mslater@cup.portal.com (06/16/88)
The P9 goes public!
Intel's long-rumored P9 processor is now officially the 80386SX. It is
fully software-compatible with the 386 but has a 16-bit data bus and a 24-
bit address bus. Bus timing is very similar to the 386. The 386SX is
available only in a 16-MHz grade and Intel claims to have no plans to
introduce faster versions. The chip is already in production and Compaq is
expected to announce the first 386SX-based PC on June 20. The 386SX is pin-
compatible with the 80376, a version without paging and backward-
compatibility functions that was introduced in April.
Intel claims that 386SX-based systems will have 85 to 90% of 386
performance running 16-bit software, and 70% running 32-bit software,
assuming equal clock and memory speeds. Since the 386 is available at
faster clock speeds (20, 25, and eventually 33 MHz), high-end 386-based
systems will still be significantly faster than 386SX-based systems. The
days of 16-MHz 386 systems appear to be numbered, however.
Numeric coprocessor support is provided by the 387SX, which is identical to
a regular 387 except for the narrower data bus. The 387SX at 16 MHz is
claimed to provide three times the performance of a 287 at 10 MHz, so SX-
based machines will have a significant edge over 286-based machines in
floating-point performance. Price for the 387SX is $393, vs $441 for the
standard 387.
------------------------------------------------------
Part No. Clock Speed Price (100-pc)
286 12.5 $138
286 (Harris) 16 $160
386SX 16 $219
386 16 $299
386 20 $484
------------------------------------------------------
The table above shows the relative pricing of the 286, 386SX, and 386. As
you would expect, the SX is priced somewhat lower than the regular 386 and
higher than the 286. The price of the 16-MHz 286 is sufficiently close to
the 386SX price that the SX will take away much of the high-end 286 market.
Early P9 rumors often stated that chip would be pin-compatible with the
286. It is not, however, because this would have constrained the design too
severely. The 386SX uses a 386-like bus structure and timing; designing the
chip to be pin-compatible with the 286 would have meant sticking to 286
timing, and therefore sacrificing performance.
(Incidentally, Intel does not formally acknowledge their internal project
numbers, so the 386SX officially has nothing to do with the P9. But with
the addition of clock speed and packaging, Microprocessor Report's
September 1987 article on the P9 would have fully described the 386SX.)
Naming for the mass market
At first glance, it seems odd that Intel chose to call the chip the 386SX,
a designation that sounds more like a Japanese sports car than a
microprocessor chip. However, Intel wants to be sure that personal computer
buyers recognize that the 386SX is fully compatible with the 386. A name
such as the 80388 may have meant something to us techies, but it would have
made life more difficulty for computer salespeople who would have had to
convince customers that a 388 could run all 386 software; now they can just
point to the CPU and say "See, it says 386." This is perhaps the first time
a new microprocessor has been introduced with such concern for the
perceptions of the end user.
The naming scheme also fits in with the approach Intel has taken with some
other product families. The numeric portion specifies the basic
architecture, and the alphabetic suffix specifies details of the particular
implementation. For example, the 80960 architecture currently has three
implementations, with the suffixes KA, KB, and MC. Motorola uses a similar
scheme for their 6804 and 6805 single-chip microcomputers.
(Intel insists, by the way, that SX doesn't stand for anything, it just
sounded good. In the case of the 80960, the K-series parts are commercial,
and the M-series is military; A, B, and C refer to successively more
sophisticated versions.)
Pricing for the Marketplace
Intel claims that the design of the 386SX is optimized for low-cost, high-
yield manufacturing, and that this emphasis means that the chip is less
likely to be made in higher speed versions. The heat dissipation in the
plastic package is also a potential problem at higher speeds. While these
issues are no doubt real to some degree, there is another motive for
keeping the 386SX at 16 MHz<F14><197><F255>it protects the market for the
more expensive 386. The 386SX allows Intel to reach into more cost-
sensitive markets with the 386 architecture without lowering the price of
the high-end devices. While the plastic package and manufacturing
optimizations certainly account for some of the price difference between
the regular 386 and the SX, the pricing of the 386SX has more to do with
positioning than with costs.
