The delphina DSP boards seemed to work quite will (different DSP) and to answer a previous question were quite happy to play mp3 properly. The 50Mhz 060 could not really play mp3 at all well with 100% CPU use.
Hello, Well done for having managed to revive the DSP planned for the Amiga. What, in my opinion, could be interesting would be to see a comparison of the performance of the DSP against the onboard 68030 and 68882, as well as against a 68040 (A3640). To me, it's a good way to realize what we missed at the time. In addition to the mandelbrot, a JPEG display comparison would be really interesting. I also hope one day to be able to hear the DSP audio 16 bit 8 channels and MP3 playback. Congratulations again to everyone who made this video possible.
Thanks for the comments. I agree with your "wish list" comparisons. The bit that is holding this back right now is that the DSP is very temperamental with the CPUs/accelerators it will work with. Some not at all (eg A3640 and onboard '030 - which is totally weird) and some only partly (eg CSMk2). Atm, I don't know why that is, and until someone with a logic analyser takes a look at it, we're not likely to make much progress.
@@wrangleramiga896 I didn't know for the onboard 68030 or A3640. All we can do is to hope that one of you can get them to work, as you did for the DSP. We waited 30 years to already see this video, we can wait a little the rest (well not 30 more years anyway :))
@@wrangleramiga896 If it's not working with the onboard CPU, then I think that the version of the 3000+ that these replicas are based on might not be the newest version that commodore did. Either that, or they never got the DSP up and running themselves.
And of course it was available a few years before the '060. Just having the DSP with the original '030 would also have eliminated all the compatability and optimisation issues with lots of existing software that the '060 would have suffered from even if Commodore hadn't gone bust, you've just got a decent ('030) with good compatability and with which developers are familiar and a distinct quasi FPU in the DSP chip. You'd probably have to wait until 1995 or 96 to be getting the best out of your '060 and of course that wouldn't solve problems with older software.
To try and eliminate some of the chip-ram access overhead you could try changing the approach a little. Doing horisontal lines instead of vertical lines and then only writing to the chipram every 32 pixel - i.e. taking 32 individual calculated pixels from the dsp-interupts and keeping them in an internal cpu-register (or even fast-ram) and then only moving a single longword to chipram in one go - seems like that would remove the most excessive overhead.
Are you the designer of this motherboard? I did see this board on an Amiga forum but it looked as though ordering the board was no longer possible. Also does Dave Haynie have one of your recreations? I was tempted to see if it would be possible to order a few as this was a machine I always wanted.
No, I don't have the skill to make anything like that! The designer is Hese and the sales channel is a thread on Amibay. I don't think these boards or the DSP software have made their way to Dave Haynie. Hopefully he would be pleased to see his prototype produced in numbers and running software!
@@wrangleramiga896 I was thinking of possibly buying one but was tempted to possibly buy one for Dave Haynie too (I would love to see his comments on this design for something which never found the light of day). The Amiga forum I came across (I think it may have been Amibench) seemed to look like no future orders so I think I would need to pay higher than usual. Also whilst I could solder it together I am not sure I would feel comfortable at all doing it. I have a very rare prototype (one of two; though I own the lesser of the two). I was tempted to lend this to either Hese or Chucky to reverse engineer and make the design available. When Dave auctioned lots of prototypes; it was the only one which I could win over his then boss (Ryan).
@@wrangleramiga896 It is the lesser of the two Gemini prototypes. There are pictures of it on The Big book of Amiga hardware. I bought the lesser one before Dave found the other one :(
Sadly, not much. Search for mandel_dsp on Aminet and you will get the Mandelbrot generator I'm showing in this video. Also on Aminet are asm3210 and gcc3210 which are the programming tools I created to write other software for the DSP. I'm waiting for someone to write a jpeg or mp3 decoder!
I'm not familiar with the Falcon but from what I read it's DSP was a Motorola DSP56001 at 33MHz whereas the AA3000+ has an AT&T DSP3210 at 50MHz. So different clock speeds and architectures eg the 560001 has 2x56-bit floating point accumulators whereas the 3210 has 4x40-bit FP accumulators. Very difficult to compare without running a head to head somehow!
@@wrangleramiga896 From Freescale'sy DSP56001/D document "The data paths are 24 bits wide thereby providing 144 dB of dynamic range; intermediate results held in the 56-bit accumulators can range over 336 dB."