer MIPS (Dhrystone) per CPU

This is much better, Thanks

"Bronze keyboard" Powerbook with Powerlogix G4500 upgrade, 512 meg ram and OS 10.3.4.

[Lateralus]

Originally posted by Shaktai:
PowerMacMan

What mikkyo said. Don''t know if you are running the "standard" download client, or one of the "optimized clients. You have plenty of RAM, if you haven't tried it, then check out the "optimized clients" HERE

Try the "super optimized" for the 7450 first, and if that doesn't work well, try the standard without debug.

I had been using the optimized client that was being clamored about at the beginning of this thread. I just downloaded the 'super optimized' one and my Whetstone has gone through the roof.

New score for Dual 1.40GHz G4/1GB:
2004-07-18 22:54:19 [---] Benchmark results:
2004-07-18 22:54:19 [---] Number of CPUs: 2
2004-07-18 22:54:19 [---] 1935 double precision MIPS (Whetstone) per CPU (Up from 981 with previous optimized client)
2004-07-18 22:54:19 [---] 2200 integer MIPS (Dhrystone) per CPU (Up from 2024 with previous optimized client)
2004-07-18 22:54:19 [---] Finished CPU benchmarks

[Shaktai]

Originally posted by PowerMacMan:
I had been using the optimized client that was being clamored about at the beginning of this thread. I just downloaded the 'super optimized' one and my Whetstone has gone through the roof.

Nothing like a little optimization to really make a G4 sing. That looks much better. Hopefully the real world results will show the benefit. Since that is for one CPU, your dual looks like it matches up well with the Athlon 2600 now.

[DrBoar]

Some questions
Is the right way of updating the client by just replacing the boink... file in the used folder or do the whole stuff thing to be replaced?

For a dual 1.25 is the superoptimized client the 7450 the one to use?

For the 7555A CPU will the 7550 work or sould I step back to the plain G4 client?

[mikkyo]

Just replace yoru existing 3.20 client or drop the new one in the directory.
You don't have to remove anything, except the old client to avoid confusion.
The most you should have to do to get the new client to run is
chmod a+rx boinc_3.20_powerpc_apple_darwin

You can use the 7400 version for any G4, and the 7450 version for any 7450 or greater G4. There is only a slight difference in the processors, and I'm not positive the specific compiles can fully take advantage of the difference.
The benchmarks vary too much on the same machine with the same client to be able to tell the difference between the clients.

[drweaser]

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

The Whetstone is at least double the normal optimized......

Thanks, hope it will help improve my credits.....

Do any of these programs actually take advantage of Altivec?

drweaser

[Shaktai]

Originally posted by drweaser:
Do any of these programs actually take advantage of Altivec?

drweaser

Not yet. BOINC is far too new for any type of optimized applications yet, outside of the core clients like mikkyo has done Too many bugs to still work out. However, once established, we might see some optimizations, if there is enough interest.

[falofolio]

Originally posted by Shaktai:
Not yet. BOINC is far too new for any type of optimized applications yet, outside of the core clients like mikkyo has done Too many bugs to still work out. However, once established, we might see some optimizations, if there is enough interest.

Hi,
I am certainly anxious to know what kind of improvement an AltiVec-optimization will give. So, you have my interest in this matter!

I am a new-comer here. So, I am a bit confused how to post my things here to make a start participating in this interesting topic. So, please, forgive me, when I am making here an attrupt interruption, but I would like everyone to look at these benchmark figures of my iBook below!

System configuration:

Macintosh type: iBook G4
Processor: ppc 7455 v3.3 (tech 0) with AltiVec
Clock speed: 1000 MHz
Bus Speed: 133 MHz
L1 Cache: 32 KB Inst, 32 KB Data
L2 Cache: 256 KB
RAM: 640 MB
OS Version: 10.3.4

Standard version 3.20:
2004-07-20 21:30:04 [---] Benchmark results:
2004-07-20 21:30:04 [---] Number of CPUs: 1
2004-07-20 21:30:04 [---] 474 double precision MIPS (Whetstone) per CPU
2004-07-20 21:30:04 [---] 1447 integer MIPS (Dhrystone) per CPU
2004-07-20 21:30:04 [---] Finished CPU benchmarks

MacNN Team's optimized version 3.20 for G4:
2004-07-20 21:26:08 [---] Benchmark results:
2004-07-20 21:26:08 [---] Number of CPUs: 1
2004-07-20 21:26:08 [---] 696 double precision MIPS (Whetstone) per CPU
2004-07-20 21:26:08 [---] 1457 integer MIPS (Dhrystone) per CPU
2004-07-20 21:26:08 [---] Finished CPU benchmarks

In percentages:
Whetstone increase of 47 %;
Dhrystone increase of 0.7 %.

Now my questions:
Will there be an super-optimized version of BOINC for ppc7455 in the future??? Or would that no significant improvement?

[mikkyo]

MacNN Team's optimized version 3.20 for G4:

Is that the Optimized or Super optimized version you are running?

Will there be an super-optimized version of BOINC for ppc7455 I have in the future??? Or would that no significant improvement?

There should be no significant difference in the 7450 version v a 7455 version.

[falofolio]

Originally posted by mikkyo:
Is that the Optimized or Super optimized version you are running?

There should be no significant difference in the 7450 version v a 7455 version.

I am running the Optimized version, because I didn't know that I was able to use the Super-Optimized version for 7450. But that's great! Will that give me another good boost?

[Lateralus]

Yes.

[falofolio]

Originally posted by PowerMacMan:
Yes.

Yes, indeed. Wow, incredible!!!
Look at these benchmark figures of my iBook G4 now, when using the Super-Optimized version 3.20 for G4, instead of the Optimzed version !?!

System configuration:
Macintosh type: iBook G4
Processor: ppc 7455 v3.3 (tech 0) with AltiVec
Clock speed: 1000 MHz
Bus Speed: 133 MHz
L1 Cache: 32 KB Inst, 32 KB Data
L2 Cache: 256 KB
RAM: 640 MB
OS Version: 10.3.4

Standard version 3.20:
2004-07-20 21:30:04 [---] Benchmark results:
2004-07-20 21:30:04 [---] Number of CPUs: 1
2004-07-20 21:30:04 [---] 474 double precision MIPS (Whetstone) per CPU
2004-07-20 21:30:04 [---] 1447 integer MIPS (Dhrystone) per CPU
2004-07-20 21:30:04 [---] Finished CPU benchmarks

MacNN Team's Super-Optimized version 3.20 for G4:
2004-07-20 22:50:35 [---] Benchmark results:
2004-07-20 22:50:35 [---] Number of CPUs: 1
2004-07-20 22:50:35 [---] 1409 double precision MIPS (Whetstone) per CPU
2004-07-20 22:50:35 [---] 1599 integer MIPS (Dhrystone) per CPU
2004-07-20 22:50:35 [---] Finished CPU benchmarks

In percentages:
Whetstone increase of 197 %;
Dhrystone increase of 11 %.

Now, another question:
May be a stupid one, but what will I gain from it when running SETI@home?

[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.