Friday Apple Quickies

An anonymous reader writes "Steve Jobs' $78 million Apple income tops Fortune magazine's list of CEOs whose companies lagged behind the S&P 500 performance last year. The number 'reflects the value of five million restricted shares Jobs got this year in exchange for 27.5 million underwater options.'" markomarko writes "Well, despite Charlie White making all us Mac users eat crow over his comparison of render times between a dual 1.25 GHz Power Mac and a Dell 3.06 GHz P4, it seems that that Dave Nagel has given us a reason to take another look at the Mac. His article shows how After Effects render speeds can be doubled with the Mac, by using both CPUs."

Comments

Re:After effects

[cyb97]

I wonder if you know what After Effects does sir, as there are different uses for Final Cut *, Shake and After Effects.
After Effects is actually what the name implies a program to add Effects after it's been edited (or even while). Final Cut is a linear editor, while After Effects is a visual effects program. The do wastly different things.
Shake on the other hand is not unsimilar to AE, but the price tag on it ($5,000) just prices it out of the competition. That's also why they marked it as a digital composition tool, aimed at filmproducers rather than the average joe just wanting to put some flashy-text in his average home-video.

Re:After effects

[seinman]

Cut is a linear editor

Great post, you're right about the difference between FCP and AE. However, FCP is a non-linear editor.

So in other words....

[Squidgee]

Adobe is technically screwing over the Mac.

By releasing a product which uses 1/2 of the Mac's power, they've crippled it. It would be similar to capping that 3ghz pentium to 1.5ghz, in which cause the Mac would (Guess what) slaughter the PC.

Yea, Adobe's a real good Mac software maker.

Re:So in other words....

[arloguthrie]

Oh, right, it's Adobe's fault. *rolls eyes*

Damn right its Adobe's fault. It's Microsoft's fault that Windows sucks, is it not? It's not Intel's fault that Microsoft is not writing an operating system that runs smoothly and efficiently on their processors and chipsets. So if Adobe isn't willing to optimize their software to take advantage of Altivec and dual-processors on the platform that made their company what it is today, then I say, "Adobe -- j'accuse!"

I have many of the same feelings about Adobe as I do Microsoft. Adobe had a nearly monopolistic place in the design industry (Photoshop, especially) and they abuse it by filling their software with bloat and not actually improving it. They have a history of making their software incompatible with others (see this article). And let's not forget when they sued Macromedia for using a fairly common UI widget.

When optimized for G4s and multiple G4s, software can be very zippy on a Mac. In raw power, they are still lacking behind in Intel, all us Mac lovers are sad to admit. But this article proves that Macs, compared to PCs in their price point neighborhood, can hold their own if developers like the marketing-driven ones at Adobe, took the time to make their software take full advantage of the machine they're working on.

Re:So in other words....

[Llywelyn]

" Does gcc or MetroWorks compile binaries that are optimized for altivec but have slower code for older processors?"

It is simple enough to check and several of the libraries that are commonly used (as opposed to directly using the altivec commands) do exactly this.

" Adobe could have rewritten some key functions to be altivec optimized. "

The vast majority of designers who use Adobe on the Mac have at least one G4.

". Furthermore, Apple has been adding free/lowcost software that competes with Adobe and other manufacturors."

Last I checked, half of Adobe's profit came from the Macintosh--at the least on certain product lines.

Re:So in other words....

[GlassHeart]

not every mac user has a dual G4. Or even a G4.

Since graphics professionals are somewhat more likely to require and afford top-of-the-line hardware, it's not an unreasonable assumption to optimize for at all.

Adobe could have rewritten some key functions to be altivec optimized. But the Macintosh market is smaller, (and therefore less potential profit).

Be careful with your math. Although the Macintosh market is comparatively small, it's not necessarily a correspondingly small part for a maker of software for graphics professionals. I would not be surprised if the Macintosh products bring 10% (double the apparent Macintosh market share) or even 20% of Adobe's sales.

Re:So in other words....

[afantee]

>> Macintosh products bring 10% (double the apparent Macintosh market share) or even 20% of Adobe's sales.

It's actually 30% according overall, according to Adobe's latest report. Excluding consumer products like Acrobat and Photoshop Album, the figure may well be over 50% for their professional lines.

Frustrating

[Niles_Stonne]

Reading the "Mac vs. PC III: Mac Slaughtered Again" article was an exercise in frustration, particularly when the other link states how to nearly double rendering performance. The writing in the article just generally pissed me off.

Anyway, perhaps it's time to send a nicely worded e-mail to Mr. Charlie White. A real benchmark may be impressive, considering a 3.06Ghz Dell system was only 2 times faster than the 1.25Ghz G4 in the best case - a Mhz/Mhz comparison should put it closer to 2.4 times faster.

