Well that's an overly simplistic way of looking at it. Apple's claim of around 200% performance is strictly from a SPECint and SPECfp perspective. Those benchmarks are synthetic, but reflect general purpose integer and floating point performance. These are areas where the G4 and the G5 have trailed the Intel based processors a little bit lately.
Bear in mind that SPEC is optimized for multiprocessors, so automatically the dual-core nature of the Core Duo gives it a leg up over the single G5 in the previous iMac.
Where the G4 and the G5 did well, and why Photoshop was still very much viable on the PowerPC was with vector performance. Specifically, with Altivec instructions. Altivec was much more elegant and much faster than Intel's equivalent, SSE, and I do not expect Rosetta emulation of Altivec to be nearly as fast as real Altivec hardware.
So what does that necessarily mean? For something like application startup, I would expect it to be a wash. App launching doesn't use altivec, so you won't see much of a slowdown there. where you will see a significant slowdown is when you start to apply altivec optimized filters in Photoshop.
SPEC is not a synthetic benchmark. Among other things, it includes some ray tracing and a round of GNUChess.
I went back to Macworld today and performed more scientific tests with the new MacBook. I'll post more details later tonight but I'll post my revised iTunes times now.
I brought the same track I used with my G5 to MacWorld and re-encoded it using iTunes. It was read off the HD from the iTunes library and all other apps were closed.
Today the MacBook ripped at 20x which is far better than I expected.
I'm not sure what impacted the previous test and it might have been related to the source file, but this time it was fairly controlled and I'd stand by these results. Again for comparison, an existing G4 PowerBook ripped at 12.5x and my Rev A G5 tower ripped at 25x.
I'll post more tonight regarding Video transcoding performance using QuickTime Player and H264 iPod content.
I want numbers of ripping speeds from CDs not from the HD.
Meh. It looks like my iMac has a bad DVD drive. It can read CDs fine, but it can't mount DVDs. Tried both of the supplied system discs plus a movie DVD, and none mount (all work fine in a powerbook). I'll probably have to take it in. I'll just have them swap out the machine if I do.
Only the SPECint_rate2000 is a test of multiple processors. The standard SPECint2000 is not.
Oddly, Apple quoted the speed increase with regard to running SPEC*_rate2000 on the previous generation over the new one. That explains some of the dramatic figure as running the multithreaded versions on a single core CPU will be slower than running the normal tests. As usual with Apple's benchmarks, there's a slight of hand trick going on.
On the last day of Macworld I revisited the prototype MacBooks and performed a series performance tests. I conducted four tests with the MacBook and previous G4/G5 systems, including encoding audio and video and playback of HD video. The goal was to evaluate MacBook Pro system performance and compare the new system to previous Macs.
Test Systems
Four systems were tested to provide results for a broad range of Macs, including various generations of Mac laptops and desktops. Each system was running off its AC power adapter and used the "Better Performance" power management setting. Only the primary display was active and it was set to its native resolution. Exact specifications for each test system and their introduction dates are as follows:
MacBook Pro Core Duo 1.83 GHz, 2 GB DDR2, ATI X1600 Mobility 128 MB @ 1440x900 - Jan. 2006 Preproduction
PowerMac G5 Dual 2 GHz, 2.5 GB DDR, ATI 9800 Pro 128 MB @ 1680x1050 - Jul. 2003
PowerBook G4 Titanium 500 MHz, 768 MB SDRAM, ATI Rage 128 Mobile 8 MB @ 1152x768 - Jan. 2001
Methods
To best judge the performance of the new MacBooks I conducted four media-related tests on each system: audio encoding using iTunes, video encoding using QuickTime Player, and HD video playback using QuickTime Player for 720p and 1080p content. These tests stress various aspects of each system, including the CPU and graphics cards. All test files resided on the machine?s internal hard disk and other applications were closed unless otherwise noted. Each test system had the latest versions of iTunes, QuickTime, and MacOS X installed, versions 6.0.3, 7.0.4, and 10.4.4 respectively.
The iTunes encoding test converted audio using the "Convert Selection to AAC..." feature and reports the encoding speed as displayed in iTunes 6.0.2. The source file was a 16 minute, unprotected AAC file of "It's A Fast Driving Rave Up With The Dandy Warhols Sixteen Minutes" by The Dandy Warhols. The source file is 160 kbps and was previously ripped from the CD "Dandy's Rule OK?" using iTunes 4.0.1 and QuickTime 6.3. The output format is iTunes? AAC default settings at 128 kbps.
