Would it be financially worth it when things like SwinSID, ARMSID and FPGASID exist?..

Interesting idea, but prices would need to come down a lot further for it to be interesting. Take CMP. I'm not up to speed enough on the process to be sure I'm covering everything, but at the very least you need the dies themselves and the package. Costs for the cheapest process are EUR 640/mm^2 for 25 dies. Now, SID was 27.2mm^2 on 0.7um process, so on 0.35um process assuming a straight shrink it would be 7mm^2. That means EUR 4480 for 25 dies. Even assuming 100% yield, that's EUR 180 per die. Then unit price for a DIP-28 package (like SID) would be EUR 58. Adding that on gives you EUR 238 each. There's also the one-off costs per project, which look like several thousand EUR, so at the very least multiple tens of Euros per chip on top of that. Assuming realistic yield and you're looking at >>EUR 350 per chip. Vastly cheaper than you'd get with EUR 500k setup costs, but as long as the chips themselves can be purchased for a small fraction of that amount, I don't see this going anywhere soon.

@dionb Aww... yeah, running those numbers really makes sense. Well, maybe it will become cheaper over time. The SID will definitely become more expensive. No denying that...

Regarding SwinSID et al: Yes, those exist, but my dream would be that we have "original" replacements, that look and feel like the original. It's just like with historic automobiles: sure, you can replace the bulbs with LED ones, but then the car wouldn't look authentic. However with a modern incandescent bulb it will look much better. Same goes for historic computers: We can stuff PLAs made from 2 GALs and a PCB in there, a VIC made from FPGA and a SwinSID with an Atmega. But it will look weird once you work on the machine -- even if the general user can't see from the outside. But we tinkerers aren't the general users! 😉

EDIT: the 27mm^2 is using a VERY old process. I figure by now this will have shrunk by factor 10 at least, right?

root42 wrote on 2020-06-18, 12:09:

EDIT: the 27mm^2 is using a VERY old process. I figure by now this will have shrunk by factor 10 at least, right?

Yes, but shrinking a die to a smaller fabrication process is not as simple as doing an Image -> Resize, 99% of the time it will not work.. So you can't really take the SID die, X-Ray it, then try to replicate it on a smaller fab.

appiah4 wrote on 2020-06-18, 12:53:

Yes, but shrinking a die to a smaller fabrication process is not as simple as doing an Image -> Resize, 99% of the time it will not work.. So you can't really take the SID die, X-Ray it, then try to replicate it on a smaller fab.

No of course not. But we would have VHDL code for the chip, see my first post. I don’t want to „photocopy“ existing chips, but people write VHDL descriptions of old chips all the time. Simply pour those into silicon, instead of emulating them or putting them into an FPGA.

Other than replicating dying chips, I'd love to see what a die shrinked 386 could be capable of x)

root42 wrote on 2020-06-18, 12:59:

appiah4 wrote on 2020-06-18, 12:53:

Yes, but shrinking a die to a smaller fabrication process is not as simple as doing an Image -> Resize, 99% of the time it will not work.. So you can't really take the SID die, X-Ray it, then try to replicate it on a smaller fab.

No of course not. But we would have VHDL code for the chip, see my first post. I don’t want to „photocopy“ existing chips, but people write VHDL descriptions of old chips all the time. Simply pour those into silicon, instead of emulating them or putting them into an FPGA.

Ah - I thought the discussion about the incomplete SID VHDLs was just an introduction, sorry. In that case you could probably shrink it down to as much as 1/1000 of the original 😁 The issue is, it would probably not be perfect at first attempt, you would need to do a few revisions before taping out a gold sample (you can't just bodge a chip..), and the cost would just keep adding up 😀

If you want to replicate the original IC in hardware I still think doing so with FPGAs would be more cost effective, there are some really cheap FPGAs out there today that can easily handle what a SID does..

@appiah4: yes, I agree. This would be the Deluxe solution for sure. And I am far from sure that there is a market for that. On the other hand I have seen so many crazy Kickstarters, way beyond the 150k USD barrier... And I think a "true" SID would be so much more worthwhile 😀

Plus once you have a working VHDL, you could crank out more chips. The SID at least as some fan following e.g. in the music scene.

root42 wrote on 2020-06-18, 13:12:

@appiah4: yes, I agree. This would be the Deluxe solution for sure. And I am far from sure that there is a market for that. On the other hand I have seen so many crazy Kickstarters, way beyond the 150k USD barrier... And I think a "true" SID would be so much more worthwhile 😀

Plus once you have a working VHDL, you could crank out more chips. The SID at least as some fan following e.g. in the music scene.

The SID is also one of the most failure-prone ICs of one of the most popular 8-bit micros, so if you produce it in volume and price it reasonably you would probably have a steady stream of income. The issue is probably the startup investment. If you can complete the VHDL, test in in an FPGA, then design the chip and launch a Kickstarter, you might break even on it even if you can't make a lot of money from it. Of course, that requires good chip design and getting it right on the silicon without too many revisions.

By all means, go for it. I'd back you up. Though, not for more than what an ARMSID costs, to be sure 😁

The problem is: I have NO knowledge of VHDL or chip design. I hear it's not that much harder than regular coding, but I have 0 background in that. But I would also back such a project. 😀 And the SID would probably be one of the best examples where it actually makes sense. Stuff like the PLA will have a hard time getting a proper replacement, especially now that the GALv8 is out, which is so cheap, even though it looks weird, it doesn't really matter.

SID has half and half digital and analog components so doing it in all digital is hard to do unless idea is split the chip into two chips solution and have pin compatible on a daughterboard.

Cheers,

What does that even mean? I heard about the SID being "half analog, half digital". However in the end all you can put on an IC (to an extent) are transistors, diodes, capacitors and resistors. And you can build analog or digital circuits with that. Anyhow, if there are any parts that can't modeled by "digital" circuits, you can probably still put something digital on there that emulates that behaviour. Be it a DAC or something else.

Well, there is a new option:

https://www.theregister.com/2020/07/03/open_c … 9Xw5_HQVPtKx904