john@prism.TMC.COM (06/21/88)
I've just been reading about the new 386SX chip, or "P9", as it has
been referred to. An article in InfoWorld, 6/20/88 talks about
the price break this chip can bring to new systems designed around it,
and how Compaq, and others are busy designing and working on such systems.
Maybe I'm totally wrong on this, but isn't this the version of the 386 chip
that you can just pop into a 286 socket and blast off at near-386 speed
and performance? And if so, wouldn't the obvious thing that everyone
would like to do be to buy one for their AT? Why build NEW systems
around the thing? If all the chips being produced are going into new
systems, it will be hard or impossible to buy one for the purpose it was
intended for originally -- upgrading from 286.
Again, maybe I'm out in space on this. Would someone who knows please
explain what is going on here? Thanks in advance.
----
JOHN DOWD john@mirror.TMC.COM
{mit-eddie, ihnp4, harvard!wjh12, cca, cbosgd, seismo}!mirror!john
Mirror Systems Cambridge, MA 02140wnp@dcs.UUCP (Wolf N. Paul) (06/21/88)
In article <206900116@prism> john@prism.TMC.COM writes: >I've just been reading about the new 386SX chip, or "P9", as it has >been referred to. An article in InfoWorld, 6/20/88 talks about >the price break this chip can bring to new systems designed around it, >and how Compaq, and others are busy designing and working on such systems. > >Maybe I'm totally wrong on this, but isn't this the version of the 386 chip >that you can just pop into a 286 socket and blast off at near-386 speed >and performance? And if so, wouldn't the obvious thing that everyone >would like to do be to buy one for their AT? Why build NEW systems >around the thing? If all the chips being produced are going into new >systems, it will be hard or impossible to buy one for the purpose it was >intended for originally -- upgrading from 286. Well, that was my original understanding too, but the InfoWorld article you refere to quotes an Intel executive as saying that the chip allows system manufacturers to use all of their components designed for the 286 -- WITH THE EXCEPTION OF THE MOTHERBOARD. So apparently the chip is not simply a drop- in replacement for the 286. -- Wolf N. Paul * 3387 Sam Rayburn Run * Carrollton TX 75007 * (214) 306-9101 UUCP: killer!dcs!wnp ESL: 62832882 DOMAIN: wnp@dcs.UUCP TLX: 910-380-0585 EES PLANO UD
scjones@sdrc.UUCP (Larry Jones) (06/22/88)
In article <206900116@prism>, john@prism.TMC.COM writes about the 386sx: > Maybe I'm totally wrong on this, but isn't this the version of the 386 chip > that you can just pop into a 286 socket and blast off at near-386 speed > and performance? And if so, wouldn't the obvious thing that everyone > would like to do be to buy one for their AT? Why build NEW systems > around the thing? If all the chips being produced are going into new > systems, it will be hard or impossible to buy one for the purpose it was > intended for originally -- upgrading from 286. Would that life were so simple.... No, you can't just pop a 386sx into a 286 socket and run. Intel in their infinite wisdom changed the pinout so you need an adaptor board at the very least (probably a few chips here and there, too). ---- Larry Jones UUCP: ...!sdrc!scjones SDRC AT&T: (513) 576-2070 2000 Eastman Dr. BIX: ltl Milford, OH 45150 "When all else fails, read the directions."
mslater@cup.portal.com (06/22/88)
> 386SX a plug-in replacement for the 286 at near 386 speed.
That is NOT what the 386SX is. First of all, it is NOT pin-compatible
with the 286. It uses 386-style control signals and timing, and needed
more power and ground pins. Intel felt that putting it in a 286-compatible
pinout would constrain its performance, and that the real market was for
new machines, not for upgrades.
You can build (and I'm sure several companies will sell) a piggy-back board
with a 386SX and three PALs, which will plug into a 286 socket. However, this
won't buy you any speed -- it just gets you 386 software compatibility, and
in fact will be a bit slower than a 286 for many applications.
The June issue of Microprocessor Report has a detailed article on the 386SX;
if you'd like a copy, send your US mail address to me and I'll send you one.
An excerpt from this article was posted in comp.arch last week.
