Image Map Image Map
Page 9 of 10 FirstFirst ... 5678910 LastLast
Results 81 to 90 of 97

Thread: Worst x86 CPUs over the years

  1. #81
    Join Date
    Jan 2007
    Location
    Pacific Northwest, USA
    Posts
    33,690
    Blog Entries
    18

    Default

    Problem was, that by 1979, there were x80 systems that had more than 64KB of memory, so it should have seemed obvious that any sort of baling-wire setup was going to be a problem. Motorola certainly realised that. I think that Intel got sidelined with the prospect of the IAPX432. My impression of the 8086 was that it was a low-end stopgap product, rather than anything revolutionary.

    What was surprising was the lack of 16-bit peripheral chips, even 3 years after introduction. How many PC platforms used the tired old 8237 DMAC, that was intended for the 8085 family? Sure, tack on enough external logic and you can get it to address 20/24 bit address buses, after a fashion. But the whole affair was nothing more than chewing gum and bailing wire. The Boeing 737 MAX of the CPU world.

  2. #82
    Join Date
    May 2011
    Location
    Outer Mongolia
    Posts
    2,055

    Default

    To me the real head-scratcher is probably the 80286. The 8086 was designed starting from 1976 and released in 1978 so I'd say its compromises were reasonably defensible, but doubling down on 16 bit segmentation for a chip with a 24 bit address bus (and adding virtual memory support on top of that; theoretically a 286 can pretend to have up to one gig of RAM)... made sense to somebody.
    My Retro-computing YouTube Channel (updates... eventually?): Paleozoic PCs

  3. #83
    Join Date
    May 2009
    Location
    Connecticut
    Posts
    4,635
    Blog Entries
    1

    Default

    Quote Originally Posted by Eudimorphodon View Post
    To me the real head-scratcher is probably the 80286. The 8086 was designed starting from 1976 and released in 1978 so I'd say its compromises were reasonably defensible, but doubling down on 16 bit segmentation for a chip with a 24 bit address bus (and adding virtual memory support on top of that; theoretically a 286 can pretend to have up to one gig of RAM)... made sense to somebody.
    The 286 included all the good bits of the 432 slapped on top of an 808x design and was thrown together in about 18 months. Another place holder for the real CPU that would be showing real soon now. Had protected mode been as easy to convert to as going from 8080 to 8086 was, the 286 would have been an absolute winner. There was a sad period in the mid-80s when every nascent protected mode operating system experimented with using LOADALL to multi-task DOS applications in protected mode. Didn't work but set the stage for the 386 finally catching up with what buyers wanted.

    The 432 was segment and protection madness. Even I, slightly more favorable to segments than many, think the 432 went a bit too far in that regard. IIRC, the 432 was several times the price of the 286 at the same clock speed but yielded maybe 1/4 the performance in the tasks the 432 was expected to be best at. The 432 looked even worse for everything else.

  4. #84
    Join Date
    May 2011
    Location
    Outer Mongolia
    Posts
    2,055

    Default

    In some alternate universe closely adjacent to ours Intel included the equivalent of VM86 mode in the 286, and everyone in that continuum remembers the CPU far more kindly.
    My Retro-computing YouTube Channel (updates... eventually?): Paleozoic PCs

  5. #85
    Join Date
    Jan 2014
    Location
    Western North Carolina, USA
    Posts
    1,289

    Default

    Quote Originally Posted by Chuck(G) View Post
    I've voiced my distaste for the original 8086 implementation before.
    You and Frank Durda were of like minds.... His .sig for years was (pulling one at random):
    Code:
    Frank Durda IV - only these addresses work:|"The Knights who say "LETNi"
       <uhclem.aug98%nemesis.lonestar.org>     | demand... A SEGMENT REGISTER!!!"
                                               |"A what?"
    This Anti-spam address expires August 31st |"LETNi! LETNi! LETNi!" - 1983
    In reality, the whole segmentation idea was basically a poor mashup of an MMU tacked on top of 16-bit addressing. IMHO, at least. For late 70's code and memory sizes segmentation made sense, at least to a few. It was a CISC design, after all.
    --
    Thus spake Tandy Xenix System III version 3.2: "Bughlt: Sckmud Shut her down Scotty, she's sucking mud again!"

