© 7-18-03 Joe C. Carson Original Article: http://www.applelust.com/oped/amc/archives/amc030718.shtml On June 23, 2003 Apple officially unveiled the PowerMac G5 machines that use IBM's new 64 bit PPC 970 processors as well as a host of very high end technologies packed into the new aluminum case. Steve Jobs hosted a presentation that showed the test results obtained by Veritest comparing a 3 GHz Pentium 4 based Dell Dimension 8300 vs. a 1.8 GHz single processor G5 and a Dell Precision 650 workstation sporting dual 3 GHz Xeon processors vs. a 2.0 GHz dual PPC 970 PowerMac. Almost before the presentation had finished the Wintel portion of the web started screaming its usual denials and making charges that Apple somehow cheated on the tests. In fact, one paid Wintel oriented writer at CNET.com actually "predicted" that Apple would cheat on G5 benchmarks. How dare Apple claim a mere Macintosh could be more powerful than a mighty Dell?! The problem is that Veritest had provided detailed proof of how they ran the tests, but never fear, the Wintellers never let mere facts and reality stand in the way of a well established prejudice. Nothing new here. This absurd behavior has been an ongoing circus for many years. What is different this time was that Apple and Veritest posted the benchmarking in gruesome detail, unlike Intel's secretive methodologies, and the Wintel Weenies immediately denied the published facts in an effort to maintain their delusions of Wintellian superiority. This time they went too far. I have spent the better part of the past several weeks checking out the facts in the case. I did something that few if any of the Apple-hating Wintel accusers did... I actually read the Veritest lab reports. I also checked out Veritest's reputation for honesty in the high tech industry and found that Veritest is squeaky clean, unlike their Wintel fanatic accusers. I frequented geek sites to find out the real facts about the technical accusations and found that the general opinion was (when they were being charitable...) that the original sources of the accusations were at best, highly incompetent. Of course, none of the real facts discussed by the in-the-know types ever made it to the general media. After all, truth is not usually a good sell to the Wintel world. Besides, the Wintel flacks might jeopardize their paychecks if they told the truth. I also discovered the point of origin of the false accusations that have been circulating all over the Wintel oriented web media. The source will be named later in this article. I think that it is possible that two companies involved in the test (Veritest and Luxology) may have their legal departments exploring the possibility of filing libel lawsuits against those who made what are appearing more and more to be deliberately false statements. That may be part of the reasons for Intel's curious response to the tests. More on that later. First of all, here is a compendium of the false claims made against Apple, Veritest and the other companies involved, and the truths I uncovered.
Falsehood: Fact:
Falsehood: Fact: Unlike the Intel ICC compiler the GCC 3.3 compiler tests the entire system's performance, not merely the processor alone. It also is the default compiler provided with the Red Hat Linux 9.0 operating system that was used to compiler and run the SPEC2000 benchmarks on the Dell machines. Red Hat is not run by stupid people and they would want to use the compiler that produces the best possible results on Red Hat Linux as is possible. The GCC compiler is in fact very highly optimized for x86 code... more so than for PPC code. This isn't because of any prejudice on the part of the providers of the GCC compilers but because they have had much more experience in fine tuning x86 code than PPC code. Besides, the GCC 3.3 compiler's 64 bit PPC compilation capability is very new and has not had time to fully fine tune the final code production capability. This will probably improve over time. Intel's proprietary ICC compiler is a poor choice for an honest test because as everyone inside the industry knows, Intel's compilers are infamously rigged for benchmarking. In the past Intel used what was known as a "reference compiler", i.e. a compiler that had only one useful function... to generate very fast but commercially useless code specifically to skew SPEC benchmark results. Over the years Intel has gotten better at this. Intel now has a nominally useful compiler that when it recognizes a SPEC benchmark being compiled will automatically use special math libraries that produce very fast SPEC benchmarks but cannot be used commercially because the resultant code is unstable and produces inaccurate numerical results. Motorola once tried using a similar trick to generate spuriously fast code on early G4 processors, the infamous "MotoMathLib" extension. They got caught and were roundly condemned by the Wintel oriented media. Apparently only Intel is allowed to cheat. Veritest could have cheated for Apple by using IBM's proprietary Power4 compiler which is compatible with the PPC 970, but that would have been called cheating by Wintellers. Again, only Intel is allowed to cheat.
