Joe Carson © 8-5-00

Just a little while ago, in early 1998 we had superpowerful Blue and White G3 machines that could stomp any x86 machine into dust. Even PC Magazine, one of several PC oriented publications known for their carefully rigged tests to enable them to perpetuate the myth of the "slow" Mac, was shocked to discover that when they tested a 400 MHz G3 Mac against a 500 MHz x86 box that the "slower" G3 Mac still outperformed it. I read with some amusement at their discomfort when they reported that even when they tried to cripple the performance of the G3 Mac by replacing the supplied SCSI drive with a slower ATA drives that the G3 still outperformed the then top of the line x86 machine. I was also very surprised that they admitted that a Mac could outperform a PC, and even more surprised that they came so close to openly admitting that they attempted to rig the test.

Then came the G4 debacle. Steve Jobs announced G4 machines in August 1999 at MacWorld NY 1999. These new machines were supposed to be faster than a G3 at matched clock speeds and would be available at 500 MHz very shortly. We all waited and then Apple had to backtrack as Motorola then announced that they were unable to get the new processors to run reliably above 450 MHz. It seems that they had a small design flaw that caused them to produce erratic results at 500 MHz. It was then another eight months before Motorola could actually produce a reliable 500 MHZ G4.

Shortly thereafter Apple pointed fingers at Motorola telling us that Motorola failed to meet their announced availability schedule for the 500 MHz G4. Motorola countered by telling us that the never said that they would have a 500 MHz G4 available for production in quantity when Apple announced the G4 machines.

Meanwhile, the x86 world wasn't sitting still. Already at 600+ MHz in late 1999, Intel and AMD then embarked on a great p*ssing contest to see who could claim the fastest production chips made. Within months we saw claims of x86 processors running at a gigahertz while the G4 was still stuck back at 500 Mhz and no sign of improvement in sight. Of course, it was still even more months before these announced processors actually appeared and even then only in extremely limited real world quantities. Never mind, at least they were announced and in the PC world, saying you have something not yet in production is counted as if you really had it in production. Bizarre.

Of course, in the Spring of 2000 Motorola announced that it had a G4+ processor running at 780 MHz, but this news was lost in the media rush to gawk at the AMD and Intel brawl over who had the biggest... er...bragging rights... to a GHz processor. A mere sub-800 MHz PPC was of no interest in the light of all of the x86 hype. This response was not lost on the Mac oriented media either.

Even the announcement of dual processor Macs still using the 500 MHz G4 only provided ammo to the anti-Apple forces both in the general Media and even in the Mac-centric media to claim that Apple was stuck with a bad [processor and that the PPC was dead in the water. One recent report to this effect surfaced at CNet, "Dual-processor Macs don't impress analysts" on July 21, 2000.

Okay, so it didn't take long for the anti-Apple claque to start sneering at the new dual processor Macs. Nothing new here. Considering that Intel owns a sizable chunk of CNet, I have to wonder if there wasn't some sort of vested interest in snubbing Apple's new products?

In the last few months there has been a flurry of postings calling for Apple to dump Motorola in favor of using Intel processors. After all, weren't they much faster and the PPC a failed processor design? In April, Jim White predicted in his posting on OS Opinion, "A Bold Prediction: Apple Will Build an x86 PC" that Apple would eventually dump the PPC and switch to Intel processor, supposedly because the PPC had failed and only Intel or AMD knew how to make a truly fast processor.

David Egger expressed the frustration that so many of us are feeling about Motorola's glacial development of the PPC in an article posted at MacMonkey, "Screw Motorola", proposing that Apple simply drop Motorola and go elsewhere since it appeared that Motorola had no apparent interest in continuing the development of the PPC.

In yet another Mac OS on Intel story, "Opinion: Mac OS after Windows Split", Mac OS Rumors reported on a Slash-Dot story that they weren't impressed with and then proceeded to make comments of their own about switching from the PPC to Intel that really didn't make any more sense than the other article they didn't like. Oh well... this IS Mac OS Rumors and we have to expect at least a small measure of irrationality in there somewhere.

