10.5.6 benchmarks - summary, no improvement

[plinden]

Aware that some people have reported that Safari is snappier with 10.5.6, I ran a few benchmarks before and after installing 10.5.6 and saw no gains (maybe a small 1-2%, but that's probably just noise).

This is the sequence I performed:
From running 10.5.5 -> SunSpider -> clear Safari cache -> Sunspider
Result:

4736 to 4777, ie, no improvement

I cleared cache to get a measurement that would match the conditions of the one after the update.

Restart 10.5.5 -> Cinebench -> XBench -> SunSpider
Results:

Cinebench Rendering (Single   CPU): 1836 
Rendering (Multiple CPU): 3452 
Multiprocessor Speedup: 1.88
Shading (OpenGL Standard)          : 2783 

XBench 114

Sunspider 4441 - 8% improvement showing restart has a large effect.

Install 10.5.6 -> restart -> Cinebench -> XBench -> SunSpider

Cinebench Rendering (Single   CPU): 1840
Rendering (Multiple CPU): 3463 
Multiprocessor Speedup: 1.88
Shading (OpenGL Standard)          : 2737 

XBench 113

Sunspider 4431 - < 1%, insignificant effect.

People who see 8% improvement in Safari javascript are probably just reporting the effect of a restart.

I'm seeing what may be a slightly better Safari memory usage. By this stage of the day, Safari's usually using 400-500MB RAM. Currently it's at 330MB RAM, but I may not have hit my usual pattern of pages.

 

[Amdahl]

That's because the performance difference has nothing to do with JavaScript. My candidates are: cache changes (the update resets Safari cache), TCP (or DNS?) changes in OSX, or changes in how many connections Safari makes (and when) or pipelining.

Given that Safari doesn't have a major version change, it is more likely the changes are in the OS level. That indicates cache or TCP.

It would be great if someone out there could test and determine exactly what has happened.

 

[rdowns]

Aware that some people have reported that Safari is snappier with 10.5.6, I ran a few benchmarks before and after installing 10.5.6 and saw no gains (maybe a small 1-2%, but that's probably just noise).

You are aware that Safari is reported as snappier for every Apple software update.

 

[plinden]

You are aware that Safari is reported as snappier for every Apple software update.

Yes, and this case some people reported actual javascript improvements of about 10% using Sunspider and V8 benchmarks - in line with my results after restarting.

You know, some time, after an update, Safari actually is going to be snappier. Perhaps when Apple introduce some of the WebKit improvements (SunSpider is 4.5x faster in WebKit than Safari on my Mac)

I personally notice no difference in browsing after installing 10.5.6. The RAM usage has remained lower after most of a day of normal usage.

As for graphics, 10.5.2 showed a significant 30% improvement in Cinebench on integrated graphics cards. 10.5.6 does nothing.

Whatever Apple has concentrated on in this release, it isn't performance.

 

[Solemony]

Mine's fine...Maybe it's just me..?

 

[Amdahl]

I don't notice as much of a difference in Safari on G5. Perhaps the changes were to device drivers and are only noticeable on certain hardware platforms.

I notice a lot of new storage device drivers and AHCI (intel) disk controller drivers in the update.