Performance in Tiger
The one thing that excited me more than anything else about getting and using Tiger was that everyone had lauded OS X for being the one OS that had improved in performance with every iteration. Being a Windows user, this was a complete reversal of my thinking - as each new iteration of Windows was an excuse for me to build a ridiculously powerful machine by my old standards, to feel even remotely at home under the latest version of Windows.But it would appear that I may have missed the boat, as performance in Tiger is a mixed bag. I'd say that overall performance in Tiger is an improvement over Panther, but there are some definite exceptions to the rule.
The biggest exception, from my perspective, is the fact that menu highlights always seem to just barely trail my mouse pointer in Tiger, whereas Panther kept up very well. At first, I thought that I was just imagining things, but then I set up a G5 test bed with two identical hard drives; the only difference being one drive had Panther while the other had Tiger. After booting them back to back, it's clear that my worries were founded. Tiger's menu highlight does seem to be slightly more laggy than Panther's.
Other areas of performance show definite improvements in Tiger, but what's interesting to me is that subjective performance impacts vary greatly (for me) depending on the type of system that I'm using. For this article, I used Tiger on four different modern day Macs: two Powermac G5 systems (dual 2.5GHz and dual 2.0GHz), one Mac mini and a 15" PowerBook G4.
Performance using Spotlight is pretty impressive on all three systems on which I've tried Tiger. On the mini and the PowerBook G4, it is definitely slower than the G5, thanks to their slower hard drives, but it is still fairly quick to search and bring up results. Spotlight on the G5 is lightning quick, although the results don't all pop up at the same time. Within a second, you have your entire list of search hits at the upper right hand corner of your screen. What's interesting here is that because Spotlight performance isn't something that I could quantify in Panther (since Spotlight obviously wasn't a feature of Panther), the PowerBook and mini feel slower in Tiger to me than they were in Panther. It's not because the OS is actually slower, but it's because my usage models have changed with Tiger - thus, putting more emphasis on fast hard disk performance. I'd also say that there's a pretty noticeable performance difference in Spotlight between the mini's 5400RPM 2.5" hard drive in comparison to my older PowerBook's 4200RPM 2.5" hard drive.
I/O performance appears to have improved by a noticeable amount in Tiger, especially on the G5. Disk accesses seem quicker, although I'm not certain if that's improvements to the disk side of I/O or if it is better at memory management, caching, or what. Regardless, those two areas of performance definitely improved.
There was a noticeable improvement in performance on the G5, with the biggest improvements being in UI performance. The UI seems a lot more responsive (especially at higher resolutions), scrolling is faster/smoother and the OS in general just feels a lot snappier. Scrolling performance throughout the OS has improved tremendously; it's not smooth enough to the point where I feel like I can enable the smooth scrolling option and get the same feel as I could on a PC, but performance is still definitely improved. My only complaint with UI performance continues to be the menu highlighting issue from earlier. It's not a show stopper, but it's definitely something noticeable.
Exposé performs basically identical to Panther under the new OS, regardless of what system I'm talking about. The Mac mini continues to have issues at higher resolutions, as Exposé stops being smooth and now even Dashboard is choppy on the mini. Again, the problem here is a sheer lack of video memory. I do hope that Apple updates the mini to include a better GPU as well as one with a larger local memory very soon, now that Tiger is out. So much of the beauty of the Dashboard is that you can access it so quickly and seamlessly, that when you get to use a choppy version of the feature, it does slightly ruin the effect of it.
Although Apple promises better battery life management on notebooks with Tiger, I didn't notice any major improvements during my multiple months with Tiger loaded on the PowerBook. Obviously, battery life improved over time, but I can't help but feel that even with the final version of Tiger, battery life is no better, and maybe even slightly worse (especially in sleep states) than Panther. I don't have any quantitative backing for these claims yet - it's just a feeling at this point.
The main thing to keep in mind is that all four of the systems I tried under Tiger performed, overall and at worst, no differently than under Panther. In many cases, there were some pretty hefty speedups that were definitely noticeable. I'd say the biggest performance gains that I noticed were on the G5 machines, despite spending more time with Tiger on the G4 based clients. If you have a G5 with enough video memory, Tiger and all of its new features should be smooth sailing for you.
I did perform some basic tests to see how Tiger stacked up to Panther in various performance categories. For these tests, I used a Powermac G5 dual 2.0GHz using a 250GB Maxtor MaXLine III SATA drive (each OS had their own drive with their own clean install of the OS).
Both Tiger and Panther took basically the same amount of time to start up, with Panther getting from pressing the power button to the desktop in about a second quicker than Tiger.
| Panther | Tiger | |
| System Startup in Seconds (Lower is Better) | 49.1 | 50.1 |
The Let1KWindowsBloom test times how long it takes to open 1000 windows:
| Panther | Tiger | |
| Window Creation Time in Seconds (Lower is Better) | 44 | 8 |
It's not the best test in the world, but it is interesting that there is an order of magnitude of performance improvement of Tiger over Panther. I'm not totally convinced that this isn't a bug with the test yet, however, so I wouldn't put too much faith in it just yet.