Falsehood: Truth:
Falsehood: Truth:
Falsehood: Truth:
In other words, Hyperthreading is just another Intel technology that does not work on currently produced machines. If Veritest had in fact wanted to rig the test the easiest way would have been by enabling Hyperthreading which, as Dell points out, can cripple a PC's performance by as much as 37%.
Falsehood: Truth: It hardly makes sense to run tests with the computer crippled by obsolete legacy systems intact if they only serve to hamper performance of the machine. However, the Wintellers seem to feel that a test is fair only of the Apple machines are crippled in some irrational manner... like leaving legacy G4 software prefetch enabled.
Falsehood: Truth: A closer look at these applications tells some inconvenient truths that the Wintel Weenies would rather the world not notice. The lies are so much more convenient. Mathematica 5 Wolfram Research, the publishers of Mathematica 5, is not particularly Mac-specific. In fact, it wasn't all that long ago that Wolfram temporarily abandoned the Mac while continuing to support Windows as some of us cursed with accurate memories recall. The explanation was that Mathematica 5 uses algorithms that work well in a Unix environment, passably well enough under Windows but were difficult to support under Classic Mac OS. When Apple moved to FreeBSD Unix based Mac OS X, Wolfram came back to the Mac for the simple minded reason that the new operating system was an ideal environment for Mathematica 5. HMMer HMMer (Hidden Markoff Methodologies) is an Open Source (i.e. Unix/Linux) gene sequencing program that can be downloaded on the web from http://hmmer.wustl.edu/ as source code. You then compile it under your favorite version of Linux or Unix using a compiler of your choice suitable to the hardware platform you choose. Since this is platform-agnostic Open Source code, claiming that it is somehow Mac-specific is pretty ludicrous. Luxology Luxology is a new startup that is developing a new 3D software package for Mac OS X, Windows XP and Linux. They were stung rather sharply by the Weenies attacking the software demonstration showing their new application to perform much faster on a new G5 than on the best dual Xeon machine Dell has to offer. If you go to their web site one of the first things that will greet you will be a "response" that has a few choice words for the Weenies who made claims against them. Luxology took it rather personally. You can read their response here: http://www.luxology.net/company/wwdc03followup.aspx. Luxology felt compelled to respond to the charges obviously because if they aren't refuted that they may suffer possible sales losses in their projected dominant area of sales: Windows users. Luxology is a new company with software still in development and has not yet released their intended product for sale. Having false charges leveled at them about rigging their product would definitely damage potential future sales in the Windows market. Luxology points out that:
For the Windows machine Luxology used dual 3 GHz Dell Xeon machine. For the Mac, a dual 2 GHz G5, of course. Both machines used Radeon 9800 graphics cards. Both machines used software based on OpenGL. Neither machine used SSE/SSE2 (Dell) or AltiVec (G5), therefore the software tested only raw computing power, not SIMD optimizations. The test was intended to push large amounts of data to compare raw processor throughput speed and bandwidth. Luxology is definitely less than pleased with the false charges from the Wintel Weenies. I wonder if their legal department has been exploring the possibility of lawsuits to claim damages? BLAST BLAST is a special gene processing program from Genentech. In case anyone hadn't noticed, biotech is their main product, not software. BLAST is a platform-agnostic application Genentech provides to professionals in the field and is fully optimized for each software/hardware platform it runs on. The Mac OS Xspecific version of the program is called A/G BLAST, but other platforms such as Windows are also fully supported. Accusing Genentech of being a Mac-specific company tends to look pretty ridiculous. Cubase VST Cubase is a professional level recording software package that supports both Mac OS X and Windows XP. If anyone has ever made a trip to the NAMM expos for professional music vendors, you will see that Cubase is far from a Mac-specific company. Their booth tends to be Windows dominated, so if they were so Mac-specific as the Weenies claimed, why all the Windows demonstrations? Sounds rather odd to me. Adobe PhotoShop 7 Here we go again! Every time a PhotoShop cookoff is run on a PC and a Mac, the false accusations fly like a manure in a Texas tornado. Why does Apple use PhotoShop for the cookoffs? Because