The QuickTime encoding test converted video to iPod format using the "Export..." feature in QuickTime Player 7.0.4. A stopwatch was used to measure the time taken to encode each video. Then, the video?s length was divided by this encoding time to produce the encoding speed. For example, if a 60 second video clip was encoded in 30 seconds the encoding speed is 2x because 60 divided by 30 is 2. The source file is Apple?s latest Intel iMac ad and was downloaded from the Apple.com website. The clip duration is 35.66 seconds at 640x360 resolution playing at 23.98 fps and is encoded in H.264 format with AAC audio. The filename is "intel20060110_h.640.mov". The output format is QuickTime player?s "Movie to iPod (320x240)" default settings which uses H.264 and AAC codecs.
The QuickTime playback tests played HD video in QuickTime Player at both actual and fullscreen window sizes. Both 720p and 1080p versions of the King Kong movie trailer were used. These files were downloaded from Apple?s QuickTime Movie Trailer website. The files were loaded from the internal Hard Drive and no downloading or buffering was needed during testing. QuickTime?s "Movie Info" window was open to record the Playing FPS and an average value was recorded for the clip. If the clip played seamlessly the Activity Monitor was used to monitor QuickTime?s CPU usage via the "% CPU" column.
Results
As you can see, the new MacBook Pro encoded audio 33% faster than the previous PowerBook, at 20x compared to 15x. Unfortunately, the MacBook Pro is still 20% slower than the original PowerMac G5, which encoded audio at 25x, and more powerful G5 quad systems are now available.
The new MacBook Pro proved to be far more adept at encoding video, at rate of 1.3x. The MacBook Pro is 160% faster than the previous PowerBook, which encoded video at 0.5x. The MacBook Pro encodes video only 24% slower than the G5 system, which encoded video at 1.7x.
When playing fullscreen 1080p video the MacBook Pro never dropped a frame, unlike the PowerBook G4 which played at 8 fps. The PowerMac G5 system also maintained the full framerate of 23.98 fps but used more system resources. The MacBook Pro used 80% of total CPU resources whereas the PowerMac used 110%. (Since these are both dual processor systems, the total CPU resources for these systems tops at 200%, 100% for each processor.) The MacBook Pro is playing the video back with far less effort than the PowerMac G5 and may not even need the second processor core, unlike the PowerMac which is leveraging both processors. The Titanium PowerBook G4 never had a chance.
When playing fullscreen 720p video most systems performed well. Neither the MacBook Pro nor the PowerMac G5 dropped frames. While the PowerBook G4?s playback was consistent, it did occasionally drop a frame. Like the 1080p tests, the MacBook Pro bested the other systems, using the least system resources to display the video.
Analysis and Conclusion
The new MacBook Pro performed admirably in all tests and surprisingly well in the video playback tests. While the Core Duo processor used in the MacBooks is very different than previous G4 and G5 processors, that did not seem to hinder it. In addition to dual processors, the MacBook features a fast ATI X1600 graphics chip and both of these are responsible for the new system?s improved performance.
While these tests paint a glamorous picture of the MacBook Pro, this analysis could be improved in many ways. Firstly it would be beneficial to include iMac G5 and iMac Core Duo systems as well as cutting edge PowerMac G5s and PCs. This would round out the analysis and provide deeper comparisons. Repeated trials for existing tests and additional tests would improve accuracy. A more precise and less subjective methodology for testing video framerates would also be preferred. The results of this analysis also lead to new questions, such as would an upgraded graphics card improve the PowerMac G5?s performance?
The overall results of this analysis indicate that Apple?s new MacBook Pro is an improved machine in many ways. While some areas are yet to be examined, it is safe to say that the new MacBook is Apple?s fastest laptop yet.
I suppose that answers the question about h.264 decode assistance in the drivers for the X1600. Any chance you can do an encoding test with a different codec? I want to see if the MBP has encoding assisted by the graphics card too.
