Where Does The G5 Really Stand?
So far there’s been much screaming and little substance. Here’s my best estimate of the situation.
Is The G5 Really Faster?
We’ve explained twice, here and here
why the Apple spec-tacle is a good deal of nonsense, so we won’t repeat ourselves.
However, per Apple’s spec score, it’s likely that the G5 implementation got rushed a bit, so the compilers for it really aren’t tuned up yet.
Looking at the results from other companies, they can often tweak the results up as much as 10% or so. In this particular case, the scores may go up more because it is unclear whether the version of MacOS X used (10.2.7) uses (or at least is highly optimized) for 64-code. So scores may go up some more because of that, perhaps as much as 15% or so.
In all likelihood, a truly tuned-up G5 will likely come within shouting distance of the best current x86 processors.
Per Photoshop, I don’t doubt tests like PSBench will show a dual G5 beating a single PIV or Hammer pretty handily, though not by the margins Apple is claiming.
PSBench is a standard Photoshop benchmark used by those in the PC and Mac world. You can find it here.
Current testing shows a 3.0GHz HT-enabled PIV nudging a dual G4-1400MHz, so a dual 2GHz system should do considerably better.
Whether Apple would win a single/single or dual/dual competition is another story. That’s at least questionable. In all likelihood, the details of Apple’s Photoshop test would reveal that a few or more of the PS filters used in the Apple test but not present in PSBench that are simply optimized much better for the Mac than the PC.
Per the other tests, the margins of victory indicate that they represent cherry-picked results. I mean, really, how often do you find AMD being even twice as fast as Intel or vice versa in any real test?
Nonetheless, it’s probably very safe to say that the G5 will get the Mac to at least within shouting distance of x86 processors.
Can IBM Keep Scaling The Heights?
One big problem Apple has had in the past is that it often manages to more-or-less catch up with x86 processors, but then it can’t keep up with them as they steadily improve (until the G5, the last few years, they haven’t even been able to even catch-up)
Primary blame for that falls, not an Apple, but Motorola. Motorola simply couldn’t keep up.
Some of it was due to technical difficulties: the PowerPC chip has traditionally had a shorter pipeline than x86 chips (and much shorter than the PIV’s), which makes scaling tough
The rest of it seemed to be that Motorola just didn’t give much of a damn about it. IBM was a second-source for PowerPC chips, and they managed to come out with better, faster PowerPC chips that ran on Apples than Motorola could.
In any case, the G5 is an IBM only CPU. IBM has also promised that they’ll be able to get the current 2GHz G5 to 3GHz in a year.
That seems optimistic given what AMD is projecting for Hammer, but given 90nm transition, it’s not implausible.
If IBM can do that, over the course of the next year, that would be at least good enough to keep up, and depending on how Intel and/or AMD times speed grades (and presuming those spec scores; it’s possible that a G5-based system could temporarily take a lead over an x86 equivalent.
The long-term question then becomes, “Can IBM keep up with x86 after that?” There’s talk of dual-core processors, but whether that will show up on a desktop at a vaguely reasonable price is at least questionable.
The Real Purpose for Apple’s Benchmarks
Of course, you’ll never see this kind of estimate from Apple.
There’s a very simple reason for that. Apple doesn’t use benchmarks to . . . benchmark. They don’t even use it for benchmarketing.
No, these benchmarks are meant as propaganda to tell the Macster disciples what they want to hear.
It’s not good to be close or even win in Apple Valley. No, the only good benchmark is one where Apple annihilates the Intel processor (At least for public statements, AMD doesn’t exist according to Apple).
True, that wasn’t the case with Spec, but the only way the G5 could have whomped a PIV or Opteron was if Apple tweaked the x86 with a blowtorch. Mark my words, once Apple has a good optimized compiler in action, they’ll substitute the new numbers and Apple will “win” that by some huge margin.
To repeat, imagine AMD claiming on the website that an Athlon or Hammer was, say, five times faster than a PIV running anything. They would be ripped apart. Even most of the AMD fans would rip them apart for something like that.
But Apple does that all the time, and outside of a relative few, Macsters believe it, and usually attack even those within the fold who say otherwise.
Macsters will tell you with a straight face, “Of course the G5 is three times faster than a Pentium IV in Photoshop. I saw the demo on the Apple stage.”
No, this is not an isolated loon. I’ve been told that quite a few times over the years by Macsters. To those people, that’s rock-solid irrefutable evidence, and when you tell them it’s not in the real world, more often than not, they’ll tell you that you’re just envious of them.
So what Apple does is tell these people what they want to hear: that they have an elite machine that crushes the common, low-class competition. That’s much more important than say, reality.
The MacMatrix
It’s mad. Just yesterday, I had someone telling me that Apple had Firewire five years before the PC world. When I told him that wasn’t true and that anybody who wanted Firewire on a PC could just buy a Firewire PCI card; it did not compute.
Then I get told how innovative Apple is. Do you know how many legitimate, significant technological innovations Apple has actually made in the last five years? Firewire. That’s it.
Most of Apple’s “innovations” are simply making emerging (and at the time expensive) PC technologies standard before the PC industry does. That’s like Dell claiming credit for hyperthreading. Even more are simply adapting PC technologies.
Outside of the CPU and mobo (along with good industrial design), the rest of a Mac are PC components with a Mac driver or BIOS. It’s been that way for years. Try telling some Macsters that.
Some say, “Why not livw and let live?”
Personally, if somebody uses a Mac because they prefer MacOS X to Windows, I couldn’t care less, and I wouldn’t say all or even most Mac owners qualify as Macsters deep into the MacMatrix.
But when Macsters come crawling out of the MacMatrix and start making all sorts of comments that are simply not true: factually, objectively not true, this bothers me. I don’t like any cause that puts itself above the truth.
Unfortunately, trying to deprogram a Macster from those particular fallacious bits and pieces (no, I don’t try to convert them into Windows use) is rough, and usually fails.
It’s like telling someone just unplugged from the Matrix that they really didn’t eat a steak last night. If you try really, really hard, you might get a Macster to talk like this:
Cypher: You know, I know this steak doesn’t exist. I know that when I put it in my mouth, the Matrix is telling my brain that it is juicy and delicious. After nine years, you know what I realize? Ignorance is bliss.
Many can’t even get that far.
It’s weird.
The problem Apple has is that not too many people want to be plugged in. What they need to do to keep the people they have happy is contradictory to what they need to do to get a lot more of them.
The illusion may not fool a lot of people, but if you end it, you may not have anybody left.
It’s a shame.
Be the first to comment