it is one of the few applications that is fully optimized to the max for each platform for which it is intended. Well... maybe not fully optimized for the Mac because like several of the commercial apps already mentioned, it is not 64 bit optimized for the G5. PhotoShop stresses a computer's entire system and components such as drives, I/O, processor at all levels, graphics, memory access, etc. Usually the Weenies claim that Apple picks and chooses a few filters that favor the Mac. Well, this time Apple chose a full suite of 45 filters, so making that claim again rings pretty hollow. Never mind, the Weenies repeated it again anyway. It's an old habit that's hard to break. Is Adobe really Mac-centric? Well, maybe not so much now that Adobe has dropped all Mac support for Adobe Premiere and hasn't updated the Mac version of After Effects to full optimization for the AltiVec registers or multiple processors. This is why After Effects is a favorite of a certain Wintel Weenie videography web site to claim that Macs are "slow". Adobe's actions are based on the fact that Premiere sales for Mac have been dwindling to nothing ever since Apple introduced first Final Cut pro, then Final Cut Express, and now has a hot deal for anyone who turns in their old Premiere CD to get Final Cut Express for a measly $99. You also have to remember that Adobe makes approximately 53% of its PhotoShop sales to Windows users. With that in mind, would it make sense for Adobe to fail to support their Windows versions fully while devoting more time to the Mac versions?
Falsehood: Truth: Without going into a lengthy breakdown of every dinky feature, I did get a fairly close feature match on both the 1.8 GHz (single processor) PowerMac G5 versus the Dell Dimension 8300 (3 GHz Pentium 4), and the 2.0 GHz (dual processor) PowerMac versus the Dell Precision 650 (dual Xeon 3 GHz processors). I had to add a large number of features to the Dell to make them as close as I could despite the lack of available features on the Dell machines. The Dell machines lacked available Serial ATA drives and lacked FireWire capability. The Dell Dimension seemed to lack PCI-X availability although it is available on the Dell Precision. Neither of the Dells offered any sort of standard wireless capability as an option. If it is available then it will be an extra cost third party add-on. I deliberately went to the nearest but smaller of the available hard drives for the Dells to keep their prices down as much as possible. A larger drive would have cost more and driven the price of the Dell up. I did try to cut the Dell machines as much slack as I could. I also chose the nearest default graphics card to models used by Apple. I also chose the lowest cost software bundle offered for the Dells to keep the prices down. The final result: The Dell Precision 650 once configured to match the G5 PowerMac as closely as I could came to: US$4,121... and this was on sale! (US$250 off original list price). This is more than the US$4,040 Apple quoted in the WWDC comparison, so I can hardly support the claim that Apple posted a misleading price comparison. The 2 GHz dual G5 PowerMac sells for $2,999 at Apple's site. There are already slightly lower prices and special deals being offered by resellers. The Dell Dimension was not mentioned by Apple in a price comparison, so to be fair I did one anyway. The Dell Dimension 8300 when cofigured as closely as possible to the 1.8 GHz single G5 PowerMac was US$2,199, or two hundred U.S. dollars less than the nearest equivalent PowerMac that goes for US$2,399. However, that two hundred dollar difference also included PCI-X, wireless capability, a larger hard drive on the PowerMac, a faster hard drive interface and a digital/analog optical drive interface on the PowerMac (not available on the Dell) and both 802.11g and BlueTooth wireless capability, also not available on the Dell. I think that might even up the cost vs. feature set a bit. Consider the Source Where did all this propaganda come from in the first place? From what I can determine, it all originated form a supposedly anonymous posting on the web at Haxial.com, an Australian based shareware company. The "anonymous" poster used a favorite tactic of Weenies of all stripes... if the straight facts tend to be inconvenient to your position, then bury the opposition under a mountain of meaningless techno-babble. The "anonymous" poster who spent some time blathering a lot of techno-babble managed to convince a lot of Wintel oriented types that Apple had somehow cheated. The problem is that some people actually can understand techno-babble and I am one of those inconvenient types. Shortly after this posting appeared, it was taken up quickly first by Paul Thurrott who operates