So despite its lack of technical innovation, the 386SX is potentially one
of the most important new processors Intel has ever produced. Look for
numerous 386SX-based machines by Fall COMDEX.
(This article is excerpted from the June issue of Microprocessor Report,
which includes additional details.)
Michael Slater, Editor and Publisher, Microprocessor Report
550 California Avenue, Suite 320, Palo Alto, CA 94306 415/494-2677
Bix: mslater uucp: mslater@cup.portal.com (sun!portal!cup.portal.com!mslater)
david@sun.uucp (David DiGiacomo) (06/17/88)
In article <6560@cup.portal.com> mslater@cup.portal.com writes: >At first glance, it seems odd that Intel chose to call the chip the 386SX, >a designation that sounds more like a Japanese sports car than a >microprocessor chip. Not at all! Intel has a long history of giving the same name to different things (iRMX-xx), giving different names (and prices) to the same thing (8275, 8276]), arbitrarily renaming second source parts (8274), giving new products names similar to successful but completely unrelated products (e.g. 82586, 82786, 80960), and of course creative yet truly stupid new concepts in naming (iAPX!).
neighorn@qiclab.UUCP (Steve Neighorn) (06/20/88)
In article <56872@sun.uucp> david@sun.uucp (David DiGiacomo) writes:
:Not at all! Intel has a long history of giving the same name to different
:things (iRMX-xx), giving different names (and prices) to the same thing
:(8275, 8276]), arbitrarily renaming second source parts (8274), giving new
:products names similar to successful but completely unrelated products
:(e.g. 82586, 82786, 80960), and of course creative yet truly stupid new
:concepts in naming (iAPX!).
Thank you for adding to the technical content of the architecture group.
Poking fun at the Intel nomenclature is truly a relevant subject that
needs further discussion. Besides, we all need a little extra SPARC in
our daily news reading. :-)
Seriously speaking, it does take a road map to keep up with all those
seemingly arbitrary identifications being given to new processors. I
don't think Intel is any more guilty of this than any other manufacturer.
--
Steven C. Neighorn !tektronix!{psu-cs,reed,ogcvax}!qiclab!neighorn
Portland Public Schools "Where we train young Star Fighters to defend the
(503) 249-2000 ext 337 frontier against Xur and the Ko-dan Armada"jim@belltec.UUCP (Mr. Jim's Own Logon) (06/20/88)
In article <6560@cup.portal.com>, mslater@cup.portal.com writes: > The P9 goes public! > > Intel's long-rumored P9 processor is now officially the 80386SX. It is > fully software-compatible with the 386 but has a 16-bit data bus and a 24- > bit address bus. Bus timing is very similar to the 386. The 386SX is.... Oh, I should have left in the part about Compaq coming out with a machine based on the 386SX. The question has to be raised: why would anyone want to base an entire new machine on this chip? The answer is: (drum roll faintly growing in the backround) COST!!. But is it worth it? Cost breakdown for a computer (all approximations for a reasonably powerful machine): Hard disk $200, 1 meg of memory $300, Chassis and cables $100, keyboard and monitor $75, disk and video controllers $75, Misc. required logic and PC board $175, and the cost of the CPU. So on a machine that costs $1100 to build, the CPU is a small percentage of that. Sure, you can still save $100 by using the 386SX, but is it worth having only the 55% to 70% performance (reasonable benchmarks are closer to 55% than to 70%)? I think not. If you want the 386 compatibility for Sys V.3 or 3.1 or Windows, or X Windows or Sun NeWs, than the performance is the key. If you are running Dos, stay with a fast 286 and really save on cost. An end user difference of $300 does not justify an entire new machine. But perhaps a daughter board to replace a 286 so you can run V.3 ..... -Jim Wall Bell Technologies Inc. "My ideas are just that, and no more" Which applies equally to a fig newton.
scott@hpcvca.HP.COM (Scott Linn) (06/24/88)
Re: Why an 80386SX? The cost advantages are not simply based on a "cheaper CPU": 1) Those companies with large inventories of 80286 support chips can use them directly with the 80386SX. 2) The board area required for an 80386SX is much smaller. The 80386 package is about 1/3 the size and can be surface-mounted. 3) External data paths are 16 bits, which makes for cheaper PC boards and easier layout.