Michael Slater, Editor and Publisher, Microprocessor Report
550 California Ave., Suite 320, Palo Alto, CA 94306 415/494-2677
uucp: mslater@cup.portal.com (sun!portal!cup.portal.com!mslater)toma@tekgvs.TEK.COM (Tom Almy) (06/22/88)
In article <206900116@prism> john@prism.TMC.COM writes: > >I've just been reading about the new 386SX chip, or "P9", as it has >been referred to. [...] >Maybe I'm totally wrong on this, but isn't this the version of the 386 chip >that you can just pop into a 286 socket and blast off at near-386 speed >and performance? The word I got (from an Intel employee) is that it is a 16 bit (external) bus version of the 386, but is not pin compatible with the 286. Expect some companies to come out with daughter boards that plug into the 286 socket which have the 386sx, some PALs, and probably a socket for the 387sx. >If all the chips being produced are going into new >systems, it will be hard or impossible to buy one for the purpose it was >intended for originally -- upgrading from 286. If that *was* the original intent. It can also be viewed as a way to build a lower cost (and performance) 80386 system, much like the purpose of using an 8088 instead of an 8086. Tom Almy toma@tekgvs.TEK.COM Usual disclaimers apply
jim@belltec.UUCP (Mr. Jim's Own Logon) (06/22/88)
The 386SX (formerly leaked as the P9) is not a 286 compatible at all.
The internals and the timings are very similar to the 386. It cannot be
added to a 286 socket directly (although a simple daughter board can be
built that will allow this, you'll see these available soon enough). The
386SX is slightly faster than a 386 with the BS16 line tied low because of
some changes to the pipelining, but not much faster. For comperable speeds,
the 386SX will be 55% to 70% as fast as a real 386.
This still raises the question: why design a new machine around the
386SX? A system cost is based on (in order of most $ to least) the memory,
the hard disk, the chassis and power supply, the controllers,monitor, and
keyboard, the support logic, and finally the CPU. So what if you can save
$100 on the CPU, it is a small percentage of the system cost. And you are
going to settle for 60% of the performance? Not me.
Watch what Compaq does, its new 386SX machine will not be a apples to
apples comparison. When you price out the machine do it for an equal 386
machine: 1 meg of RAM, same size hard disk, no cache, same size power supply.
For a end user price of $3000 there should be only $300 or so difference
in parts (and that includes the mark up).
Sorry, I got carried away.
-Jim Wall
Bell Technologies Inc.gas@ecsvax.uncecs.edu (Guerry A. Semones) (06/23/88)
In article <206900116@prism>, john@prism.TMC.COM writes: > > Maybe I'm totally wrong on this, but isn't this the version of the 386 chip > that you can just pop into a 286 socket and blast off at near-386 speed > and performance? And if so, wouldn't the obvious thing that everyone > would like to do be to buy one for their AT? Why build NEW systems > around the thing? If all the chips being produced are going into new > systems, it will be hard or impossible to buy one for the purpose it was > intended for originally -- upgrading from 286. A lot of us have been following the rumors about the 386SX or P9 chip for quite a little while now. If what I have read is true, the P9 is NOT pin-compatible with the 286. If the other rumors I heard are true, then it seems that all you would need is for some smart(?) company to produce a wafer-board that IS pin-compatible with the 286 socket and then mount the P9 on the wafer-board - hence a drop-in replacement for your 286. I'll leave it to the more technically-wise to say whether the 16 mhz P9 will operate in some of the older 6, 8 mhz AT's. -- Guerry A. Semones BITNET: drogo@tucc.BITNET Information Services USENET: gas@ecsvax.UUCP, semones@dukeac.UUCP Duke University My views are despairingly mine only. Talent Identification Program "We ain't gifted, we just work here."