the Wintel oriented site WinInformant. Mr. Thurrott is a Wintel fanatic who seems to have dedicated an excessively large portion of his waking time to bashing Apple at every opportunity by repeating the usual anti-Apple misinformation. He did it again by regurgitating the stuff from Haxial and as usual, never checked out the validity of the source. The problem is that one poster at a geek site recognized the poster from his unique style of irrationality. Adam Hinkley is the author of the Haxial diatribe making claims against Apple's benchmarking. To get a handle on this fruitcake, go to... http://macslash.org/article.pl?sid=03/06/25/109246&mode=thread ...and look for the posting titled: "I know the guy who wrote that critique" Follow the threads and Mr. Hinkley's other less-than-sane diatribes and you will get a handle on how irrational the Apple-haters are. Why does anyone take these Weenies seriously? Equating dog-breeding to elitist racism is just plain... NUTS! As far his other diatribes such as his views on religion... don't go there unless you are prepared to get angry. Mr. Hinkley also has an interesting habit of accusing other developers of stealing his code. Yeah, right. It was his postings that made most of the false claims I listed above. I found it interesting the the refutations of his misinformation was gleaned from the geek sites as other developers tore his claims apart. Oddly enough, none of the knowledgeable refutations ever appeared on the mainstream web sites. I have to wonder why. Birds of a Feather Flock Together That some dedicated Apple-hater would post another anti-Apple diatribe making the usual false claims about Apple rigging the tests and libeling several companies in the process is not unusual. That is the way the Great Wintel FUD Machine operates. That CNet would post a prediction that Apple would rig the tests and then repeat the slanders despite the solid evidence to the contrary that was available to any who would bother to take an honest look is as predictable as discovering that the Sun will rise in the East. That the Wintel oriented media would repeat the stories without checking the validity of the sources is both predictable and disappointing. It only exposes them as pseudo-journalists... hacks who will say anything to guarantee a paycheck. In the aftermath of the shouts of denial I have completely lost respect for some supposedly unbiased sources such as The Gartner Group. Long ago the Gartner Group was a studiously unbiased source of objective studies, but over the past few years that objectivity has been clouded by the presence of so much Wintel cash. The obvious loss of objectivity has taken voice in the form of Martin Reynolds, the Gartner Group's tech analyst who has consistently shown a strong pro-Wintel and anti-Apple bias, such as his slamming the XServe as something that has little impact on the market and has very little future, to joining the claque of anti-Apple shouters who declared that the SPEC results by Veritest were not believable. I am sure that Veritest may have also concluded that Mr. Reynolds and the Gartner Group now have lost credibility just as I have. Mr. Reynolds may have exposed himself unwittingly when MacFixit decided to confront the thunderous silence that was issuing from Intel. Intel declined a direct answer but instead redirected them to Mr. Reynolds of The Gartner Group. Since Mr. Reynolds was already on record of declaring that Veritest's G5 SPEC2000 benchmarking results were not believable, the following bit of backtracking emanating from Mr. Reynolds literally within hours of MacFixit's confrontation of Intel is remarkable:
Mr. Reynolds only managed to prove that he was definitely in Intel's pocket and was obliged to support whatever public position that Intel deemed necessary. Shortly after this bit of backtracking was posted by Mr. Reynolds, other Wintel media flacks started backtracking as well. In short, Intel told them to shut up and they dutifully obeyed their master. Intel's "Advanced 875 platform" means Intel's top of the line chipset/processor combination used in high end "Northbridge" machines. Reading between the lines it is obvious that Mr. Reynold's admitting that a G5 can at least equal a top end Pentium system means that the G5 actually does exceeds the current Pentium 4 capabilities, but he is quite loathe to admit it openly. Would anyone expect Intel to admit that a PPC 970 based machine can exceed Intel's best efforts? I think not. To date no one at Intel has publicly admitted that AMD's offerings also exceed Intel's Pentium 4 performance either. Would anyone expect someone such as Mr. Reynolds or anyone else who is solidly in the Wintel camp to admit this? Silencing Fellow Travellers Over the years, Intel