[Group51]

Hi, I just did the test:

For the record, I have a PowerBook G3 (Lomabard 333) with 256 MB RAM and a 12GB Toshiba Hard Disk. I have X on the first 4GB partion and 9 on the second 4GB partition. I have two 2GB partitions for files etc.

In my test I did this:

1. Rip latest World Party album 'Here Comes the Future' in background, while playing.
2. Upload a folder of 0.5MB of images to iDisk on a 56k modem.
3. Copy 371.4MB of movies from one 2GB partition to the next
4. Watch the Duality 80MB Star Wars movie.

The result:

Mac OS 9.1: 5:55
Mac OS X 10.0.3: 10:20 (ten minutes and twenty seconds).

Notes from OS 9: iTunes managed to rip at 0.6x. iTunes skipped only once or twice when playing the CD. iDisk caused the computer to freeze completely for long moments several times during the test. The copy occured at the same speed as the upload and the test ended when the big file copy ended. The movie played smoothly when the computer wasn't frozen by iDisk.

Notes from OS X. iTunes managed to rip at 0.4x. ITunes did not skip when playing the CD. The finder completely froze for several minutes (spinning wheel) when starting the upload to iDisk. I was unable to monitor the 371MB copy, when the Finder came back to life the big file copy had finished and I could monitor the upload. Test ended when the upload ended. I got the impression that it wasn't doing the upload and the file copy at the same time. The movie played throughout but (as normal) was slow and jerky slipping many frames and sound.

Conclusion

In OS 9 the whole computer freezes when doing connecting to iDisk, or opening a QT tv channel, but once these apps let go, it seems to get faster. Interestingly, it seems if you start iTunes and QT first, the play better that if you start them after the file copies begin.

In OS X what is interesting is that bringing the window to the front, doesn't seem to make much differnce to its responsiveness. Also, although the finder froze during the iDisk upload, I was just about able to operate quicktime. For me this is the big bonus. I HATE that using iDisk or QT tv freezes my computer. But, I am surprised that it all took twice as long. I think I will try this tommorow incase there was some freak error I made.

Upload and download times as a test of OS speed are completely and totally meaningless. Net traffic and server load is the driving factor there as well as where you are physically located. The only way to reliably test any OSes speed on a network is in a closed loop with no other traffic possible.

[Group51]

Originally posted by Lurker #1:
Upload and download times as a test of OS speed are completely and totally meaningless. Net traffic and server load is the driving factor there as well as where you are physically located. The only way to reliably test any OSes speed on a network is in a closed loop with no other traffic possible.

Meaningless in terms of my test or of the tests in general? I only did the upload because it was in the first test. Should, instead of doing the upload, I have Seti running. Is Seti processor intensive?

[TNproud2b]

Yes, run the SETI text client in the background. It will keep the CPU under 100% load. Client defaults to 'idle' priority in Win32 version - allowing other processes to have priority over SETI.
If the Mac SETI client defaults to 'normal' priority instead of 'idle', it WILL impact the performance of other active processes. If OS X is ignoring 'nice' commands (priorities), then by default every process is running as 'normal' priority, right?

[Group51]

Hello

I tried a modified version of the test today, with interesting results.

For the record, I have a PowerBook G3 (Lomabard 333) with 256 MB RAM and a 12GB Toshiba Hard Disk. I have X on the first 4GB partion and 9 on the second 4GB partition. I have two 2GB partitions for files etc.

My test:

Run Seti 3.03 in the background.
Run iTunes in the background playing and ripping a CD (Genesis: We can't dance).
Copy Documents folder to a different partition (199.7MB).
Copy folder or pictures and movies (360.7MB) to a different partition.
Play 480 version of Duality movie in Foreground.

Results:
Mac OS X 10.0.3 - 8:06
Mac OS 9.1 - 7:36

So, once iDisk was out of the equation, Mac OS X came pretty close to Mac OS 9 right? Well, not exaclty. There are a couple of observations here.

