The Unofficial Apple Weblog (TUAW)

Not really. That's just called "optimisation", something that basically all software developers tend to neglect, basically by relying on to the ever faster hardware to balance the sloppy programming. Glad to see that Apple tries to reverse that trend.

↓↑report

It is pretty amazing, and certain things are definitely faster, no placebo effect. However, I've said this before, and I'll say it again; they're optimizing this for future macs. When the Quad Core iMacs come out, and the 12-core dual i7 Mac Pro's come out, it'll be much more apparent.

Also, when Software Devs start taking advantage of Grand Central, like hopefully Adobe CS5, Aperture 2.5 or 3 (dying to give them my money for this one), and so on, the Snow Leopard advantage will be much more compelling.

↓↑report

I'm replying to this, as it needs to be seen higher up in the comment thread. Not that most people will read past the headline, let alone the article.

This article is a fundamental misunderstanding of benchmarks, and is worse than worthless, it actively misinforms those who also don't understand a benchmark. Issue a clarification of this ASAP.

No OS is going to make your hardware magically faster (seriously, how do you think computers work - pixie dust?). However, the software included in the OS - the API's, frameworks, and programs like the Finder may be faster, or they may employ better algorithms and techniques, such as when 10.1 enabled prebinding for all applications, or Boot Caches added several versions ago, or the new Snow Leopard shutdown procedure.

↓↑report

as joshua said... im kind of shocked that you guys dont understand that a hardware bench test has little to nothing to do with the os running it.

↓↑report

As Joshua says, this benchmark is 99.9% about the underlying hardware, and has almost nothing at all to do with the OS.

A more useful benchmark would reflect real-world actions and most likely would show the significant speed improvements we've been hoping/expecting.

How about some Automator-based task? Perform some batch image resize/transform operations (in Preview to have a chance of exercising GC/OpenCL), then export to iMovie and encode to a slideshow maybe? Whatever the tasks, they should be Apple apps all the way, since nothing else is likely to take advantage of GC/OpenCL yet. Run the same action on Leopard v.s. Snow Leopard and the results are much more likely to be interesting.

↓↑report

Microsoft is great too. It gives you more code (larger OS) for the same price and help you to make faster decision on buying newer and faster system to support their slower OS.

↓↑report

i have a 1 1/2 year old mac pro and its faster than when brand new, now everything, not just my pro apps can use all of my 8 cores and 10gb, looking forward to seeing the devs use the new tech too.

↓↑report

Chris said 11:26AM on 9-02-2009
A more useful benchmark would reflect real-world actions and most likely would show the significant speed improvements we've been hoping/expecting.


Yeah, I'm finding the most noticeable increase in speed and responsiveness is not with a speed-bump to specific apps, as the benchmarking test. But where SL shines is when you've got many apps and windows running at once, competing for resources. Even though most 3rd part apps are not yet written for SL, having the finder, dock, window manager, application switcher, expose all able to take advantage of multiple cores means that even when one application gets slogged down, you can still expose/command-tab out of it and still find a snappy finder, mail, safari which seem unaffected (or at least less so) by Photoshop churning through a filter...
this was not true in Leopard... I'm not sure I've seen the ol' beach ball since i upgraded

↓↑report

@Joshua, Microdot, Chris

How do you guys then explain the increases in GeekBench scores?

↓↑report

dagaz: The OS is still the layer inbetween you (geekbench) and the hardware. There are still small details that can be optimized to squeeze that last bit out of that. For instance, if Leopard had to call two functions to copy X bytes memory around, and they managed to do it in one in Snow Leopard, that could be a slight increase in efficiency (no second function call, saves a few instructions). If such a thing happens often enough (say, thousands of times per second) you'll notice the slight difference.

Of course, the 'stream' bench is really purely a hardware test. There's not much OS between that anymore and what's there likely hasn't changed in a while. It basically entails reading, as fast as you can, from memory and discarding the result. So what you're measuring there is simply the bandwidth between memory and the CPU. This explains their similarity (the difference is, indeed, statistical deviation most likely caused by less memory activity on the higher side.)

↓↑report

Well, it would be rather paradoxical if they managed to make software run faster by adding more code. In general, less code = less time to execute.

↓↑report

yes cause 30 bucks is so expensive... it amazes me how much of the general mac popluation are cheap skates first of you complain about 2bucks for 8021.n enabler for the iMac then say 9.99 is to much for a firmware update and now 30 bucks is to expensive for a OS. give me a break.

↓↑report

@ Mike

I am one of those cheapskates you speak of, but not because I'm a complainer; but because I like to SEE the value. I didn't want to pay $2 for the 802.11n enabler because it wasn't worth it with my current setup and in all reality should have been enabled at the factory. And eventually, I got it for free when I bought a used Time Machine.
I upgraded my iPod Touch to 3.0 on a friends computer to save some cash; and I would have a right to be upset that a $10 charge is being applied because I don't have a monthly contract and a phone built-in. In all reality, I just didn't see the value in the other paid upgrade (2.1 I think).
The way Apple talked about Snow Leopard made me think that it wasn't worth $29. I wanted it because I love my new MBPro and I knew that eventual updates would make it faster and better; but at the keynote they made it sound like nothing more than update (with the exclusion of a rewritten finder). But luckily, being the cheapskate that I am, I was able to purchase the $10 up-to-date version.
See? It all works out in the end for a Mac user.

↓↑report

I see this Phoronix stuff turn up on almost every Mac webpage, but that does not make it any more meaningful: Phoronix *only* tested the OpenGL performance on two generations of Mac minis with integrated CPUs.

As long as you don't show me a single benchmark that confirms that OpenGL is slower on *all* types of Macs, no matter what GPU they use, *then* – and not before – you can say there has been a "regression" in OpenGL performance.

↓↑report