and Microsoft have built up The Great Wintel FUD Machine. Some parts of the FUD machine are paid flacks but much of it is composed of Fellow Travellers. There is never a shortage of those who will seek to bask under the glow of perceived immense power whether it be political or religious, or due to the charisma of some sort cult or another. The Wintel Hegemony is no exception to this pattern if human misbehavior. The great mob of Wintel Cultists willingly take up the cudgels to bludgeon into oblivion any perceived threat to their chosen Cult of Technology. The problem is that on occasion the mob can get out of control. That is what has happened here. Intel has a problem. Part of their marketing over the past several years has been their fosterage of the Megahertz Myth. The problem is that Intel engineers know that the IA-32 (aka, "x86") Instruction Set Architecture is rapidly reaching its end of potential. Yes, they can crank up the clock speed but even now the Pentium 4 shows only modest actual performance gains for each bump up in processing speed. I have estimated that it is now only at about 2% performance boost for each 10% clock speed increase... and declining. Intel originally planned to replace the aging IA-32 architecture with a new EPIC based pseudo-RISC architecture originally code named "Merced" which eventually became the Itanium. This new processor flew like a lead balloon in the market place largely due to its high price, low performance and lack of developer support. Intel had no choice but to continue to pump up its aging IA-32 processors while trying to find a way to get acceptance for the Itanium and Itanium-2 processors. Intel has shot itself in the foot over their promulgated Megahertz Myth by trying to tell the public at large on the one hand that clock speed was the whole story while telling IT and Enterprise types the exact opposite about the Itanium processors. Intel has also tried to downplay the Megahertz Myth for laptops with the new low speed/higher performance processor chipset for laptops. The result is an irrational, contradictory and schizophrenic marketing position. Intel has managed to do a series of Devil's Deals with IT types to marginalize AMD's processors despite the fact that they show much better performance than Intel processors but at significantly lower clock speeds. Only hard core AMD aficionados seem cognizant of those facts... or even care. Intel's problem is that they have no viable 64 bit processors ready for desktop use, having put all of their efforts into developing an unwanted 64 bit processor that is incompatible with the established IA-32 code base. AMD has managed to take a 32 bit processor and kludge together some extensions to the instruction set that allows a modified K7 core to emulate a true 64 bit processor. There has been considerable interest in AMD's Opteron and Athlon-64 processors. The Opteron is a high end version of this extended IA-32 code set that is intended to be used in high end workstations and severs. The desktop version is the dumbed-down Athlon-64 but it won't be out until sometime in September of this year. Nonetheless, Intel has been successful in diverting attention away from these new "64 bit" processors from AMD by continually cranking up empty clock speed for the Pentium 4. Apple's new PPC 970 based PowerMacs are a serious problem for Intel. Intel can't say anything about the new G5 PowerMacs. Intel is in a no-win marketing position against the new G5 PowerMacs. Any commentary about them either pro or con only invites outside parties to check things out for themselves. This is why Intel had to order their stooges to shut up. Too much notice now only invites trouble for Intel by getting people to take a close look. Apple and IBM have conspired to set up a bad situation for Intel by deliberately baiting the Wintel FUD Machine into a loud response to their benchmarks. This was done in full knowledge that the new PowerMac G5s are more powerful than even Apple has admitted. For one thing, how many of you noticed how much better the G5 PowerMacs performed on real world applications against the Dells than the SPEC2000 benchmarks indicated? This is because Windows hampers the potential performance of a Pentium 4 processor while Linux, the operating system of choice for the Dells under the SPEC2000 allows the machines to run unencumbered by Microsoftian bloat. How many of you are aware that most of the Mac OS X applications used for the real world applications cook offs are not 64 bit optimized but are off the shelf 32 bit legacy applications while the Windows applications were fully optimized for the Dells? How many of you realized that Veritest gave the Dells 2 Gigabytes of memory but only one and a half Gigabytes for the Macs? I