By doing a little analysis of their data, we get the following table:


Test Name (Mac Time in Seconds/Dell Time in Seconds) = Time Ratio
1. After Effects: Simple Animation (14/7) = 2.0
2. After Effects: Video Composite (85/54) = 1.57
3. After Effects: Data Project (227/125) = 1.82
4. After Effects: Gambler (43/29) = 1.48
5. After Effects: Source Shapes (426/254) = 1.68
6. After Effects: Virtual Set (495/264) = 1.88
1. Photoshop: Layer styles & transformation (7.1/4.5) = 1.58
2. Photoshop: Filter Effects (62/35.1) = 1.77
3. Photoshop: Manipulations and adjustments (4.5/3.4) = 1.32

Average Time Ratio: 1.68


That means that On Average The Mac, running at 1.25Ghz only took 1.68 times longer to do the same task.

The "How to Double your After Effects Performance" article averages almost exactly a 2x speed increase, taking on average 49.2% of the time on the "long tests". Taking this into account, we can look at the two longest tests, factor in the speed increase, and look at the "final" performance:


5. After Effects: Source Shapes (209.65/254) = 0.83
6. After Effects: Virtual Set (243.61/264) = 0.92


If we go through the same calculation for _all_ of the calculations, we get the Mac BEATING the Dell, taking only 83% of the time that the Dell took. If we take the Photoshop calculations out of the average, it is still at only 86% of the time the Dell took.

Note that these are all calculated values, and as such may vary significantly from actual values, also that I used the improvement on the "long" test for everything, and did not use the improvement on the short test at all - I imagine that the two systems would end up nearly even in a real test.

Apparently all of my tables were lame when I had them formatted, thanks lameness filter!

Re:Frustrating

[Niles_Stonne]

Read both of the articles, you will see that After Effects only uses one processor - hence the ability to "double your performance" using the dual processor "tip" from the second article. My calculation is correct, given that the second processor is roughly equivalent to a brick as far as After Effects is concerned in it's standard configuration.

The entire point of my post was to extrapolate from given data what the data may have looked like if White's testing had used the dual processor tip, instead of just using a single 1.25Ghz G4 vs a 3.06Ghz Intel.

Re:Frustrating

[Llywelyn]

Do you know *anything* about SMP?

1) If the mac was only using one processor, then while a dual system is still faster (OS tasks can be sent to the other processor, other apps have more of a pipeline to be scheduled in, &c) it is nowhere near the point of using all of its processors.

2) Two single processors have advantages and disadvantages when compared to one faster processor. Some of the advantages include that it will feel more responsive and will actually be faster when performing lots of smaller or a mixed group of CPU based tasks. When performing a single large tasks (such as what AE tends to do without this hack), each one independently, the dual processor system suffers greatly.

Dual processor machines are just completely different beasts--adding their MHz and comparing them that way is a very poor way of scaling them per MHz against a single processor system, since it is completely and totally nonrepresentative of how the systems work.

Re:Frustrating

[GlassHeart]

Mr. White is [an idiot], if only because he believed the results of his tests enough to publish them, and make a fool of himself

I don't think there is reason to believe that the White benchmarks are invalid. First of all, they are out-of-the-box configurations, still valid for those who don't want to hack their software. Secondly, it predates this dual processor hack.

Its conclusion can now be called into question, but I think "idiot" is far too harsh. If he is intellectually honest, he will redo his benchmarks applying this recent discovery.

Re:Frustrating

[Gropo]

If we could, how about a fair comparison: Mac G4 dual 1.42 GHz versus dual P4 3.06 GHz. Dual versus dual, or dumb down the Mac, single versus single.

Better still, let's compare a cluster of G4's versus a cluster of Pentium 4's... Taking in to account OS X's native Multilink Multihoming capabilities and freedom from the shackles of the PCI bus.

Or, we could just compare prices, in that two 1.25 Ghz G4's from Apple cost roughly the same as a single 3.06 Ghz P4 from Dell... Any idea what two 3Ghz Xeons would cost? Any idea what two XServe Dual 1.3 Ghz "Cluster Models" would cost?

Re:Frustrating

[dtfarmer]

but I think "idiot" is far too harsh. If he is intellectually honest, he will redo his benchmarks applying this recent discovery

Yes, idiot might be too harsh, but from what I've heard of him and his demeanor towards macs, I seriously doubt he will redo the benchmarks using the full power of a dual processor mac. I certainly hope he proves me wrong, for the sake of his "intellectual honesty".

Take for instance the entire tone of the article, which can be summed up by the title he used "mac slaughtered again." Now ask why he never once questioned the performance of the Mac, and why it performed as such? He never even asks if the fault of the performance lies with Apple (hardware) or Adobe (software). He wants to lead you believe it is the Mac hardware that falls short, when the software is as much or more to blame for the performance. Had the software he was using been threaded properly (and Adobe knows how to do this, see Photoshop) to take advantage of the multi-processors, which this "hack" allows you to do, his precious new Dell with it's hyperthreading might not have fared so well.

