Advocacy
Dojo (HowTo)
Reference
Markets
Museum
News
Other
Myths
Press
General
Hack
Hardware
Interface
Software
Standards
People
Forensics
Web
CodeNames
Easter Eggs
History
Innovation
Sightings
Opinion
Martial Arts
ITIL
Thought
|
Apple has been running the Bunny-suit / Performance ads for a while now, MSNBC and the Wall Street Journal decided to defend the truth at
http://www.msnbc.com/news/153594.asp
I would find this of far more value if not for the following point --
Where were they when Intel was making claims that could not be substantiated in the real world?.
The article basically discusses the ambiguity of benchmarks (like ByteMarks) which Apple used for this test. Fine. But remember Intel was claiming that MMX made their processors 4 or 8 times faster or more? I do. I remember hearing these lame claims about what it would do. The reality is that it was only useful for some VERY specific things. Most people who saw any gains at all, saw gains at the 10-15% level. There is a slight difference between 10%+ and 800%+. Where were the ever diligent watchdogs of MSNBC and the Wall Street Journal then (both of which seem to have shown a pretty strong bias against anything not by Microsoft)?
The article has gems like --
Intel disputes Apple's claims of faster performance, but Intel spokesman Howard High says his company won't elaborate, and doesn't want to "get into a benchmarking war."
Do you know why they don't want to get into a benchmark war? Because they lose at every single benchmark out there (by varying degrees). The one they lose the least on in Spec's (SpecMarks) -- but there is an explanation. They often lose the worst the more real world the tests are -- like Application tests (using programs like Photoshop), where they usually get stomped on (sometimes by more than the "twice as fast" figure that Apple quoted). So you bet they would like to avoid a Benchmarking war -- they want to stick with ambiguous terms like "Injecting Fun" into a computer, because it has a better rings than saying "a workaround for an antiquated instruction set and outdated processor design, to make it slightly less lame and lock people into a proprietary technology".
Floating Point
For the past 4 years, the PowerPC's (604's) have been able to resoundingly pound the Pentiums into the ground in Floating Point performance -- but they were about the same speed in Integer Performance. The Intel crowd all cried that Floating Point didn't matter -- except for CAD and few Engineering Apps. So Motorola and IBM responded -- and make the chip far faster in Integer (and a little faster in Floating Point). Now the herd is crying that it is Floating Point that really matters. To a very small minority it does, but I would be far more impressed if it wasn't the same crowd that was denying that a year ago. Of course programs (and programmers) can utilize either, if one is dramatically different enough in performance to make it valuable -- so the point is somewhat moot. The only dramatic differences have been in favor of the PowerPC.
Spec
For years, people have been using the Spec suite as the benchmark. Why? Because that is the only one Intel seems to do really well in. The reasons are complex. Basically, Intel uses special hot boxes (that no one can really buy), and uses a special compiler (that no one else really uses), and sets a best conditions -- that no one else can match. In those special conditions, the processor performs adequately. But that is not the real world. The real world is about running Windows and Windows Applications -- where the processors are often 30% (or more) slower than the optimum -- and that would be running WinNT. Where are MSNBC and WSJ on that one? Then there is the truth about running 16 bit Windows apps and Windows95 (98), where the performance hits are even WORSE. Again, where are MSNBC and WSJ on that one? Seems they only want to play truth-cop for their biggest advertisers (or owners).
Application Tests
The real performance tests that matter are Application tests -- doing the work you want to do. When the Bunny Ad first ran, Intel cried foul at Apples ad, and they said, "it is the Application Performance" that counts. So some tests were run and guess what --those test showed that the PPC was even faster than Apple was claiming. Notice how quiet Intel is about "Application Performance" now? Do you notice how the Wall Street Journal and MSNBC don't even talk about comparing Photoshop filters and so on. (1)
(1) A year or so back, Intel made a brief attempt to show that the Pentiums were as fast (nearly) as the PPC's, or more so, because of MMX technology. Sure enough, if you used one of 4 specially optimized filters, only using certain parameters (sizes and effects), then MMX could beat the PPC. And if you were doing only those things in the real world, then you should consider an MMX. But since people are trying to work, not just make Intel look good, MMX is useless to them -- and the PPC trounces the Pentiums. When the truth was found out about Intel tweaking the numbers (by only running certain tests where Intel did good, and so on), where was the Wall Street Journal or MSNBC?
Conclusion
Performance is very complex. Read my series of Articles in Myths: Understanding Performance for more on the issue. The truth is that in the real world, the whole System is what matters -- not just the processor. In those cases, the Mac is better -- but not always as much better as the processor alone. But for Processor Performance, the PPC probably IS over twice as fast as the Pentiums.
What I find most troubling is that many publications are so biased as to become the guardians of their advertisers and owners truths -- and little else. They weren't defending the truth on the dozens of fallacies created and perpetuated by Intel or Microsoft -- but they are sure willing to question Apple. How often have you seen them defend the truth that computer speed is more than MHz? Where are the articles discussing how Intel has the worst history for releasing buggy processors that have been slower than any of the competing RISC machines? How about the hard-hitting articles about the demotion of Intel's leader just last week? Where are the exposes about how Microsoft has made a fortune charging people $300 a pop to upgrade from one buggy version of their Applications to the next? The truth seems to be subjective -- tainted by the reality that Intel and Microsoft have big ad budgets. It also seems that Microsoft buyout of the media (like MSNBC) has had the desired effect, and the reporters have sold their souls for a paycheck.
|