I picked up a mini-DVI cable today and hooked up my Gateway 21" widescreen display. This image (crappy cellphone cam -- sorry) is showing the iMac doing spanning, with both displays set to 1680 * 1050 (native for each).
Hi. I want to do same thing with my Apple Cinema display.
Do you notice worst performance (video performance) with the two displays active? Like expose and dashboard? is it slowed down?
What about watching a 1080p movie? Does it hurt its performance at all with two displays splitting the VRAM? Enjoy your setup! looks great.
Expose and Dashboard are silky smooth so far across both monitors
I haven't played any HD videos since I hooked up the second monitor. I'll post if I find anything that runs me out of video ram and makes things choppy.
On the last day of Macworld I revisited the prototype MacBooks and performed a series performance tests. I conducted four tests with the MacBook and previous G4/G5 systems, including encoding audio and video and playback of HD video. The goal was to evaluate MacBook Pro system performance and compare the new system to previous Macs.
Hi Xool. Thanks for all your effort writing this up. Very thorough.
Quote:
Originally posted by Xool
Test Systems
...
PowerBook G4 Titanium 500 MHz, 768 MB SDRAM, ATI Rage 128 Mobile 8 MB @ 1152x768 - Jan. 2001
My system specs exactly. Looks like I'll get a nice performance boost when I eventually upgrade.
You're welcome everyone! I posted the final MacBook Performance Analsysis on my site and even submitted it to Digg. If you're a digger help the story make the front page!
I just converted a 1-hour MP3 file to AAC and it was running at a constant 18.0x on my iMac 17" Duo Core (only 512MB RAM). This was happening while I was compiling some stuff, and with a few other apps open too.
Someone asked about video performance over dual monitors. I have the 17" with a 21" CRT running at 1600x1200 attached through a Mini-DVI adapter. I have yet to see any drag/dropped frames/etc. when doing exposé/dashboard. I used to have the same monitor plugged into my PB 17" 1.5GHz w/ 1GB RAM, and you definitely noticed dropped frames when you did exposé/dashboard on it. So definitely better video performance.
Comments
Originally posted by laughingman
Well that's an overly simplistic way of looking at it. Apple's claim of around 200% performance is strictly from a SPECint and SPECfp perspective. Those benchmarks are synthetic, but reflect general purpose integer and floating point performance. These are areas where the G4 and the G5 have trailed the Intel based processors a little bit lately.
Bear in mind that SPEC is optimized for multiprocessors, so automatically the dual-core nature of the Core Duo gives it a leg up over the single G5 in the previous iMac.
Where the G4 and the G5 did well, and why Photoshop was still very much viable on the PowerPC was with vector performance. Specifically, with Altivec instructions. Altivec was much more elegant and much faster than Intel's equivalent, SSE, and I do not expect Rosetta emulation of Altivec to be nearly as fast as real Altivec hardware.
So what does that necessarily mean? For something like application startup, I would expect it to be a wash. App launching doesn't use altivec, so you won't see much of a slowdown there. where you will see a significant slowdown is when you start to apply altivec optimized filters in Photoshop.
SPEC is not a synthetic benchmark. Among other things, it includes some ray tracing and a round of GNUChess.
Originally posted by Xool
I went back to Macworld today and performed more scientific tests with the new MacBook. I'll post more details later tonight but I'll post my revised iTunes times now.
I brought the same track I used with my G5 to MacWorld and re-encoded it using iTunes. It was read off the HD from the iTunes library and all other apps were closed.
Today the MacBook ripped at 20x which is far better than I expected.
I'm not sure what impacted the previous test and it might have been related to the source file, but this time it was fairly controlled and I'd stand by these results. Again for comparison, an existing G4 PowerBook ripped at 12.5x and my Rev A G5 tower ripped at 25x.
I'll post more tonight regarding Video transcoding performance using QuickTime Player and H264 iPod content.
I want numbers of ripping speeds from CDs not from the HD.
Bear in mind that SPEC is optimized for multiprocessors,
Only the SPECint_rate2000 is a test of multiple processors. The standard SPECint2000 is not.
I'm very curious
Originally posted by Flounder
Xool, when are you going to have a chance to post your results?
I'm very curious
Originally posted by lundy
Only the SPECint_rate2000 is a test of multiple processors. The standard SPECint2000 is not.