I wanted to see if Tiger improved the absolutely dismal Doom 3 performance of OS X:
| Panther | Tiger | |
| Doom 3 Frame Rate (Higher is Better) | 35.9 | 39.5 |
And although a 10% performance increase is nothing to scoff at, there's no getting around the fact that Doom 3 performance on the Mac is absolutely unacceptable. The average frame rates that I'm reporting here don't even begin to tell the full story. There's just far too much stuttering during the actual game for it to even be remotely playable. It looks like Tiger does improve gaming performance though.
For my final test, I ran a quick decompress test of a 140MB archive to see if I/O performance really improved in Tiger.
| Panther | Tiger | |
| Archive Extraction in Seconds (Lower is Better) | 22.93 | 21.5 |
Tiger boasted a 6% lower time, but what's interesting is that Tiger's time was much more reliable than Panther's. There were many occasions when Panther actually took significantly longer than Tiger, but the reverse was never true.
Benchmarking under OS X continues to be a pain, but there's an early try at quantifying the performance differences between Tiger and Panther. I can say, in confidence, that if you're a G5 user, you won't be disappointed. For G4 users, I can't say that the performance improvement was as drastic, but it definitely wasn't negative.
Facebook
Twitter
RSS
55 Comments
View All Comments
aliasfox - Monday, May 2, 2005 - link
#31What you're forgetting is that in 1995-6, Apple wasn't losing money, it was hemorraging money. In 1995-96, Apple was irrelevant in regards to innovation of any sort. In 1995-96, Apple was selling PowerBooks with batteries (made by Sony, no less) that caught fire. When Jobs returned to Apple in '97, Apple's stock price hovered in the low teens.
It doesn't matter all that much that Apple's marketshare hasn't grown as fast as the rest of the market (and yes, it is growing- Apple recorded most Mac sales in one quarter last month), in absolute numbers, it's not shrinking. And Apple is making money, which is more than Gateway (who has a bigger marketshare than Apple) can say.
The goal of any company is to make money, regardless of size. In this respect, Jobs has succeeded. It doesn't matter if it was the Reality Distortion Field or if it was amazing marketing or if it was quality products that got Apple there.
downtowncb - Monday, May 2, 2005 - link
#43 "The bottom line is that [recoding Mac OS software for x86] will not be a big issue for the majority of developers, as you had originally argued."My original argument (#34) said nothing about Apple switching to x86; you must mean #39 by melgross.
I simply meant to state that Apple hardware and software work together to increase Apple's bottom line. Sorry if that was unclear.
msva124 - Monday, May 2, 2005 - link
>If you want an example of an obscure motherboard, I believe that the TMC Ti5tv qualifies but perhaps not.How does the TMC Ti5tv require Microsoft OS developers to deviate from programming for the x86 spec as usual? You stated before that programming around obscure hardware causes development problems and bugs, and gave motherboards as an example. I will not waste my time contacting Microsoft, since they know as well as I do that your argument is an unproven hypothesis. A hypothesis that is contradicted by the stability of x86 Linux, and unsupported by any sort of scientific evidence or systematic testing.
>Programs for x86 that use hand written assembly code or are othersiwe highly optimized would need to be redone. But whether this is "most" programs or not I don't know. Certainly not every single program, that is true.
I doubt that even 20% of mac software uses any assembly language whatsoever. Within that 20%, it is typically just one or two heavily optimized loops per program, which would take little time to recode for the x86 platform. The bottom line is that it will not be a big issue for the majority of developers, as you had originally argued. In fact, that argument is so preposterous that I can only assume that your initial post and all subsequent ones were flamebait. I will not respond to any more of your posts until you can convince me otherwise.
downtowncb - Monday, May 2, 2005 - link
msva124,If you want an example of an obscure motherboard, I believe that the TMC Ti5tv qualifies but perhaps not. If you need other examples of obscure hardware Windows supports, contact Microsoft.
Programs for x86 that use hand written assembly code or are othersiwe highly optimized would need to be redone. But whether this is "most" programs or not I don't know. Certainly not every single program, that is true.
If Apple isn't taking away from the Windows market, then which market is it taking away from? If we're talking percentages, something's got to give. Not that Microsoft's growth in terms of numbers of copies of OS sold won't outpace any such loss to Apple or whoever. I'm curious as to what you think.
Jbog - Monday, May 2, 2005 - link
#39 melgross, you say Apple directly competes with MS on the OS front. You also say Apple has always had different hardware. Sounds to me it's more like Apple trying to come up with more appealing OS in order to sell its platform. I mean, you can't just buy Tiger OS and replace Windows XP.If any strong argument can be made, it would be between Windows and Linux instead. One can migrate from Windows to Linux without having to buy a whole new set of hardware. You can even dual-boot Windows and Linux.
msva124 - Monday, May 2, 2005 - link
>If Apple were to change to an x86, then every program would have to be redone. That would be almost impossible for the many Mac developers out there .Most OS X programs are written in C or Objective-C, using the Carbon or Cocoa apis. Only those apis must be ported to the x86 platform, not every single program.
