Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Dronecatcher

macrumors 603
Original poster
Jun 17, 2014
5,209
7,795
Lincolnshire, UK
A recent thread discussed the accuracy of Geekbench and methods by which to establish the performance differences between machines.
With that in mind, I put together this test:

A 5 second 1080P MPEG4 video clip, converted in MPEG Streamclip to a 640X360 h.264 file.

All files are bundled together here (including instructions):

https://www.dropbox.com/s/dk9x1wys0m5z05r/Video Benchmark.zip?dl=0


I've used it on my machines and here are the results with Geekbench scores in brackets:


G5 Quad Leopard 7Gb 00:01:59 (3559)
Mac Pro 1,1 4X2.66 Ghz 00:02:06 (6120)
G5 DP 2.3 Ghz Leopard 4.5Gb 00:04:56 (1888)
G4 1.33 Ghz 12" Powerbook Tiger 768Mb 00:19:04 (722)
G4 1.33 Ghz 12" iBook Tiger 1Gb 00:20:52 (706)
G4 1.25 Ghz 15" Powerbook Leopard 1Gb 00:12:05 (587)
G3 500 Mhz iMac Tiger 512Mb 04:08:39 (301)

Most interesting is the 1.25 Powerbook, Geekbench significantly lower than the the other 'Books yet completes the task in nearly half the time.
And look at the G3 iMac - what a difference Altivec makes!

EDIT: Quicktime 7.5.5 seems optimal for this - 7.7 can lead to nearly double time taken.
 
Last edited:
  • Like
Reactions: AphoticD and z970

redheeler

macrumors G3
Oct 17, 2014
8,515
9,019
Colorado, USA
Wow! That's just crazy :eek:

If you try it on your G3 be aware it'll be cooking away at 100% for 4 hours+
I could smell the warm aroma of hot circuits coming from mine...

Maybe that's not such a good idea then since I can already predict how it will do anyway. I already benchmarked the GPU using Openmark and it came out at a mere 115, unsurprising considering it lags heavily on Tiger.
 

Beavix

macrumors 6502a
Dec 1, 2010
705
549
Romania
G5 DP 1.80 GHz Leopard: 00:05:51
G4 DP 1.25 GHz Leopard: 00:09:33
G4 1.67 GHz 17" PowerBook Leopard: 00:16:25
 
Last edited:

Dronecatcher

macrumors 603
Original poster
Jun 17, 2014
5,209
7,795
Lincolnshire, UK

Attachments

  • Picture 1.png
    Picture 1.png
    89.6 KB · Views: 210

catzilla

macrumors 6502
Dec 15, 2013
384
29
Rhode Island
rMBP 15" 2.3 i7 14 seconds.
Now to try it on my PPCs.


G4 Titanium 1GHz: 00:26:17
PM G5 Dual 2.0: 00:03:00
 
Last edited:

fuzzface

macrumors newbie
Apr 1, 2011
7
3
Hi folks,

here's my input:

G4 Titanium 1GHz (7455, 1MB L3) Tiger: 00:26:47
G4 MDD Dual 1.25 Tiger: 00.09:33
G5 Quad Leopard: 00:03:12

Cheers

Thomas
 
Last edited:

Cox Orange

macrumors 68000
Jan 1, 2010
1,814
241
A recent thread discussed the accuracy of Geekbench and methods by which to establish the performance differences between machines.
With that in mind, I put together this test:

A 5 second 1080P MPEG4 video clip, converted in MPEG Streamclip to a 640X360 h.264 file.

All files are bundled together here (including instructions):

https://dl.dropboxusercontent.com/u/8979107/Video Benchmark.zip


I've used it on my machines and here are the results with Geekbench scores in brackets:

G5 DP 2.3 Ghz Leopard 4.5Gb 00:04:56 (1888)
G4 1.33 Ghz 12" Powerbook Tiger 768Mb 00:19:04 (722)
G4 1.33 Ghz 12" iBook Tiger 1Gb 00:20:52 (706)
G4 1.25 Ghz 15" Powerbook Leopard 1Gb 00:12:05 (587)
G3 500 Mhz iMac Tiger 512Mb 04:08:39 (301)

Most interesting is the 1.25 Powerbook, Geekbench significantly lower than the the other 'Books yet completes the task in nearly half the time.
And look at the G3 iMac - what a difference Altivec makes!
Glad to see this thread, since I was argumenting against Geekbench in several threads (mostly SSD-HDD comparisons and GPU) and no one seemed to agree with me. ;-)

I don't have in mind which exact CPU the books have 7447 versus 7450? One would have to look it up at everymac.com I wonder, if Caches play a Role here.

In Handbrake with an ibook 12" 1,2GHz and a 23min. DV file to h.264, 2500kbps, normal preset, 2pass and audio 320kbps AAC it took 30min. longer than with the ibook 12" 1,33GHz. They were 12:30h and 12:00h.
That is 0,13GHz makes a difference of 30min. for long files... :)

What mpegstreamclip version did you use?
What h.264 codec? (If it was already there it is the one from iMovie. Then, which iMovie version do you have?) A long time ago, when I was searching for more recent x.264 codecs then the one I had from iMovie I did find these:
- x.264 x264 QuickTime Codec 1.1 long http://www.macupdate.com/app/mac/20273/x264-quicktime-codec
- x.264 different realisation, http://www003.upp.so-net.ne.jp/mycometg3/ (scroll down till Version for 10.4/10.5)

But I prefer Handbrake, because its codec is even newer. While 0.9.4 is the last PPC version (10.5 only), there is an unofficial one, too: 0.9.5 build
http://mark.nellemann.nu/2011/04/11/...r-osx-powerpc/
and
http://matthewcom.info/osx/
It's 11 passes - seems that's the default and I couldn't see an option to change it.
You can go for a single pass - less accurate but good for quick comparisons.
I think one can quit manually after the second run is about 100%, but you have to sit in front of it; and the file will work. But I quickly changed to Handbrake, faster and better quality.
G5 DP 1.80 GHz Leopard: 00:05:51
G4 DP 1.25 GHz Leopard: 00:09:33
G4 1.67 GHz 17" PowerBook Leopard: 00:16:25
Was the Powerbook on highest processor setting?
mpegstreamclip seems to be Dualcore aware, hence the lower clocked Dual is faster than the higher clocked single.