johnl@ima.ISC.COM (John R. Levine) (06/23/88)
In article <5310@ecsvax.uncecs.edu> gas@ecsvax.uncecs.edu (Guerry A. Semones) writes: > I'll leave it to the more technically-wise to say whether the >16 mhz P9 will operate in some of the older 6, 8 mhz AT's. Sure it will. Of course it will only run at 6 or 8 mHz. At a given clock speed on the same bus, a 386 is only a little faster than a 286. The speed you see from a 386 machine is due almost entirely to its high clock rate. A P9 would be nice to let us 286 owners run some decent software like, for instance, 386/ix rather than OS/2. -- John R. Levine, IECC, PO Box 349, Cambridge MA 02238-0349, +1 617 492 3869 { ihnp4 | decvax | cbosgd | harvard | yale }!ima!johnl, Levine@YALE.something Rome fell, Babylon fell, Scarsdale will have its turn. -G. B. Shaw
scottw@ico.ISC.COM (Scott Wiesner) (06/24/88)
in article <234@belltec.UUCP>, jim@belltec.UUCP (Mr. Jim's Own Logon) says: > > 386SX is slightly faster than a 386 with the BS16 line tied low because of > some changes to the pipelining, but not much faster. For comperable speeds, > the 386SX will be 55% to 70% as fast as a real 386. My experience with the Compaq 386SX machine is that for normal unix use, it feels about the same as a 16 Mhz 386 based machine. Here are some dhrystone2 numbers: compile time execute real user sys noreg reg ---- ---- --- ----- --- Compaq 386SX: 43.0 21.6 7.9 2378 2601 Zenith 386: 49.5 15.9 7.2 3159 3425 Compaq 25 Mhz: 32.3 7.8 4.2 6747 7107 With -O: Compaq 386SX: 52.8 31.2 8.7 2825 3105 Zenith 386: 49.7 23.0 7.0 3671 3952 Compaq 25 Mhz: 39.5 11.5 4.3 7716 8130 In fairness to Compaq and Zenith, I should mention that the above machines are all pre-production boxes. Things may have changed slightly. The Zenith we have seems to have a pretty slow disk. The little Compaq is a very nice machine. It's in a smaller than normal (for and AT) chassis, and is very very quiet. For pure cpu stuff (based on the above dhrystones), it looks like it's about 75% of a standard 386 box. For what I was doing with it, (software development) it was fine. The 25 Mhz Compaq is a real screamer, but has a real bottleneck at the disk. All the above machines are running Interactive's 386/ix Unix V.3 system. Scott Wiesner Interactive Systems scottw@ico.isc.com
mslater@cup.portal.com (06/26/88)
> | You can build (and I'm sure several companies will sell) a piggy-back board > | with a 386SX and three PALs, which will plug into a 286 socket. However, this > | won't buy you any speed -- it just gets you 386 software compatibility, and > | in fact will be a bit slower than a 286 for many applications. > That's misleading. Since the AT bus can be run at original speed > while the P9 runs at 16 MHz, I would assume that all piggyback boards > will run at that speed. > More to the point, I question that the 386 will be slower in any case. > The clocks for any instruction seem to be comparable, and the 386 has a > longer pipeline, so it should run slightly faster in linear code with > slow memory. Note I don't claim this for any other conditions. A simple piggyback board will use the clock from the 286 socket, and will thus run not at 16 MHz, but at whatever speed the 286 was running. Thus, I think my original statement is correct. As for 286 vs 386 speed, I've seen a number of benchmarks that show that for 16-bit code (8086 or 286 code) the 286 runs very nearly the same speed, and sometimes slightly faster, than a 386. There are many instructions that require fewer clocks on a 286 than on a 386. The primary advantage of the 386 is that it can fetch two 16-bit words at a time; however, this apparently is not such an advantage, presumably because the pipeline can be kept nearly full with only 16-bit fetches and because at every jump there is a 50/50 chance that the second word fetched must be thrown away. The 32-bit bus of the 386 obviously has a big advantage for programs that manipulate lots of 32-bit data words, but most programs are dealing with bytes or 16-bit words. If anyone else has any insights on 286 vs 386 performance, I'd appreciate any clarifications. The data I've seen indicates that nearly all the performance gain of 386 systems over 286 systems, when running DOS or OS/2, is simply due to the faster clock, and that the 32-bit bus is of little help. Michael Slater, Microprocessor Report 415/494-2677 mslater@cup.portal.com sun!portal!cup.portal.com!mslater
pavlov@hscfvax.harvard.edu (G.Pavlov) (06/27/88)
In article <234@belltec.UUCP>, jim@belltec.UUCP (Mr. Jim's Own Logon) writes: > > The 386SX (formerly leaked as the P9) is not a 286 compatible at all. > > This still raises the question: why design a new machine around the > 386SX? A system cost is based on (in order of most $ to least) the memory, > the hard disk, the chassis and power supply, the controllers,monitor, and > keyboard, the support logic, and finally the CPU. So what if you can save > $100 on the CPU, it is a small percentage of the system cost. And you are > going to settle for 60% of the performance? Not me. > The difference is that you begin with an 80286 machine design (even tho the cpu chips are not pin-compatible) rather than with an 80386 machine. greg pavlov, fstrf, amherst, ny