Why would Apple port Mac OS X over to Intel Hardware when it is PPC hardware that Apple sells? The dumb public would only continue to buy the cheap Wintel crud-boxes and kill Apple. Not smart.

On the other hand, "eric" at big g media (the lower case letters are correct...) in his article "The Megahertz Race" attempted to inject a small measure of rationality into all of the "The Sky Is Falling" rhetoric flying about on the Mac oriented web. This article has a good measure of plain old horse sense sorely lacking in most of the other posts I have seen.

MacSpeedZone has provided a listing of some of the controversy over the "slow" G4 compared to the "fast" x86 chips. This can be found at "MacSpeedZone's Performance News And Information Page - Should Apple Drop The PowerPC, And Other Performance Tales From Around The Net!". It is really remarkable how many of the supposedly "expert" Mac Pundits can panic at any setback Apple may encounter.

So, now we see "slow G4s and "fast" x86 processors. Does this mean Apple is in serious trouble and that the x86 family will emerge triumphant over the PPC? Not quite.

As a current Mercedes Benz ad campaign says, "Not everything is what it seems."

Let's take a close look at the "fast" Intel chips that we have been hearing about.

First of all, I noticed that in the listed specs from AMD and Intel that even the current gigahertz processors are being made in .18 Mm die sizes. Now that is really odd.

To get a processor to run faster, you have several choices of options: Make the die size smaller, lengthen the data pipeline, and then crank up the processor clock. However, I see that the gigahertz PIII and Athlon chips use exactly the same die size and data pipes as the 800 Mhz versions. That means the only real change from 800 Mhz to 1 Ghz is to crank up the clock speed. In other words, they merely overclocked a standard 800 Mhz processor and chose the few survivors on the wafer that could run that fast and still work. Since each chip on the wafer will vary in how fast it can run, there will always be a few that can be pushed significantly faster than the original target speed for the wafer. This also explains why so few Ghz x86 chips are really available and why they cost so much. There aren't really very many that will be able to run this fast in any batch.

This also explains why Intel has announced that they will concentrate on making more efficient versions of the Ghz chips, In other words, try to develop a true Ghz chips that isn't merely an overclocked standard chip. AMD has made no comment about this that I am aware of, but I am sure that they are doing the same.

There is also another little fly in the ointment that neither AMD nor Intel have bothered to mention in their press hype: Thermal downclocking.

Both Intel's and AMD's x86 chips run HOT! They are power hogs and one version of the AMD chip actually uses 65 watts of power! That translates into tremendous heat production, which is why so many PCs emit a lot of noise from the plethora of fans they use to keep themselves from literally melting. Of course, there has to be some sort of safety valve to keep the computer from simply shutting down and crashing on a hot day. That is where thermal downclocking comes in. Each processor has a tiny thermistor included that will send signals to the clock chip to slow down when the chip gets too hot. Since the x86 family run notoriously hot, this means in reality that your GHz processor will not ever run this fast unless you are running it at the South Pole in mid winter. I wonder why neither AMD not Intel ever tell the potential PC buyer about this?

In fact, Gateway has run into this problem with some of their offerings of late. PC World reported this problem in the article "Gateway Hits 1-GHz Snag" It seems that systems using the GHz Thunderbird chip may lock up and Gateway has company suspended shipments.

Then there is the simple fact that the chips of the x86 family are still archaic CISC chips with all of the unavoidable inefficiencies that are inherent in such designs. Despite misleading claims about "RISC Cores" in new x86 chips, they are avoiding the basic fact that all CISC chips have RISC cores. The interpreter circuits translate the machine code into RISC code for the RISC core which then does the real processing, then retranslates it into machine code for the rest of the machine to use. These translator circuits are the primary reason that CISC chips are so large, hot and inefficient. All a RISC processor does is dispense with all of the unnecessary CISC translator baggage. Without this outdated baggage you have a very small and very fast little core CPU chip. That is exactly what a PPC chip is.

