People talk about the great history of the Pentium, and what a good job Intel has done at keeping up with RISC. I think they need to study a little more processor history. Intel has done a great job with hype -- but the actually processors themselves have not done so well. Intel was good at pre-announcing things to take the wind out of the sails of others (and they succeeded). Intel consistantly promised to deliver tommorrow, what others were delivering today, only "better". The truth was that Intel usually failed to deliver on promises (Intel has a rich history of doing so, that run's back for decades). This is the first article in a series, that will look at what really happened. This article only looks at the early Pentiums vs the early PowerPC's.
In fact, there were many problems and delays, and the PowerPC's seem to be consistantly stomping the Pentiums -- if you look at the facts, and not the hype. Instead of just looking at what Intel was promising, and comparing that to what AIM and others delivered, lelet's look at Apple's to Apple's (so to speak) and compare what was delivered, and when.
In this corner...
In 1991 the ACE (Advance Computing Environment) consortium was created. This would allow for many vendors to finally make a PC that was better than the current PC. It could use the MIPS R3000 processor (or follow-ons), or the 386sx. It would run many different Operating Systems including; OSF's Unix, SCO's ODT and Microsoft's WindowsNT (which was supposed to be out that year). Many companies happily joined this consortium, primarily to get out from under Intel's grinding thumb, and to create a better PC -- the companies included; DEC, Compaq, Silicon Graphics, SCO, and many others. It was going to be a beautiful thing.
Intel was also scared because the R3000 was outperforming their 486. More than that, in late '91, Apple, IBM and Motorola announced their AIM alliance, and promised the PowerPC in just under a year (a very aggressive schedule), and with performance that gave Intel great concern -- and consider the performance of the RS/6000 (Power Architecture) that IBM was delivering, and the PowerPC was based on, there was plenty to be concerned about.
Intel was scared, and so reacted with hype. They promised that the Pentium would be out in a year -- they would be delivered in late '91 or early '92, with volume production to follow in fall of '92. Intel promised that 66MHz would be the starting speed for the Pentium (with higher speeds as well), it would be 2 times faster than 486 (int) and 4-10 times faster than 486 (fp), and a far better performer than the MIPS R3000. It would slice bread, and mow your lawn, and complete an infinite loop in just 5 seconds. All of Intel's usual garbage promises. This was all FUD (Fear Uncertainty and Doubt) to defraud the public, and break the ACE Consortium, and scare people away from the AIM Consortium.
For more on Intel's sleasy behavior in the ACE battle, read History of Ace.
So lets look at the actual deliveries, and how Intel did compared to their hype.
Early Pentiums (P5, P54C and P24T)
The Pentium (P5) was promised by late '91 -- it sampled late by basically a year (fall of '92). 1 year late on a one year schedule is pretty bad -- but it got much worse. It was supposed to ship in volume in "fall of '92". Intel claimed that they were shpping in volume in middle of '93, but all the vendors couldn't actually get the chips until the end of '93 or early '94. That is well over a year late (and somewhat fradulent of Intel) -- and it gets worse still.
Finally in '94 things seemed to smooth out, and production started to go, and people could actually get the processors Intel was claiming they could. Actually though, P5-66MHz chips were aweful scarce (people had to go with the slower 60MHz flavors). Remember, the 66MHz part was supposed to be the slowest flavor -- not the top end. And as for performance, it was a big improvement over the 486, but the Pentium fell far short of all claims (and it was not competitive with the MIPS chips). It also went way over heat/power budget, required a special little fan, and would burn up when they failed. But the media and masses were buying in that everything was great and on schedule.