davidsen@steinmetz.ge.com (William E. Davidsen Jr) (06/27/88)
In article <6859@cup.portal.com> mslater@cup.portal.com writes: | As for 286 vs 386 speed, I've seen a number of benchmarks that show that for | 16-bit code (8086 or 286 code) the 286 runs very nearly the same speed, and | sometimes slightly faster, than a 386. There are many instructions that | require fewer clocks on a 286 than on a 386. I checked 25 instructions at random in the manuals without finding these instructions. I can't say that you're incorrect since I haven't check every one, but I really would like to know which instructions are slower... -- bill davidsen (wedu@ge-crd.arpa) {uunet | philabs | seismo}!steinmetz!crdos1!davidsen "Stupidity, like virtue, is its own reward" -me
dave@micropen (David F. Carlson) (06/29/88)
In article <6859@cup.portal.com>, mslater@cup.portal.com writes: > > As for 286 vs 386 speed, I've seen a number of benchmarks that show that for > 16-bit code (8086 or 286 code) the 286 runs very nearly the same speed, and > sometimes slightly faster, than a 386. There are many instructions that > require fewer clocks on a 286 than on a 386. The primary advantage of the > 386 is that it can fetch two 16-bit words at a time; however, this apparently > is not such an advantage, presumably because the pipeline can be kept nearly > full with only 16-bit fetches and because at every jump there is a 50/50 > chance that the second word fetched must be thrown away. The 32-bit bus of > the 386 obviously has a big advantage for programs that manipulate lots of > 32-bit data words, but most programs are dealing with bytes or 16-bit words. Most programs deal with word size for the machine. Most useful programs I have used have 32 bit ints and floats. I can run "useful" programs on the 386SX, whereas I lost much hair running most anything "useful" on the 80286. > > If anyone else has any insights on 286 vs 386 performance, I'd appreciate any > clarifications. The data I've seen indicates that nearly all the performance > gain of 386 systems over 286 systems, when running DOS or OS/2, is simply due > to the faster clock, and that the 32-bit bus is of little help. > Michael Slater, Microprocessor Report 415/494-2677 > mslater@cup.portal.com sun!portal!cup.portal.com!mslater The stipulation of DOS or OS/2 is probably correct. The 386 requires more time to do segment loads due to 2 levels of tables indirection whereas the 286 has just a GDT, LDT, etc. The big point of the 386SX is that you don't have to close the door on Xenix, UNIX, OS/2-386, etc. just because the 32bit memory scene is such a mess now. For most single user apps (workstation model), a 16 bit data bus is not as much of a bottleneck as is the 250Kbps of the WD MFM disk controller. Speed isn't the big question but future compatibility is. -- David F. Carlson, Micropen, Inc. ...!{ames|harvard|rutgers|topaz|...}!rochester!ur-valhalla!micropen!dave "The faster I go, the behinder I get." --Lewis Carroll
james@bigtex.uucp (James Van Artsdalen) (07/02/88)
IN article <11390@steinmetz.ge.com>, davidsen@crdos1.UUCP (bill davidsen) wrote: > In article <6859@cup.portal.com> mslater@cup.portal.com writes: [ some 286 instructions are faster than 386 versions ] > I checked 25 instructions at random in the manuals without finding > these instructions. I can't say that you're incorrect since I haven't > check every one, but I really would like to know which instructions are > slower... You obviously didn't start at the front of the instruction set section. :-) The first four instructions (alphabetically) are faster on the 286. However, the fifth (ADC) shows that the 386 *is* faster when dealing with immediate operands. I would hazard a guess this is really a function of the 32-bit data width and not the ALU itself, and that the 386SX will not be faster than the 286 in this case. All of these timings should be taken with a grain of salt: the manuals tend to gloss over or ignore a lot of strangeness in processor activity, particularly with regards to branches, and the only way to really know these things is to get out a hardware probe. Even then, the 68030 is hard to measure: several instructions can be running at once (as far as counting cycles goes), and many instructions appear to execute in zero time due the the caches & parallel execution. -- James R. Van Artsdalen ...!ut-sally!utastro!bigtex!james "Live Free or Die" Home: 512-346-2444 Work: 328-0282; 110 Wild Basin Rd. Ste #230, Austin TX 78746