Image Map Image Map
Page 3 of 4 FirstFirst 1234 LastLast
Results 21 to 30 of 31

Thread: Which version of MS-DOS has the most software written for it?

  1. #21
    Join Date
    Jan 2007
    Location
    Pacific Northwest, USA
    Posts
    31,227
    Blog Entries
    20

    Default

    Hence the term "BIU" for "Bus Interface Unit"--the thing that connects the CPU core with outside memory space. The core of the 8086 and 8088, as mentioned above, is the same; the BIUs differ.

  2. #22
    Join Date
    Mar 2011
    Location
    Atlanta, GA, USA
    Posts
    1,394

    Default

    Quote Originally Posted by SomeGuy View Post
    A 486 with 8-bit RAM would be a 2 bit computer :P
    I don't remember them costing $.25 USD!

    The BIU really isn't indicative of anything related to the architecture of the CPU or ISA - which is what most people refer to when saying "a x-bit CPU or machine". The best example above is jfir pointing out the Pentium has a 64-bit BIU but it certainly isn't a 64-bit machine by anyone's criteria. The BIU in the 8086/88 is the equivalent of the FSB or Front Side Bus in later processors - only it went all the way to the back! The idea of a North/South bridge started with VESA to run the two halves of the machine a different speeds and architectures. The BIUs in modern CPUs (formerly North Bridge / FSB) went from parallel busses to multiple 1-bit wide high-speed differential serial lanes more than a decade ago with QPI, UPI, HyperTransport, et al. But that doesn't change x86_64 (or ia64) still being considered a 64-bit CPU, machine, and architecture.

    8088 is a 16-bit CPU, ISA and machine. 386sx is still a 32-bit CPU, ISA, and machine. Pentium is still a 32-bit CPU, ISA, and machine. i7 is still a 64-bit CPU, ISA, and machine. The only thing in-common is they all have differently sized data paths for packaging, pin-count, and size reasons. Remember 8-bits wide x 100 MHz is the same bandwidth as 16-bit wide x 50 MHz.

    I'm not saying you are wrong. I'm just saying the vast majority of people disagree with you... If you want to change their minds, please provide some logic we can wrap our heads around.
    "Good engineers keep thick authoritative books on their shelf. Not for their own reference, but to throw at people who ask stupid questions; hoping a small fragment of knowledge will osmotically transfer with each cranial impact." - Me

  3. #23
    Join Date
    Jan 2017
    Location
    Galicia, Spain
    Posts
    148

    Default

    Quote Originally Posted by Svenska View Post
    I remember most software requiring at least MS-DOS 3.3 or 5. But then, MS-DOS 3.1 was the last version without 3.5" disk support (OEM versions excluded), so maybe that's the reason and the software would have worked.
    I still remember a program that listed MS-DOS 5.0 as a requisite. Actually the program worked fine with any DOS version > 3.0 but it used about 600KB or RAM so they assumed that anybody using a version of DR/PC/MS-DOS >5.0 had correctly configured UMBs and was able to get that amount of memory free. Of course you can got more or less the same amount of free memory with an older DOS version using QEMM/386MAX (and I did) but they wanted to "simplify" the matter and avoid lot of support calls ...

  4. #24
    Join Date
    Jan 2007
    Location
    Pacific Northwest, USA
    Posts
    31,227
    Blog Entries
    20

    Default

    Quote Originally Posted by eeguru View Post
    8088 is a 16-bit CPU, ISA and machine. 386sx is still a 32-bit CPU, ISA, and machine. Pentium is still a 32-bit CPU, ISA, and machine. i7 is still a 64-bit CPU, ISA, and machine. The only thing in-common is they all have differently sized data paths for packaging, pin-count, and size reasons. Remember 8-bits wide x 100 MHz is the same bandwidth as 16-bit wide x 50 MHz.
    I've always wondered why an 8-bit MPU (or a PDP "straight 8") using a bit-serial ALU isn't considered to be a one-bit machine, if the width of the bus interface doesn't determine the "bit-edness"...

