lindsay@gandalf.cs.cmu.edu (Donald Lindsay) (10/23/90)
I've heard some comments lately about how hard it will be for 386 clone makers to get a debugged product to market. An example is Nick Tredennick's article in Microprocessor Report, which implies that some of the clone makers haven't any clues to rub together. But, wait a second here. There are three things they have to get right: pinout-related issues: the 286 emulation: the native mode. (They don't have to get the cycle counts "right", do they?) I would have nominated the 286 emulation as the killer, since the native mode is mostly "more of the same" with paging. But, AMD is mass-producing a perfectly good 286, and the 386's microcode can be reverse-engineered. So, what is to stop AMD (at least) from getting its 386 debugged and out? Is the pessimism only justified for the smaller houses, or have I mis-estimated where the problems lie? -- Don D.C.Lindsay
mslater@cup.portal.com (Michael Z Slater) (10/24/90)
>I've heard some comments lately about how hard it will be for 386 >clone makers to get a debugged product to market. An example is Nick >Tredennick's article in Microprocessor Report, which implies that >some of the clone makers haven't any clues to rub together. > >But, wait a second here. There are three things they have to get >right: pinout-related issues: the 286 emulation: the native mode. >(They don't have to get the cycle counts "right", do they?) > >I would have nominated the 286 emulation as the killer, since the >native mode is mostly "more of the same" with paging. But, AMD is >mass-producing a perfectly good 286, and the 386's microcode can be >reverse-engineered. So, what is to stop AMD (at least) from getting >its 386 debugged and out? Is the pessimism only justified for the >smaller houses, or have I mis-estimated where the problems lie? Nick is more pessimistic than I am personally on this subject. (Note that Microprocessor Report serves as a forum for a variety of viewpoints; it isn't all stuff that I agree with.) I think most of what Nick says applies to companies other than AMD. AMD is a special case because: 1. They have the rights to the 286 2. They have an undisputed right to Intel's patents 3. They may have the right (now under dispute) to Intel's microcode 4. Jerry Sanders is really mad! AMD is sampling parts that are rumored to be in much better shape than Nick's piece would imply. These parts use Intel's microcode, and don't have to work around Intel's patents. The mask design is original, but probably follow's Intel's logic design very closely. Michael Slater, Editor and Publisher, Microprocessor Report mslater@cup.portal.com 707/823-4004 fax: 707/823-0504
davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) (10/25/90)
In article <35204@cup.portal.com> mslater@cup.portal.com (Michael Z Slater) writes: | 1. They have the rights to the 286 | 2. They have an undisputed right to Intel's patents | 3. They may have the right (now under dispute) to Intel's microcode | 4. Jerry Sanders is really mad! | | AMD is sampling parts that are rumored to be in much better shape than Nick's | piece would imply. These parts use Intel's microcode, and don't have to work | around Intel's patents. The mask design is original, but probably follow's | Intel's logic design very closely. I wonder if JS is mad enough to contemplate making a 486-like chip which fits in a 386 socket. I wonder what that would do to Intel's market? Obviously it would have to look like a 386 from bus signal level, but the cycles/opcode could be lower, there could be an FPU, about the only really serious performance hit you take is the loss of burst mode memory access. -- bill davidsen (davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen) VMS is a text-only adventure game. If you win you can use unix.
crisp@mips.COM (Richard Crisp) (10/25/90)
In article <35204@cup.portal.com> mslater@cup.portal.com (Michael Z Slater) writes: >AMD is sampling parts that are rumored to be in much better shape than Nick's >piece would imply. These parts use Intel's microcode, and don't have to work >around Intel's patents. The mask design is original, but probably follow's >Intel's logic design very closely. After the landmark Brooktree versus AMD case involving infringement under the Semiconductor Protection Act of 1984 (the Mask Work Act), I'd be careful about claiming the maskwork of AMD's '386 clone being original. In the Brooktree case, a single cell (the static RAM cell of the RAMDAC) was found to have been infringed by AMD. There sure a lot of cells in a '386! My bet is that Intel will closely examine every square micron of the AMD die and if they find any similarity, will probably add maskwork infringement claims to whatever they sue AMD for (and they *will* sue AMD)! -- Richard Crisp crisp@mips.com MIPS Computer Systems !decwrl!mips!crisp 928 Arques MS 2-02 (408) 524-8177 Sunnyvale, Ca 94086
phil@brahms.amd.com (Phil Ngai) (10/26/90)
In article <10833@pt.cs.cmu.edu> lindsay@gandalf.cs.cmu.edu (Donald Lindsay) writes: |I would have nominated the 286 emulation as the killer, since the |native mode is mostly "more of the same" with paging. But, AMD is |mass-producing a perfectly good 286, and the 386's microcode can be |reverse-engineered. So, what is to stop AMD (at least) from getting |its 386 debugged and out? Here's some more data for your query. Intel stopped delivering usable 286 manufacturing packages with the C step August 1984. AMD came to realize that they were on their own and started reverse engineering the 286 in March 1986, culminating with the introduction of the 286 S Step in Q3 1987. (the judge described Intel's deliveries Q1 1986 as "deliberately incomplete, deliberately indecipherable and delliberately unusable".) So you might believe that AMD does possess a great deal of 286 experience and expertise. -- This article contains my opinions and does not represent the views of the company.
