Indivision ECS V4: New features, development progress

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.


Please understand that you need to create an account to be able to post, guest posting was disabled as an anti spam measure.

    There is stock left and you can order now. We can ship today (it's already Monday morning here - IIRR, you're downunder?).


    It's just that the shop stopped taking orders for this particular product during the weekend because it detected a scalping attempt.


    Jens

    Oh, no, I understand that supply is currently available. My ability to purchase - the money I'm waiting on - doesn't happen for another couple of weeks, so I'm hoping there's still stock when the time comes. I'm hoping to get an ACE2B at the same time, also. Stocking up on iComp goodies seems to be an annual hobby of mine. :)


    So that's all I'm wary of: I know there's stock now, will there be when I come to buy. I hope so.

    Quote from LineOf7s

    so I'm hoping there's still stock when the time comes.

    I'm confident that we'll make it all the way into next year with the currently-available stock. That's taking two "order rushes" into account: The one that I'm expecting when we release the new core and Picasso96 V3.6.0 and the usual Xmas rush. No need to hurry, no need to pay Scalper's prices, no need to buy on credit.


    Jens

    Just made a quick module with the same A500/68030-33 setup, but the old CPLD version that does not allow 32-bit access to the frame buffer. Surprisingly, almost all benchmarks are the same speed, and only HiColour writepixel gains 38% with the faster access.


    I did run with a slightly changed software setup, though, as the TCP/IP stack wasn't running. Some tests that should be identical (as they are not gfx related) are 7% faster. So the stack running in the background does eat some performance. I'll do future modules without network running, so we have an equal base, especially with older modules.


    Jens

    Quote

    LineOf7s

    What kind of resolution(s) do you run? Are they "fine", or perhaps "a bit slow but usable" or... what? Would any speed increase be the icing on the cake or might it be most of the cake? :)

    I on my A600 with turbo A630 030/50 run at 800x600 resolution with 256 colors. I find it the most responsive :). Here's a comparison of ECS vs IndiV4 at 800x600 resolution:


    and Sysspeed module:

    A630_800x600_Iv4_ECS.lha


    And here's a comparison A600 with turbo card A630 with an A500 (from Jens) with ACA500plus and an old ACA1232 running at 33MHz (unnamed) at 1280x1024 resolution:


    and Sysspeed module:

    A630_Iv4_1280x1024.lha

    Given I'd be running the V4 on an A500 with ACA500plus and ACA1234 running at 50MHz, this is all very useful and relevant information, thank you.

    Benchmarking P96 on Indivision ECS V4 is mainly benchmarking the accelerator. The only thing that's truly relevant is the "percieved speed", IMHO. And that part would be the transfer from fastmem-framebuffer into Indivision ECS V4, which gets a major speedup from the ACA500plus CPLD update. However, that can't be measured, as current benchmarks will only measure how long it takes for the P96 API to return a function call. However, that's not when all the work is done - it's only the time when the API is ready to take the next command, but the changes still have to be transferred to the actual frame buffer.


    For things that are done with the blitter, you win a little, but it's still "double work", as we need to keep two frame buffers (one in fastmem, and one for displaying). Since there is no blitter for the frame buffer in fastmem, the biggest influence on performance is with the accelerator.


    Jens

    As for performance, Doom (ADoom), for example, goes just as fast on the Indi as it does on the ECS with the difference that there are 256 colors on the Indi and not EHB mode.

    ECS -- 6.1FPS:


    RTG Indi v4 -- 6.9FPS


    Well, and you can run some other games that use RTG modes :). Unfortunately, it's impossible to compare if Indi speeds up anything in them, because they are mostly AGA + RTG versions of games (like Abuse). You can also play on a Mac emulator in 256 colors.

    Definitely still have to wait for the Indi software update :).

    Thank you for that. The good news is I really only want it for the bigger resolutions and extra colours, so the inability to play RTG games quickly bothers me little. That's more of a Pistorm thing, really (for me, at least).


    So as long as the Workbench feels snappy, is nice and big and is plenty colourful, I'll be happy. :)

    There's still one bug in the driver that makes it crash when changing resolutions. Thor didn't have any time debugging that these past few weeks, but my personal hope is that we can release it before our summer break.


    Jens

    New software in our Wiki!


    Config tool V1.1 (release candidate) contains the latest core with Framebuffer and 2D acceleration. The tool will also complain if it's opened on an RTG screen. If P96 has taken over, it really did, and also takes control of output frequencies. The software would compete for output modes and hardware resources in general, which can't be resolved in a meaningful way. The best solution is to launch the tool when it's on a native graphics mode.


    Launch this tool and make a rescue disk first. Then choose "Flash Firmware" from the drop-down menu (press and hold the right mouse button and move the mouse to the "Firmware" text in the title bar of the screen). The screen will turn black and the monitor will lose sync during the update process, which takes around 15 seconds. Don't interrupt this process - let it run until the picture comes back. The new core will be active immediately, which you can verify on the core version displayed in the OSD


    I have also added a tiny archive with the new P96 driver for Indivision ECS V4. Install by copying it to the libs:Picasso96/ drawer. This driver makes use of the blitter. Further, transfers from an A1200 accelerator on ACA500plus into the ECS V4 frame buffer are twice as fast if you have hardware ACA500plus HW ID $0B. This is shipped as standard with current cards, and we'll make a JTAG Updater available for all older versions of the ACA500plus (all the way back to units shipped in 2016!) in a few weeks.


    Please report back if you find any errors - or if it just works for you. The P96 driver should work with older versions of P96 as well (back to V3.2.0), so in order to participate in thie beta test, you don't have to buy the latest version but use the one you've purchased between December 2020 and today.


    Jens

    My ACA500plus HW ID is showing $08. Assuming $0B isn't a typo (or misread), am I correct in thinking I'm three versions too old and I should hold off for a few weeks for the JTAG adaptor before attempting to evaluate the new V4 P96 driver with my ACA1234?

    Quote from LineOf7s

    My ACA500plus HW ID is showing $08. Assuming $0B isn't a typo (or misread), am I correct in thinking I'm three versions too old and I should hold off for a few weeks for the JTAG adaptor before attempting to evaluate the new V4 P96 driver with my ACA1234?

    No need to hold off: It will just work, but not at the accelerated transfer speed. When the JTAG updater is available, you'll just connect that one-time, and it'll be faster, no need for further installation.


    Jens

    Now trying the new firmware. Not really sure what I was expecting but this seems to be a huge speedup vs the old version.

    Yes I have a aca plus + aca 020 card but not installed in this machine.
    I’m currently testing with a DIY Matze (a1k) 68030tk2@60 mhz in the cpu socket.