  6. Default

    Chuck(G): The original conception of the segmentation and 8086 was the following:

    1. You have an operating system with a GUI, disk system and file management, video and audio drivers, program code interpreter/compiler, text editor, and so on. Loaded at boot, and running in a 64kbyte chuck of RAM. (64k was considered plenty in 1979, and all of these should fit by those standards, including the stack and some spare memory).

    2. If you start a program, then it will receive a free 64kb chuck and it will be loaded to there. Like, for example, you execute a game. This game never supposed to touch memory outside of its segment. But the cpu supports touching the another segment. This has multiple reasons. There are no priviliges, so the cpu cant know if system mode or user mode program is running, thus it must be able to write and modify everything. Paging is not a thing yet. Also, they was not sure how the OS will work - will the os be only available through an interrupt, or a process should be able to directly access memory there, or call a function there - so they didnt blocked the ability to reach memory outside the segment. They could not know in 1979, how the operating systems will work later on. They assumed an operating system with GUI and multitasking, running multiple programs, and they have assumed 64k space for programs will be enough. Which was true in 1979.

    3. So the operating system is taking up your first 64k, and the game you are running, taking up another 64k. Then you start a word processor, which is taking up another 64k. Then you finally start an excel software, that takes another 64k, and then prety much your RAM ran out aniway, you already running 3 programs and an OS with a GUI, prety much pushed the architecture to the limits. Of course you can add 640k ram in total, which will allow you to run 6 or 7 programs in parallel, well 640 is 64*10 but it can be assumed that the firmware (bios) will dirt some for you, so you cant use the whole.

    4. All of the above... never happened. DOS was designed to be a single-process environment, without GUI. Its accessing the whole memory, and then the programs written for the system were also accessing the whole memory. Then this design became obsolete. But for the backward compatibility (which was one of the main point of x86), they had to preserve this design. Originally, however, a random program wasnt supposed to just randomly touch outside from its segment, this was never the intended use-case of this architecture, more like the inability of microsoft and ibm to understand how to produce a multitasking environment for the system.

  7. #87
    Join Date
    Jan 2007
    Location
    Pacific Northwest, USA
    Posts
    33,690
    Blog Entries
    18

    Default

    I might buy that, but the point of going to 16 bits for many vendors was the prospect of managing large (>64K) data structures.

    There were multitasking systems running on x80 hardware, but accessing large data structures was very difficult. For x80, we simply inserted a bipolar RAM chip between the CPU and memory to translate the 16-bit addresses into 18-bit ones. Easy--and you can even have "pages" mapped wherever you want.

    So Intel designs a 16-bit CPU without a barrel shifter and sets the granularity of segments at 16 bytes. What kind of sense does that make? Why not 256 bytes? At least address computation would be easier.

    And Xenix existed for the 8086.

  8. #88
    Join Date
    Sep 2006
    Location
    Silicon Valley
    Posts
    2,199

    Default

    Quote Originally Posted by Geri View Post
    All of the above... never happened.
    Actually, it DID happen.
    Convergent Technologies CTOS, which managed to work just fine in a multitasking environment on a 80186 with GUI and networking.

  9. Default

    Quote Originally Posted by Al Kossow View Post
    Actually, it DID happen.
    It didnt happened. DOS totally dominated the early x86 market. Everything else is just a boring and irrelevant side-note or an experiment with 40-ish users including all the *nix variations for it. Thats why Chuck(G) wasnt able to put the goal of segmentation into context as well.
    Last edited by Geri; January 29th, 2020 at 01:50 PM.

  10. #90
    Join Date
    Sep 2006
    Location
    Silicon Valley
    Posts
    2,199

    Default

    Quote Originally Posted by Geri View Post
    It didnt happened.
    sure thing, sunshine

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
  •