Nonetheless, when we get right down to brass tacks, my only point (which I admit I overdid in using the term idiot) is that Mr White has a bias, and his tests were meant solely to reinforce that bias, just as Photoshop bakeoffs are meant to reinforce a pro-mac bias.

Re:Frustrating

[yomegaman]

That's nice that a 3 GHz G4 would whup a 3 GHz P4, but who cares? There's no such part. I agree that clock speed isn't everything, but Intel processors still whip any G4 that actually exists. Sure, you can go dual, but you could also go dual Intel with the same trick and we're right back where we started.

iMovie 3

[Llywelyn]

"iMovie3 on the other hand, sux so hard it would put Miss Lewinsky to shame."

That being said, iMovie 3 is free and better than any other free editor I have ever seen.

I also should say that, in my limited working with it, I have not had any problems.

For those who are doing more than making home movies, Final Cut Pro and Final Cut Express are reasonably priced.

Sorry Charlie, Dave, my fault

[markomarko]

The last thing I wanted to achieve with my posting was another shouting match over the validity of Charlie White's benchmarks. If some more of you had bothered to read the second article, you would realize that this isn't a software "hack", it's a technique that employs adobe software included with the After Effects production bundle, and that it does, in real world terms, double render speeds on many After Effects tasks. Charlie and Dave were both good enough to reply to my inquiries about their tests, and the relationship between them. Seems Charlie is planning another test soon, using the "render farm" technique on both the Mac and the P4, but with the latest hardware. And yes, the technique provided by Dave did come after Charlie's article. Two things I'm disappointed with: One, that adobe didn't recognize that the benchmarks they published didn't make full use of both processors in the mac, when they have software that will do so. Charlie wasn't aware of this trick, and that's fine, but I would have expected Adobe to know more about their own software. Two, (yes I'm saying it again)the number of slashdotters that don't actually read the articles they comment on. To Charlie and Dave: I publicly apologize if my invitation to revisit the results has resulted in a mass of e-mail you don't need, or otherwise inconvenienced you. To Charlie in particular: I in no way meant to slander your test. I just wanted to show that the dual 1.25 is actually capable of much better performance with both processors being used to their full potential.

Oh yeah,

[markomarko]

And mod me down to (-1 obviously impaired) for thinking a hardware comparison between Apples and P4's would be commented on sanely.

please... It's not a "HACK"

[Satchel Buddah]

It's called "using the provided software". Unless you consider using expert features of a program a hack. It is not uncommon to use the network renderer in pro houses for heavy comps. I never thought of using two "render nodes" on the same machine tho, but it makes total sense. The big difference of AE for NT/2K/XP vs AE for Mac is that AE for windows is threaded and uses multiprocessoring (has been for years), while AE for Mac is not using both processors. It is pure sillyness and lazyness from Adobe (or strategic evil payback to FCP by trying to move people out of Macs), since ALL the professional Mac animation/compositing monkeys use dual proc machines. Mac Duallies are standard in the industry for over 3 years.. Anyway, Adobe needs to seriouly overhaul its rendering sub system (practically unchanged since they bought "Causa After Effects 2.0", way back). They should use a self replicating render deamon, like many more efficient compositing systems. Maybe do something fancy like rendez-vous rendering a la Shake 3. While not "top end" AE remains a bargain for the all the functionality it offers and remains the standard swiss knife for many effects/animation shops. As for the guy stating that Mac users prefer to use FCP and Shake, obviously he does not know what he is talking about. Totally different stuff. FCP is a (very nice) rather unexpensive, feature packed non linear editor with some compositing functions, but not a full fledged compositor like AE. Shake is a great compositor for the wealthy ($5k license on macs, while AE production is what, $500?) with many great features (color tools, keying, easy scripting all over the place, SPEED) and unparalleled control, a toy for rather expert compositors. It also inherits from its "linuxy geeky" past many annoying inconsistancies (undos are ***ed up, UI is on the garish/unreadable/confusing side, file acess looks linuxy...). it is also lacking some features (hello ? where is the mesh warping ? Crappy timeline also). It will turn out beautifully in version 4 once apple spends some time polishing it. Version 2.5 was just a quick port on the mac platform ("look Steve ! it does run on mac") but gave a feel of work in progress, performance wise and UI wise. Didnt try new vers 3 yet.

is speed important to everybody?

[nuckin futs]

Speed is not the reason why some people use Macs. They use it because of the OS and bundled apps. Having a nice looking computer,complete with dual processors and altivec to speed things up is just a bonus.

Re:Are you kidding?

[jlaxson]

Umm, no. The Mac vs. Dell comparison was done using one instance of the AE renderer. It was only using 1 CPU. Thus, he was comparing a 3GHz P4 to a single 1.25GHz G4.

That's kind of interesting, considering that the Mac scores were (usually) only 5-20% slower than the Dell, even though the Mac CPU was running 60% slower.

I would like to see the results of the test if one were to use both CPUs (2 AE Rendering Instances) in the Mac. Perhaps the 'Puts it to shame' comparison will be pointing the other direction.