Oddly, Apple quoted the speed increase with regard to running SPEC*_rate2000 on the previous generation over the new one. That explains some of the dramatic figure as running the multithreaded versions on a single core CPU will be slower than running the normal tests. As usual with Apple's benchmarks, there's a slight of hand trick going on.
-----------
MacBook Pro Performance Analysis
Draft Report - My final analysis is on craigtheguru.com.
Introduction
On the last day of Macworld I revisited the prototype MacBooks and performed a series performance tests. I conducted four tests with the MacBook and previous G4/G5 systems, including encoding audio and video and playback of HD video. The goal was to evaluate MacBook Pro system performance and compare the new system to previous Macs.
Test Systems
Four systems were tested to provide results for a broad range of Macs, including various generations of Mac laptops and desktops. Each system was running off its AC power adapter and used the "Better Performance" power management setting. Only the primary display was active and it was set to its native resolution. Exact specifications for each test system and their introduction dates are as follows:
- MacBook Pro Core Duo 1.83 GHz, 2 GB DDR2, ATI X1600 Mobility 128 MB @ 1440x900 - Jan. 2006 Preproduction
- PowerBook G4 15" 1.67 GHz, 1.5 GB DDR, ATI 9700 Mobility 64 MB @ 1280x854 - Jan. 2005
- PowerMac G5 Dual 2 GHz, 2.5 GB DDR, ATI 9800 Pro 128 MB @ 1680x1050 - Jul. 2003
- PowerBook G4 Titanium 500 MHz, 768 MB SDRAM, ATI Rage 128 Mobile 8 MB @ 1152x768 - Jan. 2001
MethodsTo best judge the performance of the new MacBooks I conducted four media-related tests on each system: audio encoding using iTunes, video encoding using QuickTime Player, and HD video playback using QuickTime Player for 720p and 1080p content. These tests stress various aspects of each system, including the CPU and graphics cards. All test files resided on the machine?s internal hard disk and other applications were closed unless otherwise noted. Each test system had the latest versions of iTunes, QuickTime, and MacOS X installed, versions 6.0.3, 7.0.4, and 10.4.4 respectively.
The iTunes encoding test converted audio using the "Convert Selection to AAC..." feature and reports the encoding speed as displayed in iTunes 6.0.2. The source file was a 16 minute, unprotected AAC file of "It's A Fast Driving Rave Up With The Dandy Warhols Sixteen Minutes" by The Dandy Warhols. The source file is 160 kbps and was previously ripped from the CD "Dandy's Rule OK?" using iTunes 4.0.1 and QuickTime 6.3. The output format is iTunes? AAC default settings at 128 kbps.
The QuickTime encoding test converted video to iPod format using the "Export..." feature in QuickTime Player 7.0.4. A stopwatch was used to measure the time taken to encode each video. Then, the video?s length was divided by this encoding time to produce the encoding speed. For example, if a 60 second video clip was encoded in 30 seconds the encoding speed is 2x because 60 divided by 30 is 2. The source file is Apple?s latest Intel iMac ad and was downloaded from the Apple.com website. The clip duration is 35.66 seconds at 640x360 resolution playing at 23.98 fps and is encoded in H.264 format with AAC audio. The filename is "intel20060110_h.640.mov". The output format is QuickTime player?s "Movie to iPod (320x240)" default settings which uses H.264 and AAC codecs.
The QuickTime playback tests played HD video in QuickTime Player at both actual and fullscreen window sizes. Both 720p and 1080p versions of the King Kong movie trailer were used. These files were downloaded from Apple?s QuickTime Movie Trailer website. The files were loaded from the internal Hard Drive and no downloading or buffering was needed during testing. QuickTime?s "Movie Info" window was open to record the Playing FPS and an average value was recorded for the clip. If the clip played seamlessly the Activity Monitor was used to monitor QuickTime?s CPU usage via the "% CPU" column.
Results
As you can see, the new MacBook Pro encoded audio 33% faster than the previous PowerBook, at 20x compared to 15x. Unfortunately, the MacBook Pro is still 20% slower than the original PowerMac G5, which encoded audio at 25x, and more powerful G5 quad systems are now available.
The new MacBook Pro proved to be far more adept at encoding video, at rate of 1.3x. The MacBook Pro is 160% faster than the previous PowerBook, which encoded video at 0.5x. The MacBook Pro encodes video only 24% slower than the G5 system, which encoded video at 1.7x.