am sure that none of the Wintellers accusing Apple and Veritest of cheating bothered to notice those little details. They are much too inconvenient. Not only that, but at least two web sites noticed that in fact the new PowerMac G5s are workstation class machines in everything but price. The presence of 8X AGP, PCI-X expansion, very fast I/O connections (FireWire 800, Gigabit Ethernet), and one of the fastest system buses available (200 MHz faster than the fastest Pentium 4 system bus), Hypertransport's wide open system bus as compared to a Wintel PC's bottlenecked single chokepoint system bus, not to mention a true 64 bit RISC processor with a fast bus system, superfast caching and true Unix based operating system all sound more like a workstation... but at desktop prices. You might want to check out "Is the Mac G5 a PC or a workstation?" posted by Nebojsa Novakovic at the Inquirer, U.K. Nebojsa Novakovic asked the question, "Is the G5 a PC or a workstation" and answered it as follows:
In short, Apple is selling a powerful 64 bit Unix workstation but at desktop prices! No wonder the Wintel Weenies are screaming in rage at the SPEC2000 benchmarks Apple published. The reality of what the G5 is must be scaring the bejesus out of them. The Smell of Fear This whole brouhaha over the Veritest benchmarking stinks mightily. I smell a rat... in fact I smell a lot of rats. I smell the stink of hypocrisy from some corners but in the end all is overpowerd by The Smell of Fear. Why? Because deep down underneath the Wintel Hegemony's longstanding arrogance dwells the gnawing fear that eventually the public might wake up and realize that the Wintel Hegemony has built everything on stolen, copied and second rate technology. They have kept this under control with paid stooges in the media, well placed FUD plants to divert the public away from the sleight of hand tricks that have been played on them. They have spent a lot of PR time loudly proclaiming "innovations" that they conveniently forget to mention are almost invariably something they copied from Apple, often years later and usually copied badly. The problem is that they are aware that Apple technology has always had the potential of marginalizing them if the Masses of the Great Unwashed ever found them out. The problem came home to roost when Apple announced the PowerMac G5. The PowerMac G5 delivers everything Wintel has only promised. Most of the general public won't become aware of what has happened for some time to come, but already the Wintel Hegemony has been losing ground in content production to Macs, even when the G4 failed to be developed in a timely manner. The new G5 leapfrogs Wintel's now-obsolescent 32 bit technology with Apple's solidly developed user interface blended with a powerful Unix cored operating system running on workstation class hardware but available at desktop prices. Major content creators cannot avoid the obvious, that the PowerMac G5 is exactly what they want. Apple's new moves sidesteps Wintel's domination of cheap commoditized low end home machines by thrusting sharply into the heart of serious scientific, graphic, music, video and movie production. Only the most arrogant, obtuse and fanatic Apple haters in those fields can ignore the obvious, that Apple now really does have the fastest desktop machine in existence despite the loud denials issuing from the usual suspects. Now to make it worse... The PPC 970 is only a relatively inefficient design when compared to its planned successor, the PPC 980. Already the PPC 980 is being developed and exists now in prototype form, being readied for introduction sometime late in 2004 according to current schedules. The PPC 980 is supposed to have as much as four times the performance capabilities at matched clock speeds over the PPC 970. There is the planned PPC 990 to appear sometime by 2006 and reach clock speeds up to 10 GHz. Then there is the recently mentioned PPC 9900... Lordy! Does it ever end? Yes, I detect the rank stench of The Smell of Fear rising from the piles of verbal manure being tossed around by the Wintel fanatics. The only reason that they have reacted so violently to the Veritest SPEC2000 benchmarks is that they know full well that they are real, and that reality spells the eventual end of Wintel. My nose bleeds for them. © 2000-2006 Applelust.com.
All rights reserved. No part of this publication may be reproduced in any
way without prior, expressed permission from the Publisher. It is the sole
property of Applelust.com and its writers, who retain copyright to their
own works. If you wish to link to us, please see our Privacy
Statement for conditions. Apple, Macintosh, and Mac are trademarks of
Apple Computer, Inc, with whom we are in no way affiliated or endorsed.
|