Then a major problem was made a dozen times worse. In early '94, Intel learned of a floating point that would seriously corrupt users data. Rather than inform users and do some spin control (and replace some units), Intel tried to hide the problem. This allowed for far more bad chips to get shipped to customers. Many months later, at the end of 1994, the story got out -- the Pentiums were flawed at floating point math (the one feature that made the Pentium so superior to the 486). Intel tmade matters worse by trying to pretend that it was no big deal -- this angered people even more. Intel claimed it was only a problem at high degrees of accurace (very insignificant digits), and so on -- the reality was that it could completely destroy results and cause very significant errors (even if it was statistically rare to run into the problem). Intel cleared up the mess, when they finally agreed to replace people's processors (which they should have done all along). A year later they actually got around to doing it -- and it was already the end of '95.
The reality of the Pentium (P5) was that you couldn't get a WORKING Pentium until 3 years AFTER Intels projected date. Remember, that during all this time, the Pentium really didn't perform as was initially advertised (by probably about half). To get the claimed perfomance numbers, users really had to wait for the P54C -- the second generation of Pentium, and 100MHz part. That Pentium would finally delivered the speeds (in MHz and Performance) of the first Pentiums promised.
With the P55C, Intel again tried to pull the wool over some people's eyes -- and they said that they were shipping quantity from the moment that the chips were introduced (in early '94). The truth was that these chips were very limited, and you could only get the 75's and 90's, and not the 100's. (Again with the speed fraud). To make matters more fraudulent, you could get the Processors in '94, but the I/O chips that you needed to make Systems work -- those weren't delivered until early '95. So in this reality (not Intel's) you really couldn't find P54C Systems until late in '95. Again, Intel delivered on '92's promises 3 years (and 2 generations) late.
But Intel had also claimed that the PC's (and 486's) would be "upgradable" to Pentiums (and implied this would happen about the same time as the release of the first Pentiums). You couldn't upgrade your machine with the P5 or the P54C, at least not unless you replaced your entire motherboard -- twice (the P5 and P54C were completely incompatible with each other, as well as the 486). Actually, you not only had to replace the motherboard, but you probably had to change other things like I/O cards, memory, and other devices -- or basically replace most of your computer. So if you were an early adopter of the P5, you not only had to replace your System once, but you got punished when the P54C came out, and had to replace your System again -- to get the performance that what was promised. (Imagine that expense). Some upgrade. But Intel kept promising the real upgrade, the P24T Overdrive, for all those 486 machines -- any year now.
The problem is that Intel had no motivation on making the P24T -- they made more money off of forced upgrades to the P5 or P55C (since people had to buy new Intel support chips as well as more expensive versions of the processor). Intel quickly backed down from their promises of a timely release of that chip, claimed that the P24T would be out in late '93. (ACE had been crushed, FUD had worked, and Intel was no longer in any hurry). So despite having a "socket" for the processor, you couldn't use it for much until the P24T shipped. All those users who bought "Overdrive Capable" 486 Systems had to wait, and wait, and wait.
When did the P24T ship? Intel claimed they sampled and shipped the Processor in early '95. The reality was that you couldn't get the chip in any volume until the very end of '95. (I think Intel thinks that November is "early" in the year).
With all three flavors of early Pentiums, the promise was "early '92", the truth was closer to early '96. All early adopters were punished with chips that over-heated and had bugs, and they had to replace their motheboard multiple times, and they got what they deserved for believing Intel's hype (1).
(1) This was not the first time Intel did this either. I hope to document all their broken promises with the 286, 386, and 486 time frames. But it goes all the way back to 1971 and Intels first claims for the 4004 Microprocessor, which they claimed was, "A computer on a chip". It actually took about 28 chips to achieve what Intel claimed -- but the truth has never stood in the way of their marketing.
Well all marketing people lie, right? So how does this compare to AIM and the PowerPC?
Early PowerPC's (601, 603, 604).
AIM (Apple, IBM and Motorola) were far better able to hit their schedules. In late '91 they gave a prediction that they would have the processors out in late '92, with Systems available later (implication from Apple was end of '94 to early '95). AIM made claims of 66 MHz parts and faster, and that they would quickly follow the 601 with the second generation of PowerPC's (the 603 and 604's).