phil@brahms.amd.com (Phil Ngai) (10/26/90)
In article <2790@crdos1.crd.ge.COM> davidsen@crdos1.crd.ge.com (bill davidsen) writes: | I wonder if JS is mad enough to contemplate making a 486-like chip |which fits in a 386 socket. Just imagine the benefit to the consumer if Intel had some competition in the 386/486 market... -- This article contains my opinions and does not represent the views of the company.
seanf@sco.COM (Sean Fagan) (10/28/90)
In article <10833@pt.cs.cmu.edu> lindsay@gandalf.cs.cmu.edu (Donald Lindsay) writes: >I would have nominated the 286 emulation as the killer, since the >native mode is mostly "more of the same" with paging. I wouldn't. There's little difference between the 16-bit (protected) mode and the 32-bit (protected) mode, except for sizes and addresses. All of the same operations and checks come into play for both; the only difference is a bit set in the segment descriptor which says, for example, that registers are 16-bits or 32-bits, and granularity is byte or page, or somesuch. Note that these are different bits, so you can have 32-bit registers, but still be limited to 16Mb through 64K segments. The problem is *getting* the [23]86 stuff in the first place. Once you've got the basics, getting the other isn't that difficult (or doesn't seem so, to me). But the basic architecture is a bitch, to be sure... -- -----------------+ Sean Eric Fagan | "*Never* knock on Death's door: ring the bell and seanf@sco.COM | run away! Death hates that!" uunet!sco!seanf | -- Dr. Mike Stratford (Matt Frewer, "Doctor, Doctor") (408) 458-1422 | Any opinions expressed are my own, not my employers'.
seanf@sco.COM (Sean Fagan) (10/29/90)
In article <1990Oct26.015244.586@amd.com> phil@brahms.amd.com (Phil Ngai) writes: >In article <2790@crdos1.crd.ge.COM> davidsen@crdos1.crd.ge.com (bill davidsen) writes: >| I wonder if JS is mad enough to contemplate making a 486-like chip >|which fits in a 386 socket. >Just imagine the benefit to the consumer if Intel had some competition >in the 386/486 market... Well... the Cyrix '387 clones are pin compatable, are *much* faster than the '387. From what I've read of the cycle counts for the Cyrix chip and the '486, the Cyrix is faster than the "487" as well. (Unfortunately, you have to use it in "memory mapped" mode, a la Weitek, because of the way the '486 was designed.) If/when Cyrix does a '386 clone, I hope it has similar speed advantages over the Intel chips as does their FPU clones... -- -----------------+ Sean Eric Fagan | "Quoth the raven," seanf@sco.COM | "Eat my shorts!" uunet!sco!seanf | -- Lisa and Bart Simpson (408) 458-1422 | Any opinions expressed are my own, not my employers'.
davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) (10/31/90)
In article <8464@scolex.sco.COM> seanf (Sean Fagan) writes: | Well... the Cyrix '387 clones are pin compatable, are *much* faster than the | '387. From what I've read of the cycle counts for the Cyrix chip and the | '486, the Cyrix is faster than the "487" as well. (Unfortunately, you have | to use it in "memory mapped" mode, a la Weitek, because of the way the '486 | was designed.) If I were in the business of producing niche products, I would be highly tempted to build a board with several of these chips, a DMA interface, some memory for workspace and status, and call it a vector unit. If I had a lot of time I would hack a compiler like gcc to produce code for it, but since Cray and Convex users are willing to do operations by calling subroutines, I would probably code up a set of subroutines in assembler to deal with this. I suspect that a board with three FPUs would cost <$2k to build and market, and could sell for $5k. The market would be MS-DOS and 386ix users who wanted to use FORTRAN and stay in a PC environment, but have better than workstation performance with IEEE results. At the time I looked at this (several years ago) there were a few boards in this market, I'm not sure they were shipping, and I don't believe they were IEEE. I suspect you could do the same thing with the 860 and get better performance, but I am not sure the cost could be kept low enough for this market, nor that there is a market one price step higher ($10-15k). -- bill davidsen (davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen) The Twin Peaks Haloween costume: stark naked in a body bag
dik@cwi.nl (Dik T. Winter) (11/01/90)
In article <2816@crdos1.crd.ge.COM> davidsen@crdos1.crd.ge.com (bill davidsen) writes: ... > Cray and Convex users ... > At the time I looked at this (several years ago) there were a few > boards in this market, I'm not sure they were shipping, and I don't > believe they were IEEE. ... The true Cray and Convex users do not care about IEEE. Their machines do not have IEEE conformant arithmetic; far from it. They want results; fast. They do not care about correctness. :-) -- dik t. winter, cwi, amsterdam, nederland dik@cwi.nl