  5. #25
    Join Date
    May 2011
    Location
    Outer Mongolia
    Posts
    1,137

    Default

    The question of whether to call the 68000 a 16 bit or a 32 bit machine still gives me pause, because even though it does use a (mostly) 32 bit programming/addressing model my recollection at least is on the hardware level the 16 bit-ishness extends beyond the BIU into the ALU and affects the cycle times of most 32 bit operations even when they involve only internal registers. Thus resulting in a CPU that could arguably can kind of be described as a 16 bit chip emulating a 32 bit one?

    Of course, that's also what bit-serial ALUs do, so... yeah, I guess I'd say it's a 32 bit CPU, but then I *also* recall arguments that point out some specifically 16 bit-isms in the instruction set/addressing modes of the 68000, and this is where my eyes glaze over.

  6. #26
    Join Date
    Jan 2007
    Location
    Pacific Northwest, USA
    Posts
    31,227
    Blog Entries
    20

    Default

    Addressing granularity has nothing to do with it. I've worked on a supercomputer with 1-bit granularity in addresses; it isn't that uncommon in vector architectures.

  7. #27
    Join Date
    Oct 2014
    Location
    near frankfurt/m, germany
    Posts
    848

    Default

    68000 is not emulating 32 Bit. It is 32 bit internal design for all data path and registers. Only external bus interface is 16 bit. And there is 68008 CPU with external 8 bit data path for low cost (Sinclair QL). But they run the same software as 68020 does.

  8. #28

    Default

    Quote Originally Posted by Chuck(G) View Post
    I've always wondered why an 8-bit MPU (or a PDP "straight 8") using a bit-serial ALU isn't considered to be a one-bit machine, if the width of the bus interface doesn't determine the "bit-edness"...
    Keep in mind that the Z80 only has a 4-bit ALU...

    Quote Originally Posted by 1ST1 View Post
    68000 is not emulating 32 Bit. It is 32 bit internal design for all data path and registers. Only external bus interface is 16 bit. And there is 68008 CPU with external 8 bit data path for low cost (Sinclair QL). But they run the same software as 68020 does.
    Well. the 68000 is actually a mostly 16-bit chip, but it's ISA has been designed with a 32-bit future in mind. I wouldn't call it emulation, but its native language doesn't match its implementation.

    Contrast that to the 80286, where the forward-thinking/future-proofing only applied to some CPU-internal data structures...

  9. #29
    Join Date
    May 2011
    Location
    Outer Mongolia
    Posts
    1,137

    Default

    Quote Originally Posted by 1ST1 View Post
    68000 is not emulating 32 Bit. It is 32 bit internal design for all data path and registers.
    The registers are 32 bits wide, but the data ALU is 16 bit and because of that there's a significant subset of instructions that require more cycles to execute on long words than they do on bytes or 16 bits words (even on register operations), which wouldn't be true if it had a fully 32 bit ALU.

    Here's an article by a Motorola engineer pointing out reasons why it's hard to fully characterize the 68000 as "either" a 16 or 32 bit CPU. I totally agree that it's fair to call it either both or neither, depending on your feelings about water glasses.

    Also note that the datasheet for the 68010/68012, which were improved versions of the 68000 that supported MMUs and, in the case of the 68012, up to a 2GB physical address bus, says "The MC68010 is the third member of a family of advanced microprocessors from Motorola. Utilizing VLSI technology, the MC68010 is a fully-implemented 16-bit microprocessor with 32-bit registers, a rich basic instruction set, and versatile addressing modes." (See also chapter 7 for the typical execution times and note they're usually at least 2 to 4 cycles longer for long words vs. short ones, obviously even more if they're hitting memory.)

    Note that by pointing this out I'm not a criticizing the CPU, it was obviously a fantastic effort for 1979 and it was genius move to design an API that could be used for a fully 32 bit implementation when it was possible to stick enough transistors on a chip to do it. (Unlike what Intel did with the aforementioned 80286, which was double-down *hard* on strict 16-bit-ness, and their kind of weird implementation of it at that.) I was just mulling that it's a more complicated case compared to, say, the 386SX, which has the same 32 bit ALU as the full 386 paired with a narrower BIU.

  10. #30
    Join Date
    Jan 2007
    Location
    Pacific Northwest, USA
    Posts
    31,227
    Blog Entries
    20

    Default

    And to be sure, the notion of "bit-edness" of a CPU has always been a bit of a complex issue. Some CPUs have several functional units, each with different bit-edness.

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •