Dojo (HowTo)







  Easter Eggs




  Martial Arts

603's versus 601's

By:David K. Every
©Copyright 1999

The 603 unfairly got pinned with a bad reputation early on. People assumed that the 603 was an inferior chip to the 601 because some magazines erred in how they explained the system performance differences. The history of this is an interesting look at engineering tradeoffs and how a little misinformation can go a long way.


The first generation of 601's were built, and they had some legacy instructions to make them compatible with IBM's older POWER RISC chips. This made the 601's a little less efficient than they could have been (1). The 603's were meant to alleviate this minor design flaw, and reduce size, heat, cost and increase performance. It was another engineering success, and a marketing failure.

(1) Remember, this little inefficiency was nothing compared to the far bigger inefficiency in the Pentium and x86 instruction set that Intel has by using an anachronistic design, that has its roots planted nearly 30 years in the past (1971). But it was enough that AIM (the PowerPC consortium) felt that they could do better. <Sigh> If only Intel had the same standards.

The strategy was that they were going to use a split cache design (instead of a more complex unified cache), remove the "unused" instructions, and reduce the size of the 603 -- which would reduce the cost and the power, and be able to manufacturer the processor for much cheaper, and at higher speeds. The simpler processor had a slight performance penalty (per MHz) but the chips could be made at higher speeds -- which would more than make up for it. In fact, the cost savings overall (for the cheaper part) could be used to add an L2 cache or other tricks to increase the performance even further.

Chip wise, it worked beautifully. The 603 was anywhere from 4%-10% slower in integer, and about 8%-24% slower in floating point than a 601 -- but neither is enough to really matter to most users. It would require a stopwatch to notice any performance differences. The cost of the 603 was so much less (compared to the 601) that companies could offer more MHz (in the 603 over the 601) for the same cost. In other words, an 80 MHz or 100 MHz 603 cost less than a 66 MHz 601, and offered more performance. It also required less power, which meant that it could be put in portables, and it could require a smaller (cheaper) Power Supply (to lower System costs further). Big design wins.

Apples, Performas and Gotchas

Apple was producing 601's in their business line, and so Apple used the first 603's in portables and low-end consumer desktops. They were going to then push the high-end desktop line (business line) to the superior 604 chip. Everything was staged for success -- but marketing defeat was snatched from the jaws of engineering victory.

Apple had low-end machines called Performa's (like the 520's and 620's), which were low-end desktop units. Apple wanted to keep these machines as good (inexpensive) home and education machines -- with enough power to be usable, but not so much performance that they would pirate the higher end desktop line or increase costs too much. They also wanted to update the design quickly, and get on the market. So Apple basically stuffed the 603 chip into the design that was supposed to support the 68040. This had a smaller bus than the PowerPC could handle (which didn't perform as well as the other PowerPC's of the time), and Apple didn't put in a lot of extra expense into the L2 cache (to keep the cost down), and other design decisions like this. These were econo-boxes.

Again, the project was a success. The 6200's were far faster than the 620's and 630's and previous non-RISC Performas (if you were running new native apps). Yet there were many minor marketing problems that compounded into some serious troubles. Everything that could be measured wrong, and could go wrong, did go wrong. Apple got far more shit than they deserved -- all because of erroneous expectations.

Magazines and Magnification

The first generation of PowerPC Performa's (5200/75, 6200/75's) were built around the 603, and were designed for economy. It worked, they were inexpensive for PowerPC's -- they performed well compared to Pentiums. They were cheap, and usable.

I don't count the Performa 6100 as a Performa. It was just the desktop PowerMac 6100 with a different sticker. It had nothing in common with the other Performas as far as design.

The second generation of Performas (5200/6200's from 100-120 MHz) were built much more for speed -- they used the 603e's (with larger L1 caches). Not only was the 603e a bit faster than the 601 or 603, but Apple also increased the bus size, and added an L2 cache and some other tweaks to make these second generation systems far faster.

Then the third generation of Performa's (5300/6300's from 150-180 MHz ) came out. The concept of slow 603's should have died before this, but misinformation dies slowly -- people still mistakenly credited the newer Performas with the older Performas performance. Faster bus, PCI slot, and so on. These were really fast machines.

Apple bumped them up again and stuffed them in a Tower design, and put in a video accelerator. The 5400 and 6400's were great (180-225 MHz).