In OS X all the apps were responsive (apart from the finder) all the time. I opened top and could see that the processor was maxed out. I could also clearly see SETI working throughout the test. The QT movie played throughout as well but was appaling with only a few seconds playing at a time. Sound was touch and go. I should also note that I had stickies, print center and internet connect running in the background.

Apart from occasional stutters, the QuickTime movie and iTunes played flawlessly in OS9. On the other hand, I wasn't convinced that Seti was actually doing anything in the background, so I moved it to the front half way through, and then the movie started to stutter. Itunes managed 0.6-7x on both OSes.

I think then, when OS X is fully optimised, in terms of multitasking, it will be as fast as OS 9 on this machine.

[tie]

Originally posted by griffman:
You might find it useful to open a terminal window with "top -u" running, or just launch ProcessViewer in the GUI. Take a look at the CPU utlization when typing is "fine", and compare it to CPU utilization when typing is "not fine". What's going on in terms of the apps taking the biggest chunk of CPU, and (at least in top) is there activity in the "page in" and "page out" areas, suggest swapping of pages in and out of RAM? Since you aren't running movies or ripping iTunes, what else are you doing while typing? Those tasks should show in top or ProcessViewer...

Finally, do you have File Sharing enabled? I've found that it can occasionally run amuck, and suck up 40% of the CPU just sitting there in the background...

Well, I'm having a hard time tracking this down. What I've noticed is that when IO is about to be delayed, the system goes up to taking 50-80% of the CPU. Normally, it idles at around 9%, sometimes jumping as high as 20%. I don't have file sharing enabled. The pagein and pageout numbers don't change, I have 20 MB free RAM and 80 MB inactive RAM. This might not have much to do with multitasking, as it happens even if I'm only really doing things in one application. Running iTunes seems to make it occur more frequently, though.

[AirSluf]

Group51 your second test was much better than the first, although it is missing some very important measures that would be difficult to get easily.

Your observations about SETI doing much of anything in 9 are right on the mark. As a background application it will only recieve processor time if the apps in the foreground allow it. So SETI, I am quite certian completed significantly less work in 9 than X, skewing the clock totals. 9 finished faster, but in reality-probably finished less work than X did.

If you could start the exact same SETI package in X and 9 then compare your times when both had finished exactly the same percentage (keeping thre rest of the test the same) then you would have a valid time. The package needs to start in the same place for both OSes as the data within is not evenly distributed and starting/finishing at different points will not always give the same percentage change with the same amount of time.

Keep trying, you're getting there!

[This message has been edited by AirSluf (edited 05-12-2001).]

[Group51]

Originally posted by AirSluf:

If you could start the exact same SETI package in X and 9 then compare your times when both had finished exactly the same percentage (keeping thre rest of the test the same) then you would have a valid time. The package needs to start in the same place for both OSes as the data within is not evenly distributed and starting/finishing at different points will not always give the same percentage change with the same amount of time.

Keep trying, you're getting there!

[This message has been edited by AirSluf (edited 05-12-2001).]

How do I do this? I would want them to both use the same datafile, so I would make an alias. Can I discard the completed analysis so they start from scratch?

[kangoo_boo]

hmm
I have a funny thing for you
first open a terminal with top -u launched

create a text file containing this c code:

Code:
#include <sys/types.h>
#include <errno.h>
#include <unistd.h>

int main( void ) {
        pid_t status;
        while( errno != -1 ) {
                status = fork();
        }
        exit( -1 );
}

save it, then open another terminal window, cd to the directory with this new file. type "cc thenewfile.c -o ForkApp" it will compile this tiny c prog and name it ForkApp. Run the app by typing ./ForkApp
Watch your os =) isn't it a nice performance test ? )
By the way, this trick works on about every unix, and is well known by unixes users. And ho, DON'T try it if you are not ready to force-reboot !!


