June 2008 – Page 4 – Dan Dickinson: The Primary Vivid Weblog

Javascript Benchmarks: Firefox 3.0RC2 vs. Webkit r34367

The latest Firefox build, versus the latest Webkit build, using SunSpider. Testing done on my iMac (24″, 2.8 GHz Intel Core 2 Duo, 2 GB Ram) under normal conditions (other apps open, browser frontmost during the test).

COMPARISON         FIREFOX 3.0RC2       WEBKIT r34367
** TOTAL **:           1.56x as fast      2597.0ms +/- 2.1%    1666.8ms +/- 0.3%
=============================================================================
3d:                  1.44x as fast      327.4ms +/- 5.1%     227.2ms +/- 2.1%
cube:              1.62x as fast      120.2ms +/- 1.1%     74.0ms +/- 2.9%
morph:             1.43x as fast      106.2ms +/- 14.3%    74.2ms +/- 3.6%
raytrace:          1.28x as fast      101.0ms +/- 2.8%     79.0ms +/- 0.0%
access:              1.48x as fast      363.4ms +/- 2.0%     246.0ms +/- 0.9%
binary-trees:      1.54x as fast       44.0ms +/- 2.0%     28.6ms +/- 2.4%
fannkuch:          1.61x as fast      136.4ms +/- 1.4%     84.8ms +/- 2.2%
nbody:             1.28x as fast      137.2ms +/- 6.1%     107.4ms +/- 0.6%
nsieve:            1.82x as fast       45.8ms +/- 3.0%     25.2ms +/- 2.2%
bitops:              1.53x as fast      245.0ms +/- 2.6%     159.8ms +/- 1.0%
3bit-bits-in-byte: 1.62x as fast       39.2ms +/- 5.2%     24.2ms +/- 2.3%
bits-in-byte:      2.07x as fast       62.2ms +/- 0.9%     30.0ms +/- 0.0%
bitwise-and:       1.35x as fast       65.4ms +/- 1.7%     48.6ms +/- 3.9%
nsieve-bits:       1.37x as fast       78.2ms +/- 4.8%     57.0ms +/- 0.0%
controlflow:         1.75x as fast       30.8ms +/- 1.8%     17.6ms +/- 3.9%
recursive:         1.75x as fast       30.8ms +/- 1.8%     17.6ms +/- 3.9%
crypto:              1.41x as fast      155.8ms +/- 2.6%     110.8ms +/- 0.9%
aes:               1.43x as fast       61.4ms +/- 3.9%     42.8ms +/- 1.3%
md5:               1.40x as fast       46.8ms +/- 1.2%     33.4ms +/- 2.0%
sha1:              1.38x as fast       47.6ms +/- 3.5%     34.6ms +/- 3.2%
date:                2.15x as fast      321.6ms +/- 7.7%     149.6ms +/- 0.7%
format-tofte:      2.16x as fast      195.0ms +/- 13.2%    90.2ms +/- 0.6%
format-xparb:      2.13x as fast      126.6ms +/- 1.6%     59.4ms +/- 1.1%
math:                1.60x as fast      277.4ms +/- 4.4%     173.0ms +/- 0.7%
cordic:            1.81x as fast      104.8ms +/- 2.1%     57.8ms +/- 1.8%
partial-sums:      1.48x as fast      124.0ms +/- 8.7%     83.6ms +/- 0.8%
spectral-norm:     1.54x as fast       48.6ms +/- 1.4%     31.6ms +/- 2.2%
regexp:              1.35x as fast      225.4ms +/- 0.5%     166.8ms +/- 0.3%
dna:               1.35x as fast      225.4ms +/- 0.5%     166.8ms +/- 0.3%
string:              1.56x as fast      650.2ms +/- 0.5%     416.0ms +/- 0.5%
base64:            1.16x as fast       77.2ms +/- 1.8%     66.6ms +/- 1.0%
fasta:             2.30x as fast      173.8ms +/- 0.9%     75.6ms +/- 0.9%
tagcloud:          1.34x as fast      133.4ms +/- 0.5%     99.8ms +/- 1.0%
unpack-code:       1.67x as fast      173.4ms +/- 0.6%     103.8ms +/- 1.0%
validate-input:    1.32x as fast       92.4ms +/- 0.7%     70.2ms +/- 1.5%

But don’t worry, Firefox! You may be 56% slower on overall Javascript performance, but you’ve been promised to be bug free!

The Alleged End Of OS X for PPC

History lesson: Back on June 6, 2005, Apple announced that they were transitioning away from the PowerPC processor line to ones made by Intel. Rumblings formed quickly – *how long until they drop PowerPC support from the OS?*

TUAW wrote yesterday about “10.6” being unveiled at WWDC next week:

> We have also learned that OS X 10.6 may go gold master by December 2008 in an effort to start shipping it in January ’09 at Macworld Expo. Mac OS X 10.6 will be a milestone release for Apple, as it will leave the PowerPC behind: a fully 64-bit clean, Intel-only Mac OS X.

John Gruber weighed in this morning:

> I still think it seems too soon by at least a year to drop PowerPC support — especially for G5s, which are still extremely capable machines by today’s standards — but that’s the word on the street.

Gruber’s on the right track here, but not because the machines are still capable (which they are).

Consider two truths about Applecare:

* Applecare typically comes in two quantities: the free one-year that comes with each machine, and a three-year extended service warranty.
* Applecare provides support for all point revisions of the current operating system and the final point release of the previous operating system.

As far as I’m aware, Apple has never released an OS that cannot by installed on computers that fall within the three-year window provided by Applecare. Once a machine falls outside of that window, a machine is not guaranteed to run any new major OS revisions.
The PowerMac G5 was the final machine to be killed in the Intel transition – a slight irony given that the developer kits for the Intel transition were in PowerMac towers. The MacPro was introduced to market on August 7, 2006.

For these reasons, I would not expect Apple to kill off PowerPC support in their OS until after August 6th, 2009.

(An aside: that’s also the day that Classic finally becomes unsupported, as there won’t be any Applecare eligible machines that can still run Classic.)