The final bump was to the 5500 and 6500's, which got an even better video accelerator, and 250 - 300 MHz.

Then AIM (Apple, IBM, Motorola), revised the 603 design again, and made it into the G3.

One of the biggest problems was that the magazines didn't compare the 6200's to the older Quadra and Perform 630's (68000 Performa's) for performance (with native Apps) -- they compared these low end machines to the high-end desktop PowerPC's. They expected the econo-box to perform as well as the sports-cars. The design compromises guaranteed that the Performas were not going to hold up. The slower and smaller bus, the nonexistent or smaller cache, all worked to make the 603's underperform compared to the 6100, 7100 and 8100's, let alone the 7500, 8500, and 9500's (that were soon on the market). But that didn't mean that they were bad machines -- they were fine, and more power than most home users needed at that time. They were just not going to compete with the top end. Most of the performance issues had nothing to do with the chip itself, but the inexpensive systems that they became a part of -- yet the reviews labeled the chip as "inferior".

Magazines also compared the 603 to the 601 in MHz to MHz -- and came to the obvious conclusion that the 603 was slower. It was. But you were buying a 75 MHz 603 for the cost of about a 50 MHz 601. They weren't normalizing for cost. So buyers got the impression that they were getting ripped off -- and in some ways, it was true. The 75 MHz 603 was not competitive with an 80 MHz 601 as users had hoped -- but that was never had that intent. The 603 was certainly competitive with (or faster than) the Pentiums at the same speed -- so it wasn't a case of fraud, or users getting "ripped-off", just a case of over-expectations.

The one true performance surprise was that the 603's smaller cache, and the Performa's lack of an L2 cache. This wasn't supposed to be a big deal -- but it caused a far bigger preformance hit with the 68000 emulator (used to run older Mac Apps) than anyone expected. This meant that running older (emulated) Apps, offered poor performance. But the emulated Apps were getting replaced quickly with native Apps -- so even that performance issue was short term. Later Performas fixed the emulator performance problem by using the 603e (which had a larger L1 cache), and by adding an L2 cache, and a wider bus, which all contributed dramatically to performance (especially of emulated Apps).

The point being that the magazines magnified the problems by miscomparing. It wasn't malicious, I think the reviewers really expected the Performa 5200/75's and 6200/75's (which offered nearly the same native performance as the 6100/60's and 7100/66's, but at a fraction the price), to dramatically outperform those machines -- and to dramatically outperform the 680x0 based Performa's (using emulation). It did neither. So they said that the Performas were junk. While they certainly didn't compete with the desktops (8100/100 or 7500/8500's), the Performas weren't junk either.

I wouldn't have had a problem selling a 603-based Performa to my grandmother, or to a school for education, or to most homes for puttering and Internet -- but they were not going to be good machines for geeks who wanted real power demanding things like heavy Photoshop use or for doing Video editing. So in many cases, people were sold the wrong machine by the stores and salespeople, and they blamed Apple for it. But that problem still exists today for all companies (not just Apple).

To be fair, Apple did have some quality problems with many of the early Performas (mostly the second generation Performa's) -- which compounded the bad repulation of the first ones. The reason was that Apple had other manufacturers making many of the parts (chips), and some of those OEM's were interpreting the specs creatively (or Apple was not specific enough of the specs). The results were that a combo-chip (which included the cache) was buggy, and caused a lot of failures. Apple offered upgrades and utilities to identify the crashy machines, and was pretty good about it -- once they figured out what it was. But bad reputations are far harder to lose than to gain. Still, that whole second generation of Performa's had more problems than any Macs I've dealt with, and really hurt Apple's quality perception -- even if only one out of ten was bad, that was enough pissed-off people to hurt.


To prove that the 603 (and successors) wasn't a bad chip, one only need to look at some later systems that had the same chips. The 6500/300's were very fast boxes (and the fastest consumer box of its time). The PowerBook 3400 603e/240 was the fastest Portable of its era, by a good margin. Lastly, the 603 design was the foundation of the G3's (750's) that are "up to twice as fast" as Pentiums. So it was never the 603 that was the problem -- and most of the System problems were blown out of proportions.

Created: 05/26/97
Updated: 11/09/02

Top of page

Top of Section