In my test back then, I had 3h iMovie and 1h Handbrake on the same PowerMac G5, file and settings (settings as similar as possible and with different x.264 codec). Although I once heard iMovie was SMP aware... hm.


MORE INPUT:
Would you be interested in compiling some other tests for us?

video transcoding is one thing, but machines might behave differently in other apps.
I suggest a Photoshop test. rendering something. But I am no photo guy, so I don't know, what test would be interesting.
On the LogicUser Forum, there is a testfile with multiple tracks and effects, that could be interesting for those having Logic Pro. There are also benchmark results of this test already.

To not use OpenMark, it would be interesting to find a test, to compare GPUs. For example with my PowerMac G4 1,2GHz 7455 and ATI 9200 (flashed AGP) I can play *mts2 files as stop-motion-picture show in VLC, on my ibook G4 1,33GHz (7450) with ATI 9550 I can play it almost smoothly in VLC. (Of course it depends on the file's specs). I could have been smarter and watch the fps in VLC...

PS: I also tested the ibooks with 512GB and 1,5GB RAM and it made no difference, this is actually interesting. RAM should play a bigger role, though with Photoshop I guess (and of course Logic)
 
Last edited:

Dronecatcher

macrumors 603
Original poster
Jun 17, 2014
5,209
7,795
Lincolnshire, UK
What mpegstreamclip version did you use?
What h.264 codec?

It's version 1.9.2 - as for the h.264 codec, it'll be whatever is supplied with Quicktime 7 (only my G5 has iMovie installed). I wanted the test to be basic and accessible, so the only extra requirement was Quicktime which everyone probably already has.
I guess slight variations in Quicktime version could make a slight difference?
My main objective was to highlight that Geekbench scores alone aren't really conclusive.

As for the merits of better transcoding, I've always found ffmpegx to have the edge for efficiency although Handbrake is easier to use.

The results from Intel machines are a real eye opener though - shows the gulf in technologies in such a short amount of time.

As for Photoshop tests, well I only have it on my G5 so personally, it would be in isolation and specific to the CS3 version.
 

Cox Orange

macrumors 68000
Jan 1, 2010
1,814
241
Ah, I didn't know it came with QT. I thought QT has just the h.264 playback codec, but not the encoding codec, hence I was assuming it is in iMovie.
 

flyrod

macrumors 6502
Jan 12, 2015
425
124
I tried this on my dual 1.8 G4 cube running tiger and got 8:33.

That progress bar is straight out of Office Space :) It does seem like 10.4 vs 10.5 makes a difference based on what others have posted. If I get around to it I'll boot 10.5 off an external firewire drive and try that.
 

MagicBoy

macrumors 68040
May 28, 2006
3,948
1,026
Manchester, UK
Late 2013 15" MacBook Pro 2.3GHz. 14.32 seconds. :D

Never goes above 210% CPU either, so it's using one and a fraction of the 4 cores.
 

Dronecatcher

macrumors 603
Original poster
Jun 17, 2014
5,209
7,795
Lincolnshire, UK
Late 2013 15" MacBook Pro 2.3GHz. 14.32 seconds. :D

Never goes above 210% CPU either, so it's using one and a fraction of the 4 cores.

How long does ripping/transcoding a 90 minute film take? I seem to remember my old C2D iMac was faster than my Quad but not massively so.
What is it that gives x64 this massive edge over PPC?
 

MagicBoy

macrumors 68040
May 28, 2006
3,948
1,026
Manchester, UK
How long does ripping/transcoding a 90 minute film take? I seem to remember my old C2D iMac was faster than my Quad but not massively so.
What is it that gives x64 this massive edge over PPC?

Not long! A handbrake transcode of 10 Mbps (ish) 1080p H.264 to a TiBook friendly MPEG4 720x300 of about 2.5Mbps flies through at about 200fps, so around 10 minutes.

If I run the same thing though my Windows gaming rig with a 3.4GHz i5 it's a few seconds quicker.
 

MagicBoy

macrumors 68040
May 28, 2006
3,948
1,026
Manchester, UK
Went looking for a 90 minute video and found a 1080P H.264 one of 15Mbps average and 10.5Gb total. Transcoding to the settings above with the source file on a server connected via Gigabit network was 12m 17s.

That's with the rMBP also running a Time Machine backup to a NAS box and serving iTunes out to another room at the same time. I wasn't about to kill a 400Gb initial backup partway through.

Screenshot%202015-06-13%2021.25.12.png
 
Last edited:

Dronecatcher

macrumors 603
Original poster
Jun 17, 2014
5,209
7,795
Lincolnshire, UK
Went looking for a 90 minute video and found a 1080P H.264 one of 15Mbps average and 10.5Gb total. Transcoding to the settings above with the source file on a server connected via Gigabit network was 12m 17s.

Thanks for getting for those figures. It's real world stats like that, that put things into perspective better than any point system benchmarks can.
I'm a long way off getting a new machine but when I do it's going to have to be back to Intel - those conversion times are unreal :)
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.