When playing fullscreen 1080p video the MacBook Pro never dropped a frame, unlike the PowerBook G4 which played at 8 fps. The PowerMac G5 system also maintained the full framerate of 23.98 fps but used more system resources. The MacBook Pro used 80% of total CPU resources whereas the PowerMac used 110%. (Since these are both dual processor systems, the total CPU resources for these systems tops at 200%, 100% for each processor.) The MacBook Pro is playing the video back with far less effort than the PowerMac G5 and may not even need the second processor core, unlike the PowerMac which is leveraging both processors. The Titanium PowerBook G4 never had a chance.
When playing fullscreen 720p video most systems performed well. Neither the MacBook Pro nor the PowerMac G5 dropped frames. While the PowerBook G4?s playback was consistent, it did occasionally drop a frame. Like the 1080p tests, the MacBook Pro bested the other systems, using the least system resources to display the video.
Analysis and Conclusion
The new MacBook Pro performed admirably in all tests and surprisingly well in the video playback tests. While the Core Duo processor used in the MacBooks is very different than previous G4 and G5 processors, that did not seem to hinder it. In addition to dual processors, the MacBook features a fast ATI X1600 graphics chip and both of these are responsible for the new system?s improved performance.
While these tests paint a glamorous picture of the MacBook Pro, this analysis could be improved in many ways. Firstly it would be beneficial to include iMac G5 and iMac Core Duo systems as well as cutting edge PowerMac G5s and PCs. This would round out the analysis and provide deeper comparisons. Repeated trials for existing tests and additional tests would improve accuracy. A more precise and less subjective methodology for testing video framerates would also be preferred. The results of this analysis also lead to new questions, such as would an upgraded graphics card improve the PowerMac G5?s performance?
The overall results of this analysis indicate that Apple?s new MacBook Pro is an improved machine in many ways. While some areas are yet to be examined, it is safe to say that the new MacBook is Apple?s fastest laptop yet.
-----
Edit: Updated images and added link to the final analysis on craigtheguru.com.
Huge boost over the old Powerbooks anyway.
Originally posted by ciparis
I picked up a mini-DVI cable today and hooked up my Gateway 21" widescreen display. This image (crappy cellphone cam -- sorry) is showing the iMac doing spanning, with both displays set to 1680 * 1050 (native for each).
Hi. I want to do same thing with my Apple Cinema display.
Do you notice worst performance (video performance) with the two displays active? Like expose and dashboard? is it slowed down?
What about watching a 1080p movie? Does it hurt its performance at all with two displays splitting the VRAM? Enjoy your setup! looks great.
Thanks
I haven't played any HD videos since I hooked up the second monitor. I'll post if I find anything that runs me out of video ram and makes things choppy.
Originally posted by Xool
Introduction
On the last day of Macworld I revisited the prototype MacBooks and performed a series performance tests. I conducted four tests with the MacBook and previous G4/G5 systems, including encoding audio and video and playback of HD video. The goal was to evaluate MacBook Pro system performance and compare the new system to previous Macs.
Hi Xool. Thanks for all your effort writing this up. Very thorough.
Originally posted by Xool
Test Systems
...
PowerBook G4 Titanium 500 MHz, 768 MB SDRAM, ATI Rage 128 Mobile 8 MB @ 1152x768 - Jan. 2001
My system specs exactly. Looks like I'll get a nice performance boost when I eventually upgrade.
Originally posted by Xool
MacBook Pro Performance Analysis
Awesome Xool! Thank you so much for the details.
Thank you Xool.
And THANK YOU INTEL
Anders: Thanks Anders for Thanking Intel.
Someone asked about video performance over dual monitors. I have the 17" with a 21" CRT running at 1600x1200 attached through a Mini-DVI adapter. I have yet to see any drag/dropped frames/etc. when doing exposé/dashboard. I used to have the same monitor plugged into my PB 17" 1.5GHz w/ 1GB RAM, and you definitely noticed dropped frames when you did exposé/dashboard on it. So definitely better video performance.
-Percy
Can anyone try if PGP Desktop and Glowl are working with MacIntel Rosetta? I'd appreciate a lot.