Much of the press poo-poo'ed the idea, and implied that they would never make their agressive deadlines. But they got very silent a year later when IBM nailed the first milestone and showed the 601 in late '92. Of course, the AIM camp basically hit all their Performance and Power goals the first time, and there were no floating point bugs from "rushing". Not only that, but Apple was seeding PowerMac's (in a very rough state) to developers soon after. IBM surprised people again by shipping Systems (and volume) in late '93, before they even had their OS/2 for PPC's ready (that OS took years longer than IBM expected and was finally killed) -- they shipped the Systems with AIX (Unix) instead.
Microsoft was already withdrawing its claims for WindowsNT, which was supposed to be the cross-platform and RISC version of Windows. Once people started buying into Microsofts promises, they changed its goals from being on all RISC platforms, to only being on Alpha's, and only as long as DEC (now Compaq) pays for it to be there. This effected Motorola, and IBM, who were planning on supporting WinNT of PowerPC. Everyone that gets in bed with Microsoft, gets out feeling screwed.
Apple soon came out with PowerMacs in early '94, and they surprised people with an 80MHz version of the PowerPC (when they had only promised 66MHz) (2). Then as a further surprise, IBM die shrinks the 601 so that it used less power. This was called the 601+, but got named the 601v. And Apple starts selling 110MHz PowerPC's, just 6 months after the first PowerMac Systems -- (this was nearly a year before you could actually get 100MHz P54C's).
(2) Intel quickly responded by saying that the 100Mhz P54C was shipping, but remember, you couldn't get that chip in an actual System for at least a year and half later.
If all that wasn't enough, IBM and Motorola begin volume shipping 603's and 604's, right on targe (before the end of '94), and by early/mid '95, Apple has PowerMacs and Performa's based on these processors. Intel wasn't coming close to keeping up, so they quickly started pre-selling the next generation. Intel wasn't actually volume shipping the P54C's and P24T's yet, but they were responding to the PowerPC's superiority with the only thing they could -- hyping the P6 (later called the PentiumPro) processors. Sure Intel had failed on everything promised for this generation -- but the next one would be better. That has been Intel's promise for the last 5 generations, "we'll fix it next time".
During '94, while AIM was delivering a 1...2...3 punch of 3 excellent processors, delivered on time, and meeting claims, Intel announces that it will spend $150 million in marketing. This was the start of the flying Pentium ads (that were produced on PowerMacs), and the big battle to keep people from realizing the truth.
There is far more to the story than just this little article. But I think you can start to see the picture (and the trend). There were waves of promises by Intel on what they would ship -- and the press and the masses bought it. Almost none were good at holding them accountable for their claims, and everyone was willing to play Intel's game. Intel was claiming the P5 was out before the PowerPC's, and it was if you didn't need a working one. They claimed that they had 100MHz first, you just couldn't use one. Magazines would compare prototype Pentium PC's, with shipping Apple PowerMacs, and still end with a note that while the PowerPC's might be a little faster than THESE machines (the machines that wouldn't be available to the public for months), that the next Pentiums would be out soon too. They were often selling two generations ahead of what was actually available -- it was quite frustrating to watch.
At least things haven't changed much. I still see the same MO (Modus Operandi), with reporters still comparing Intel's hype, to what Apple and others are delivering.
Shipping schedules, and MHz certainly aren't the whole story either. In other articles I will go into Cost Comparisons (of chips, and Systems), Performance (claimed and actual), and Power Consumption. All using actual times (when models were shipping -- not just Intel's claims). This makes the history far more accurate than the claims -- and tells a far different story from what most reporters and IS/IT types seem to remember. I also hope to further the series, and compare the later G2's and G3's against the P6 (Pentium Pro) and P55C. And end the series comparing likely results and deliveries of the future, with future P6's (PentiumII's) and the P7's (Merceds) versus the G4's and G2000's.