MiSTer FPGA Forum

Quick question:

I read "there is not enough room in the FPGA" a lot on this forum, usually when the question is to emulate systems from 1997 or newer. (n64, dreamcast, etc)

The DE-10 has its limits, but could those limits be broken if a second DE-10 nano is added to the stack?
Some form of parallel processing? and would it make a real difference if possible ?

This is just a banter thread, I dont know if its even possible.

I'm just curious

not an expert here, but i would think this would introduce a lot of additional (needless?) complexity to the system.
as the cores would need to interop with eachother there would a need for a highspeed communication bus which the de10 doesn't feature i believe?

only way i could imagine it would be a totally new board with the two processors close to eachother designed around that whole idea or a new board with a fpga that has more logical blocks?

and what about in terms of costs, it would drive it way up high and render it unaffordable for most people.

from what i've gathered Sorgelig wanted a modular and affordable system for all people where they have a choice what they want / need.

held wrote: ↑Sat Apr 30, 2022 9:52 am Quick question:

I read "there is not enough room in the FPGA" a lot on this forum, usually when the question is to emulate systems from 1997 or newer. (n64, dreamcast, etc)

The DE-10 has its limits, but could those limits be broken if a second DE-10 nano is added to the stack?
Some form of parallel processing? and would it make a real difference if possible ?

This is just a banter thread, I dont know if its even possible.

I'm just curious

Not possible, sorry. For a multitude of reasons. As repeated over and over again; N64 and the more modern systems need considerably more memory bandwidth and their custom silicon are orders of magnitude more complex. The sort of interconnection you are talking about just isn't feasible either. The N64 will happen, no doubt, but on a more advanced FPGA. Dreamcast and beyond are simply not in the realms of possibility for now as a) you would need a large specialist team of hardware engineers and b) it is arguably not worth the time (we are potentially talking YEARS) and effort (with little to no documentation). Commercially these engineers are worth their weight in gold and these are not trivial undertakings.

You can probably connect two DE-10´s.
The problem is that you have too connect the two FPGA´s to gain anything significant.
Connecting the two ARM cpu´s would be possible but you would not gain anything for complex cores like N64 that way.
If you do not use ARM emulation that is but then you will have one to two FPGA chips doing nothing instead and a pi/etc would be better.

And the problem with connecting the two FPGA´s is that you would have to use the pins that the memory and analog Io boards use and that would probably not be enough pins anyway.
Which in the N64 case will mean you have even less memory to use then you would have with only one DE-10!

Too bad, I also have two toothbrushes.. but they do not make me brush faster.

How about offloading parts that need not be in the FPGA to an RPI4 module via the USB/IO connection or would this be too slow ?

This is already done with the MT32-Pi over the IO port, but in that case it doesn't need a super fast (Mhz) connection.

Connecting two DE-10 together via GPIO is technically doable but I don't know what kind of speed you could get out of it.
There would be a limitation in terms of distance, and if you cram two DE-10 close together it may be harder to dissipate heat.

You could perhaps use it to make one DE-10 a dedicated video upscaler (and overall video processor for VGA, composite, SVideo, etc), freeing up resources on the core DE-10. But I don't think you could use it to make a "discrete" GPU since that would require a faster bus. Perhaps there would be enough pins to also move audio processing to the second board.

It would require rewritten cores to reassign the pins so the whole endeavour might end up as a separate fork. All of that for the dubious benefit of saving some LEs... it's probably not worth the hassle.

A more fun thing to do would be to network two DE-10 together via link cable; the Gameboy core a priori already allows this, and in theory the PSX core could do it for some multiplayer games. For computer cores it's probably simpler to just use Ethernet, though.

The DE10-nano has the FPGA and the HPS connected over a PCIe bus, which as far as I know is internal to the Cyclone V chip and not exposed to users. If there was a way to access that bus, then presumably it would be possible to add additional hardware.

The thing is, at this point when we're talking about chaining together multiple DE10-nano boards it would probably be easier to produce a custom board. That's how this all started, the original MiST board was a custom FPGA board meant to replicate the Amiga and the Atari ST. MiSTer was originally started because Sorg saw limitations in how MiST was designed and realized a Commercial Off-The-Shelf board paired with custom daughter boards could do what MiST does but better. Now that we're reaching the limits of what the De10-Nano is capable of it's probably worthwhile to go back to looking at what a custom board could do. It wouldn't even have to be a single-board computer, it could be designed as a custom PCIe card that could be installed in a regular computer, combining the advantages of FPGA and the advantages of software emulation.

Mr. Encyclopedia wrote: ↑Mon Jun 27, 2022 6:35 pm ... it could be designed as a custom PCIe card that could be installed in a regular computer, combining the advantages of FPGA and the advantages of software emulation.

That's a really bad idea. MISTer is self contained and can be carried around in your pocket, no need to lug a PC around.

jca wrote: ↑Mon Jun 27, 2022 6:41 pm That's a really bad idea. MISTer is self contained and can be carried around in your pocket, no need to lug a PC around.

I for one would like a pci-e card that encompasses a large fpga and fast memory for the cores?
It's not like i carry my MiSTer all over the place so it would be all good for me, modularity on that card would be harder then i guess.

Though probably this won't happen very soon, i however would be quite open to the idea depending on the price ofcourse.
Seeing that the DE10 nano was heavily subsidized this could also mean that the pricetag would be quite steep.

All heavy speculation ofcourse.

An ITX computer can be pretty small, but I understand the objection. A MiSTer successor could just as easily be a custom SBC with a high end ARM chipset for graphical power and an FPGA for retro fidelity. But that would probably be quite a bit more expensive than a board meant to slot into an existing PC.