Let's also mention one other small fact in the x86 world that few have mentioned. One of the PC publications mentioned in passing that a Ghz PIII produces only a 40% performance gain over a 500 MHz PIII. If you do some quick calculations, that means that currently, an Intel chip can produce only a 4% throughput gain for each 10% of clock speed increase. This is down from the previous 5%. In other words, the point of diminishing returns is not far off for the x86 family, regardless of all the money that either Intel or AMD put into what amounts to turbo-charging a Model-T Ford engine.

Well, this is all nice and good, but we still have to contend with the fact that the G4 is still stuck at 500 Mhz while the x86 world has not only gone to a gigahertz but Intel has even announced a new 1.13 Ghz chip. (In other words, Intel's p*ssing contest with AMD isn't over yet...) So, why is the PPC we still running so slow and what will happen next?

For over a year now, Motorola, Nokia and Ericsson have been in a battle to seek dominance in the wireless communications market. All competitors are eying the gigantic Chinese market. The Chinese government has decided to use cell phone technology instead of hard wire telephone systems. this would eliminate the need for a massive telephone wiring project and require only cell phones for receivers and a few satellites in place of the wiring system. Motorola wants that market, and they have diverted funding from CPU development into the wireless communications divisions in order to wage the battle at full strength.

Fortunately, that particular battle has settled down for the moment and Motorola has gone back to funding the development of the PPC. According to previous reports, Motorola has plans to release the first production G4+ processors in September. They have already demonstrated a 780 Mhz version several months ago, so now the problem seems more of moving from prototype to production versions. Motorola didn't really want to waste more time developing the current G4 processor since it was originally planned only as an entry stop-gap version to be replaced with the G4+ version within a few months. Unfortunately, the originally planned introduction date of the G4+ has been delayed during the little dominance fight Motorola has waged with Ericsson and Nokia over who will be the Big Frog in the Pond of Wireless Communication.

Apple probably had planned to wait until January 2001 at MacWorld SF to announce the production of dual processor Macs to accompany the rollout of the production version of Mac OS X. However, slow PPC development forced them to release it earlier and with only the current processors. However, Steve Jobs has already hinted in one of his recent interviews that Apple will have several some major product announcements over the next six months. I am very sure that new Macs using the UMA-2 motherboard, G4+ processors starting at around 700 Mhz will appear sometime in that time frame, and if we are all very good little children (and Motorola's Dilbert-esque management doesn't foul up...) we may even see such powerful wonders as dual gigahertz G4+ machines by early 2001.

If Apple does manage to lap ahead of the current slow period by early 2001, even PC Magazine's infamous test rigging will have a problem. The latest example is that one where they ran a PowerMac 500 Mhz (single processor) machine against current high speed x86 boxes, "Apple Power Mac G4 500 MHz" and of course, they predictably had the G4 fail the test.

What they didn't mention was that they chose a series of RAM hog applications and conveniently left the memory allocations to the minimum default ones. Any experienced Mac user knows that all of the apps they chose require enormous amounts of available RAM to run at full speed. It is one of the few advantages that Windows has over Mac OS that Windows uses dynamic memory allocation compared to Mac OS' manual memory settings.

To get an idea how badly such test rigging can cripple a Mac, check out MacAddict's June 2000 Issue (Vol.5, Issue 6) on page 18. In the sidebar is a section called "Unfair Benchmarking 101". When the default memory partitions in PhotoShop are left as is, a 500 Mhz PowerMac runs poorly against a GHz Athlon, but when you adjust the memory partitions correctly, a 500 Mhz PowerMac runs roughly neck and neck with the GHz Athlon. Of course, no PC publications were mentioned, but I have a feeling they were telling us how PC Magazine rigged the test.

Oops... I digress...

Back To The Future...

Over the next several months we are going to see a series of high performance machines appear in rapid succession culminating with some incredibly fast machines by early 2001 that will leave Apple's detractors both in the Mac camp and in the PC cap sitting with egg on their faces.

At the same time, we will see the PIII top out at 1.2 Ghz and with relatively flat performance, be replaced with a new processor that delivers higher clock speed but precious little real performance. This will be while Apple is moving into affordable multi-processing machines that can eat anything fro the PC side alive.

I wonder what new nonsense the nay-sayers will dream up by then?

joecarson@applelust.com