New Buster chip

Caution: Non registered users only see threads and messages in the currently selected language, which is determined by their browser. Please create an account and log in to see all content by default. This is a limitation of the forum software.


Also users that are not logged in can not create new threads. This is a, unfortunately needed, counter measure against spam. Please create an account and log in to start new threads.

Don't Panic. Please wash hands.
  • Hi.


    I heard a new buster chip is being developed by Dave Haynie.

    I tried to figure out more about this but I cant find anything.

    How will this be better then the old buster chip?

    Is there an estimate on when it can be released?

  • Sorry, no details yet.


    The existing Buster chip has a few bugs, mainly with Busmaster DMA. These are "fewer" with the -11 version than with the -9 version, but they still trigger in the form of complete bus-hangs.


    With Buster-11 being sold out for good, I thought it's a good idea to make a replacement that appeals to the whole Amiga community, and makes fake chips completely unattractive.

  • It's a fantastic project, Jens - big credit to you to make this move, with Dave in particular.

    I understand the goal is to deliver (sell?) by the second half of 2021 which, from a development perspective, means probably that your production prototype will be available at least 3-4 months before? Did you plan to communicate regularly on the progress?

    Do you intend to deliver a fully finalized solution or something that will see few upgrades with times?

    From a technical point of view, how have you (or will you) resolve the 2 main physical issues I see: the manufacturing of a chip (FGPA?) and its physical installation on existing socket? (simply dropped in? welded?)

    Thanks a lot.

    French Amiga fan living in Australia ;)

    Owns all Amiga machines except A3000T and CDTV. Owns 2 or more of each kind except A3000 and CD32 (only one of each!)

    Jetski fisherman as well. 8)

  • Did you plan to communicate regularly on the progress?

    I hardly ever do that. Remember that iComp is an expensive hobby for me, nothing that I can make money on, so posting progress updates and goals/outlooks on things will only add pressure that won't really help anybody.


    Do you intend to deliver a fully finalized solution or something that will see few upgrades with times?

    The product will not be field-upgradable. It will therefore have to pass all tests before being shipped.

    From a technical point of view, how have you (or will you) resolve the 2 main physical issues I see: the manufacturing of a chip (FGPA?) and its physical installation on existing socket? (simply dropped in? welded?)

    As you might have seen on the new ACE2b product, I've had tooling made for a PLCC84-adapter. The same component will be used for the Buster-12 board, which contains a small FPGA/CPLD (crossover - the unit is "in between" the two technologies). The shape and looks are comparable to ACE2b.

  • Thanks Jens. Appreciate your personal replies.

    French Amiga fan living in Australia ;)

    Owns all Amiga machines except A3000T and CDTV. Owns 2 or more of each kind except A3000 and CD32 (only one of each!)

    Jetski fisherman as well. 8)

  • this is a very good news, but will this new buster chip allow an increase in zorro3 speed too?

    I'm thinking about Zorro-3 fast ram that is well known to be very very slow.

  • but will this new buster chip allow an increase in zorro3 speed too?

    No, that would introduce incompatibility with pretty much any Z3 card. The goal is to fix bugs, not introduce new ones :-)


    I'm thinking about Zorro-3 fast ram that is well known to be very very slow.

    Memory is only truly fast if it's "near the CPU". On an A3000/A4000, there is a 68030 interface and the Z3 bus "in between" for any 68040/68060 accelerator, so you won't get the same speed from Z3 memory as you can get with local memory. That's the whole secret of a fast accelerator: Not the CPU clock, but the low latency to memory is key to performance, as a CPU can only execute it's commands as fast as it can read them from memory.

  • Memory is only truly fast if it's "near the CPU". On an A3000/A4000, there is a 68030 interface and the Z3 bus "in between" for any 68040/68060 accelerator, so you won't get the same speed from Z3 memory as you can get with local memory. That's the whole secret of a fast accelerator: Not the CPU clock, but the low latency to memory is key to performance, as a CPU can only execute it's commands as fast as it can read them from memory.

    Yes i know this and I was not expecting that Z3 memory speed would get any closer to CPU board one, i was just hoping in a faster memory speed compared to what we get with buster-11. Low memory is a problem on big box amigas as Cyberstorm supports just 128mb and with and hack you can just add 64 or 112 mb on the mobo, the only way to increase memory seems to be adding Z3 memory but it's really too slow. Anyway buster-12 is still a very good project.

  • What 68k application requires this much memory? I mean, 256M sells on a regular basis with BigRamPlus, and the ACA1240/1260 will also get a bump to 256M just because it's easier to explain than 192MBytes. However, I always thought that even the 128M of my 68030 accelerators is an oscene amount for a 68k Amiga. Even when compiling huge projects like Poseidon, you're fine with 32M RAM.

  • Any news on the project? What is the best place to hear from it?

    French Amiga fan living in Australia ;)

    Owns all Amiga machines except A3000T and CDTV. Owns 2 or more of each kind except A3000 and CD32 (only one of each!)

    Jetski fisherman as well. 8)

  • Buster-12 will be designed with 25MHz board frequency in mind. However, since we're using PLLs (instead of glitchy-clutchy logic) to generate a local clock for a fully-sync design, there's a chance that it will also work in one of the early 16MHz A3000 boards. You may need to set a jumper on Buster-12 for that to work, as the PLL needs to be configured differently.