Panther Benchmark Shows 40% Quartz Speed Boost
This article over at Xlr8yourmac shows a 40% boost (34 vs 57s) in the 'very simple benchmark Let1k WindowsBloom (creates/closes 1000 windows)'. All this booting panther off an external firewire drive and jaguar off the internal drive. Surely this means quartz has seen some major quartz architectural optimizations, at least in the Panther dev. build. I am certainly impressed. Opinions?
Full Article Text
Full Article Text
Comments
And you doubters who thought it was just the application being enhanced.
Remember, that Let1KWindowsBloom test is not well optimized, itself...
I'm pretty optimistic about Panther's speed, given the increased interface 'snappiness' in iChat AV under Jaguar; if the increased speed with which the window resizes upon connecting is indicative of the sort of coded-in-speed ups that we'll be seeing in Panther... well... I'll be a happy camper.
Originally posted by Hobbes
Hmm, I wonder why Apple at WWDC demonstrated Quartz's improvements as an enhancement to Preview? It seems like it'd be more impressive to say Quartz had been optimized, and boast improvements across the entire OS... including Preview.
Because it was a *developer* conference, and they *get it*?
Focussing on Preview was for the press and public who might hear of it. The developers knew what it meant.
In any case *I* feel vindicated...
I'll be the first to tell you that I was really skeptical about the sometime contradictory performance claims made about Panther in different forums. But now that I have seen it with my own eyes, I am a believer. Yes, Panther really brings us back to OS 9 levels: buttons react immediately, menus react immediately, sheets slide in and out without delay.
Even iTunes's CPU consumption (a really sore point for me, since I am running it pretty much all the time) has decreased from 15-20% to 5-10%!
Great job Apple! Keep it up!!
Originally posted by Kickaha
Because it was a *developer* conference, and they *get it*?
Focussing on Preview was for the press and public who might hear of it. The developers knew what it meant.
In any case *I* feel vindicated...
Yep, it looks like you were right. I was a skeptic. Any guesses as to how they did it? Software optimizations in Quartz itself, or perhaps more use of GPU ala Quartz Extreme?
Originally posted by BRussell
Yep, it looks like you were right. I was a skeptic. Any guesses as to how they did it? Software optimizations in Quartz itself, or perhaps more use of GPU ala Quartz Extreme?
Probably just raw algorithmic optimization... it's amazing how much of a speedup you can get with some clever work. (FYI: optimized a meshing algorithm for doing salphasic clock simulations on circuit boards a few years ago and cut the time on an HP Unix workstation from 52 minutes to under 2 seconds.)
None of the PDF computation can be handed off to a 2D board, then back to Quartz for anti-aliasing, etc... this is almost *certainly* just good coding.
Originally posted by BRussell
Yep, it looks like you were right. I was a skeptic. Any guesses as to how they did it? Software optimizations in Quartz itself, or perhaps more use of GPU ala Quartz Extreme?
Since people on Rage128s are feel'n the love, I'd guess the former.
Originally posted by jamm
This article over at Xlr8yourmac shows a 40% boost (34 vs 57s) in the 'very simple benchmark Let1k WindowsBloom (creates/closes 1000 windows)'. All this booting panther off an external firewire drive and jaguar off the internal drive. Surely this means quartz has seen some major quartz architectural optimizations, at least in the Panther dev. build. I am certainly impressed. Opinions?
Full Article Text
I spoke with the author of this "benchmark" -- very nice guy, btw -- and he's sorry he ever wrote it, because it doesn't test what it was supposed to test at all (comparing OS 9 window opening/closing to OS X).
I suppose it is reasonable to use it to compare OS X versions to each other, but it doesn't compare what you think it does. It's more intensive on memory allocation than it is on the window server.
Under OS X, the window server and Quartz are two very different things.
Originally posted by cygsid
I am not surprised at all. I installed Panther on my iPod to try it out, and oh my God does it scream!
Can you navigate the dock with the scrollwheel now? Did they enable sub-pixel rendering on the B/W LCD?
Man, this is going to kick the crap out of XP on my Nokia 8290!
Originally posted by cygsid
I am not surprised at all. I installed Panther on my iPod to try it out, and oh my God does it scream!
I don't think that's the best idea. OS X will abuse that little poor hard drive
Except window resizing of course. It is faster than Jag though.
Preview is so fast it could not be made any faster.
Originally posted by lundy
It does indeed scream. Basic GUI tasks are about 300% faster than Jag by my estimate.
Except window resizing of course. It is faster than Jag though.
Preview is so fast it could not be made any faster.
What HW are you using?
Originally posted by BRussell
Software optimizations in Quartz itself, or perhaps more use of GPU ala Quartz Extreme?
It seems that it is the result of optimizations and good programming. Regarding Quartz and GPU, let me post here too this link.
Even better, this performance is acheived while one CPU is maxed out by the kernel for some reason. It switches back and forth, pegging either of the CPUs at 100%. Still, its quicker at almost everything (when its not hung on another process).
I can't wait.
Well, it sounds like a nice OS upgrade, but I can't believe that Apple is going to make people pay $130 for it. No way is an OS upgrade worth $130. Apple just keeps milking their customers with upgrades that don't really offer much functionality.
</sarcasm>
Way to go Apple! You are making it harder and harder for me to not go out and buy a mac... urge to switch growing... must wait for altivec ibook or G5 powerbook...
Originally posted by moki
I spoke with the author of this "benchmark" -- very nice guy, btw -- and he's sorry he ever wrote it, because it doesn't test what it was supposed to test at all (comparing OS 9 window opening/closing to OS X).
I suppose it is reasonable to use it to compare OS X versions to each other, but it doesn't compare what you think it does. It's more intensive on memory allocation than it is on the window server.
Under OS X, the window server and Quartz are two very different things.
Well then I'll retain some of my skepticism about system-wide Quartz speed-ups. We've just heard it too often - "OS X 4K78 is as fast as OS 9!" "Puma is faster than OS 9!" "Jaguar is..." Looking forward to seeing for myself, though.
What would be a good benchmark of system-wide Quartz performance? Bouncemarks?