>Apple's increase in marketshare takes away from the Windows market itself
Right, except it doesn't.
melgross - Monday, May 2, 2005 - link
#32 Most of what you said is ridiculous.First of all, Apple's market share is growing. Of course other platforms sales aren't Apples sales. But most other sales aren't either Apple/Sun, etc.
This is mostly an Apple/Wintel market. Sun is only servers and Apple doesn't compete much in their space yet. Apple's server sales are increasing, but are only now ramping up. Except in the scientific Unix space, Apples server sales would be against Windows servers. They don't yet have the breath to compete in the higher areas yet.
Otherwise, it's Apple vs. MS.
Sure, other pc companies, or rather company (Dell) are growing, but that takes sales away from each other. Apple's increase in marketshare takes away from the Windows market itself. If Dell takes sale from Gateway, it's still a sale for MS. That's the point.
I suppose that Apple is taking away a few Linux sale as well, but it's almost all MS's.
All AMD did was to finally come out with better processors that they could actually make, rather than just announce, and then NOT make. While the 64 bit extensions was a little balsy, it's true, it didn't take away from their chips either. Even if it didn't go over, the chips would still have had the same characteristics as before. The extensions would not have been used, that's all.
Apple does directly compete with MS on the OS front. Apple has always had different hardware. When Apple went to the 68000 rather than the 8088 way back when, there were few arguments that the 68000 was not a better chip. Apple simply went on through from there.
If Apple were to change to an x86, then every program would have to be redone. That would be almost impossible for the many Mac developers out there . It's just like the Itanium. Little software development has been done for it. Why should Apple be caught in that trap?
michael2k - Monday, May 2, 2005 - link
If you want to call it "RDF", that is your choice.I'm using Microsoft's publically announced information to compare Longhorn to Tiger.
Nothing I've mentioned is 'rumored'. It's all been 'confirmed' by Microsoft.
WaltC - Monday, May 2, 2005 - link
#33 "Um, you do know from 1985 until 1996 Steve Jobs had no presence at Apple?"I believe I said the better part of the last decade to begin with, as I recall...;) Yes, I know that he was originally fired/pushed out by the Apple board around '85 and went on to NeXt--which failed, btw.
#33 "Release a simple, affordable, powerful, computer in 1984: Original Mac, which became a strong model of the computing industry for the next 21 years."
Oh yes--I suppose that's why the board fired him in '85...;)
33# "Release a powerful, modern, OS and computer in 1989: NeXTStep, which is now the foundation for Mac OS X and is now Tiger, and is AGAIN a strong model for the computer industry (Longhorn, Linux)"
Both NeXt and NeXTstep failed commercially as I recall. The "foundation" for OS X was kind of forced into play, you know...;) And, it was years late and initially very lacking in promised features (many of which it still lacks.)
#33 "Create the world's most popular mp3 player, the iPod, in 2001: It's a computer in every sense of the word, with a display, input, storage, and output functionality. It's 'revolutionary' status is because it was the first, smallest, fastest, highest capacity (all at once) device, though there were smaller, with smaller capacities, or larger, with larger capacities, and none with faster upload or UI in 2001."
Sorry, but the iPod is *not* a personal computer. But that's RDF thinking for you without a doubt...;)
#33" Again, as for why compare Longhorn to Tiger?
Because everything Longhorn WANTS to do, Tiger does.
Longhorn wants a DBFS, called WinFS, not due until next year. Tiger achieves 90% of that now, and by next year will be even better.
Longhorn wants better search, to be achieved with WinFS, not due until next year, when Tiger has Spotlight now.
Longhorn wants a 3d accelerated display layer, and is not due until next year. OS X has achieved that since 10.2 in 2002 (a small step with hardware accelerated compositing), now more fully implemented since 10.3 and 10.4 with 3d and 2d acceleration, and with even more to come by the time Tiger comes out.
Longhorn wants a 'modern' UI, which is not due until next year, where OS X has had it since 2001, with each year bringing out more usability and functionality to the UI (Dock, transparency, animation, Expose, Dashboard, etc).
Longhorn wants better security, again not due until next year, while OS X has it now, and since 2001
Longhorn wants a shell and CLI, again next year, while OS X has had it since 2001
You ask why we compare: I think it's stupid NOT to compare. Longhorn wants to be a 'next generation' OS, and it's prototype and model 'next generation' OS is available now, and has been for four years, in Mac OS X.
We're not the only ones comparing. As I said before, Allchin of Microsoft has made direct comparisons, to Microsoft's detriment."
Ah, yes, the RDF again...;) The truth of course is that *nobody knows* what Longhorn will be since Longhorn is a long way out. I see nothing wrong with a Tiger-x64 comparison because MS is *shipping* x64. Pretty simple, really.
superduperjacob - Sunday, May 1, 2005 - link
Anand - on page 7 you say:"To remove a widget, you have to still be in the add/remove widget mode and just click the x that appears next to all of the widgets."
If you hold down the option key in normal widget mode, the x appears and you can close the widget.