Announcement

Collapse
No announcement yet.

Vortex86 Processors Finally Seeing Work In 2021 For Proper Handling Under Linux

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Vortex86 Processors Finally Seeing Work In 2021 For Proper Handling Under Linux

    Phoronix: Vortex86 Processors Finally Seeing Work In 2021 For Proper Handling Under Linux

    The Vortex86 32-bit SoCs have worked under Linux for those distributions still maintaining 32-bit x86 support and where not hitting corner-cases of some i686 level features not being supported by some Vortex86 cores, but there is finally a pending kernel patch to provide proper CPU detection for Vortex86 hardware...

    https://www.phoronix.com/scan.php?pa...inux-Detection

  • #2
    Funny enough, I had actually heard of these.

    Using a Vortex86 System-on-Module to build a super-tiny MSDOS gaming PC.

    Comment


    • #3
      Originally posted by ssokolow View Post
      Funny enough, I had actually heard of these.
      Those things actually look really cool. I have seen them on a few MS-DOS groups.

      Comment


      • #4
        Btw, the first x86_64 patents run out around 2025.

        Comment


        • #5
          Originally posted by ssokolow View Post
          Funny enough, I had actually heard of these.
          I've never heard of this, but it does look very interesting. Thanks

          Comment


          • #6
            what's wrong with running msdos in vm on real computer?

            Comment


            • #7
              Originally posted by pal666 View Post
              what's wrong with running msdos in vm on real computer?
              Hard real-time behavior and complexity.

              Hard real-time behavior: There are DOS programs out there driving (old) hardware that rely on having predictable timing. A CNC machine that usually reacts within 1ms when it's told to stop the stepper motor but sometimes takes 500ms can screw up an hours-long production. Besides modern operating systems usually not optimizing for that very well (even though there are patches), there's firmware level code that can interrupt at any time for any period of time (e.g. the System Management Mode) that can give you a really hard time.
              Going through the Vortex86 data sheets it doesn't appear to have an SMM implementation (although the docs speak of "SMM memory" at 0xa0000, but that's the only reference), so once you know the reaction time of the hardware, you know how fast it will react in any given situation. No nasty surprises.

              Meanwhile, modern firmware uses SMM for all kinds of things. Some update in the background that updates some UEFI variable? CPU temperature approaching some magic number? The real time clock flips through some special state that the firmware decides to double check (to work around a hardware bug)? If the milling bit was moving while any (or more) of these happened, boom: the $200k part you've been building for the last 48 hours is now scrap metal.

              Complexity leads to similar outcomes: Yes, you _can_ route a virtualized MSDOS to access the actual I/O ports, but every time the DOS application triggers such an I/O port, you get a number of context switches in and out of kernel mode before anything actually happens (and even if you can optimize those away, there are still transitions in and out of the CPU virtualization mode), each with an indeterminate number and size of cache flushes, each of which take some semi-random amount of time.

              If you know precisely what you need, why suffer through all this?
              Last edited by pgeorgi; 17 October 2021, 01:47 PM.

              Comment


              • #8
                Wonder when they'll go 28nm. Interested to see how far a 486-based design can go.

                Comment


                • #9
                  Originally posted by monkeynut View Post
                  Wonder when they'll go 28nm. Interested to see how far a 486-based design can go.
                  I haven't looked at the data sheets, but, from what I've heard, the IPC on Vortex86 is much lower than original 486s because it's a RISC design with x86 implemented in microcode. (Typically followed by "Not that it's a problem for a DOS retro-machine. What's more relevant is how, if it's still too fast, you can divide the clock with a simple dropdown in the BIOS".)
                  Last edited by ssokolow; 17 October 2021, 08:55 PM.

                  Comment


                  • #10
                    Vortex86 has an interesting niche. I wish I could buy vintage Windows and Microsoft Office on GOG with right to repair applying to software and have Vortex86 implement SSE2 and have the ultimate Win 9x offline experience. As much as I like Linux, I do miss the days when the internet was optional, like you didn't need the internet even in 2005 to watch a movie on your computer, people bought DVDs and didn't have rent only centralization.


                    The internet is like antibiotics, it's useful and revolutionary, but we're over dependent on it working and we over use it and we over used it so much, it doesn't work right anymore. SEO ruined search engines, search engines index reddit, stack exchange and websites bloated with javascript and SEO fluff or pseudo-intellectual fluff that tricks the bias of natural language processing AIs to make that website more "relevant" with an opening paragraph full of fluff.

                    Comment

                    Working...
                    X