I agree with Dale, that the 29K _was_ a general-purpose architecture
Saturn V wrote:
"Did not get time to react to your statement. AMD developed the 29000 several years after the microcontroller market had been defined. And developing a new microcontroller is not pioneering a new general purpose processor. You do not have to worry about standards and infra structure headaches. It takes a much smaller investment."
I just saw Dale's point, and I agree with it. The 29K was not planned to just be a laser printer engine...AMD hoped it would be a major architecture. As someone, maybe Dale, said a few days ago, there was the YARC add-in board for the Macintosh...I had forgotten about this, but it was once very hot stuff for "kicking the Mac up a few notches."
Interestingly, the 29K's main competitor was the 960, written about here by one of the Intel folks who worked on it. It ended up as a microcontroller, albeit a powerful one, but it started out as part of the 8800/432/Gemini/Biin product series. One of the design variants even use tag bits, like the LISP machines of the day.
"Microcontrollers have also been the graveyard of several failed general purpose processors! The National 32000 and the MIPS come to mind right now."
Better double up on your gingko bilobas tonight, Saturn! :-)
The National 32K was also, like the 29K and Intel chip above, intended as a fully-capable architecture. In fact, it was very regular in architecture, and many people admired it a lot. Sequent, the Intel spin-off, used the 32K in their first general purpose machine in the mid-80s. (There were several variants, but the 32032 was the best known. A 32016 also existed, and some later variants.)
As for MIPS, that was most emphatically not originally planned to be a microcontroller! The MIPS chips were designed for high-performance workstation use. SGI used them and eventually bought MIPS. (IIRC, Tandem also used the MIPS chip. Several other companies did as well.)
That the MIPS architecture ended its life as the CPU in game machines (NeoGeo, I think was the first, then machines like the Nintendo N64 and so on) does not mean it was designed for this.
"The same team which created the excellent 29000, was responsible for the disasters on the K5. They made such major goofs, that the entire team was fired en masse after the NextGen acquisition. "
Yes, but my point was that the 29K counts as evidence that AMD has in fact attempted in the past to do innovative work. It was an effort not unlike the National 32K or the MIPS or either the 960 or the 860. Dale discussed the 860 in his message.
That all of these processor families either failed or were relegated to cul-de-sacs does not mean they were not innovative efforts.
(I don't know the $$$ budgeted to these various efforts, but my hunch is that the 960 cost the most, cumulatively, then the MIPS, then the 32K, then the 29K, and the 860 had the lowest budget of the bunch.)
"My point is that a new general purpose computing architecture takes an order of magnitude greater effort than a microcontroller."
Alas, Saturn, you have retrospectively dismissed the 32K and 29K and MIPS, not to mention presumably the 960 and 860, as mere microcontrollers. By this dismissal, all that is really left from the 1980s would be the SPARC and the early versions of the IBM RIOS chip (which evolved into the PowerPC series).
(What of Motorola? Their 68K follow-ons were fairly successful in early Sun machines and, of course, the Mac. The Amiga, too. And other workstations (Apollo, etc.). Not a major new architecture, but neither did Intel have a major new architecture/ISP until the Itanium. Motorola _did_ attempt a new architecture, the 88K. It never really appeared and is long gone. Had it been produced and used as a microcontroller, it could be added to the list of chips above.)
So, I think it off-base to characterize the 29K, 32K, MIPS, 960, and 860 architectures as "microcontrollers." That they didn't catch on as the Next Big Thing doesn't change how they should be viewed.
An interesting thread, though.
--Tim May |