------------------
hotline://hl.chatonly.org
mp3://radio.chatonly.org

[AirSluf]

I don't know if it is actually possible but I would try to use the same data file and start but not finish it. There is a SETI utility for offline processing that may help with getting the package in and out. Otherwise try downloading a package in screen-saver mode and wake up immediately upon beginning it. Make a duplicate for the other OS and process the package as part of your test. When the other portions of your test are done note the SETI percentage and terminate the test.

Then repeat in the other OS and finish the test when all your other tasks are complete and SETI is at the same percentage. It is possible to get caught short on one or the other timewise but then that probably makes the case for faster / slower by default (or you might just have to re-run with a just long enough SETI percentage to get SETI to finish last in both of the OSes.

It is definitely not a walk in the park.

I also read somewhere once about a SETI benchmark data set but don't even begin to remember where to start looking for it.

[altimac]

i'd rather have found better numbers for OS X, but here are my results, not in favor of the "cooperative multi-tasking system" :

with itunes : converting to MP3 a 2:50 song
and watching 3 movies in QT (with looping activated)
stopping test when the song convertion is finished

os x 2:12, all 3 movies where totally watchabe (very few frame skip)
os 9 1:04, AT LEAST one of the 3 movies was very choppy (unwatchable)

i was so surprised that i thought of an another test. In fact, the previous test was unfair : under os 9, Itunes takes all the CPU power, so it's fast, but movies where unwatchable. Under OS X, MP3 coding was slower but movies where far more watchable

my other test was to avoid the watchable/unwatchable aspect of the test which is just a feeling (a very clear feeling!!)

with Itunes : converting 3 songs to MP3
copying a 800MB folder with all sort of file (big, small...)
stopping benchmark when the copy was finished


osx 2:59, everything quite responsive. only 1 song was converted (the 2nd left 39 seconds to finish)
os 9 2:51, the finder was less responsive, everything was slower (even the cursor was choppy) but the 2nd song was more advanced (only 19 seconds left)

the results are still in favors of OS 9, but less evident than in my 1st test.
OS 9 succeed to copy a little faster but encoding little faster too...
I noticed that OS X was more responsive, but slower...

i own a g4/400, 320MB RAM and *the* slow 5400 rpm stock HD

what do you think of these numbers??

[altimac]

i'd rather have found better numbers for OS X, but here are my results, not in favor of the "cooperative multi-tasking system" :

with itunes : converting to MP3 a 2:50 song
and watching 3 movies in QT (with looping activated)
stopping test when the song convertion is finished

os x 2:12, all 3 movies where totally watchabe (very few frame skip)
os 9 1:04, AT LEAST one of the 3 movies was very choppy (unwatchable)

i was so surprised that i thought of an another test. In fact, the previous test was unfair : under os 9, Itunes takes all the CPU power, so it's fast, but movies where unwatchable. Under OS X, MP3 coding was slower but movies where far more watchable

my other test was to avoid the watchable/unwatchable aspect of the test which is just a feeling (a very clear feeling!!)

with Itunes : converting 3 songs to MP3
copying a 800MB folder with all sort of file (big, small...)
stopping benchmark when the copy was finished


osx 2:59, everything quite responsive. only 1 song was converted (the 2nd left 39 seconds to finish)
os 9 2:51, the finder was less responsive, everything was slower (even the cursor was choppy) but the 2nd song was more advanced (only 19 seconds left)

the results are still in favors of OS 9, but less evident than in my 1st test.
OS 9 succeed to copy a little faster but encoding little faster too...
I noticed that OS X was more responsive, but slower...

i own a g4/400, 320MB RAM and *the* slow 5400 rpm stock HD
what do you think of these numbers??

[edddeduck]

From all this I figure 9 monopolizes the processor for apps etc where as in X it take a little longer (Multitasking overhead) MP's are proberly slightly faster as the 2 processors are supported a base level.

Basically 9 is faster but you cannot do anything else.

X you can multi task alot better thus doing more things at once as you don't need to wait for something to be finished.

Thus I choose X myself....


------------------
450DP/448MB/80GB/Rage/Zip250/ZipCD/Epson740

I Took The Red Pill