New FPGA Hardware Design Ideas

[dobredanielstelian]

Hello All,

I want to design a board for Mister that incorporates all the peripherals in a mITX form factor.
I was thinking of using Xilinx FPGA XC7A200T (>200k logic cells)

I am waiting for your feedback on the other peripherals, RAM capacity, etc.

Best regards,
Daniel

[FPGA64]

if its not using a DE10 its not a Mister

[MostroW]

personally i'd want to see a pci-e card for my pc that allows me to run cores locally on my OS, depending on how well things would work concerning input latency and such.

[softtest9]

200k logic cells is a considerable improvement. But is it enough of an improvement to enable developers to make more demanding cores? We know that Nintendo DS and Nintendo 64 won't work on the (de10-nano) MiSTer. And of course Dreamcast is out of the question. The MiSTer is beginning to show its age, so I think that for any replacement to catch on, it would need to address this.

[Sectorseven]

Are you trying to build a new design that incorporates the add on boards or a successor board to the whole project?

If it's the latter, you should have some kind of target platform in mind e.g. "It should be capable of supporting Dreamcast" (or whatever).

Then you should consult someone with a strong understanding of that hardware and what it would require using off the shelf components.

Probably something Intel Cyclone based would make the most sense as that is what MiSTer already uses.

[Lukage]

There is a reason for using 5CSEBA6 Altera/Intel FPGA. Because it is used in education kit, this part is supported in free version of toolchain ( Quartus Lite), and the price of the whole kit/board is (was before chip crisis) nowhere near of unit price of FPGA chip itself. Also, it is a SoC FPGA with ARM cores.

Artix-7 7A200T does not incorporate ARM cores, all source would need to be rewritten using Xilinx IP cores, and even if it is supported in free (can't remember) version of Vivado, free version still needs licence and registration.

To be able to implement some more advanced machines/consoles, you would need to go for Xilinx Zynq Ultrascale+ SoC FPGAs, if I were to stay at Xilinx as a vendor. You would need to aim for 300/350MHz+ for FPGA logic timings and multiple channels of fast RAM. For example, PS2 ran at 294.912 MHz and had 3.2 GB/s memory throughput.

That being said, mini-itx board with 5CSEBA6 incorporating better I/O than implemented on MiSTer I/O boards (because of limited number of pins on headers) with all bells and whistles would be a better choice than completely changing platform.

[uigiflip]

Replay two maybe what your looking for but details are still to be revealed. Soon TM @Mike J aka fpgaarcade

[dobredanielstelian]

Hello and Thank you all for your ideas/suggestions.

I have researched further and it seems that Xilinx XC7A200T will not be a solution because it requires a license. The Xilinx XC7A140T does not require a license.
Keeping the same FPGA for the project I think An intel FPGA will be a good solution. Unfortunately the free version of INTEL Quartus Prime Software supports only Cyclone V. An idea is Cyclone V 5CGXC9 FPGA (301000 LE) but it doesn't have an ARM Cortex A9 (I don't know how important this is for the project).
Also how much RAM is optimal for this project and what are the peripherals you want to have on the board.

Thank you.
Best regards,
Daniel

[Mr. Encyclopedia]

personally i'd want to see a pci-e card for my pc that allows me to run cores locally on my OS, depending on how well things would work concerning input latency and such.

I agree with this. If there was a plan to make a MiSTer successor from scratch I'd want to see it as a card that can be installed in a PC. MiSTer was built around a subsidized FPGA board, which had its advantages and limitations. A PC card wouldn't require daughter boards and it would have all the bandwidth it could ever need for access and memory. Most users interested in such a device probably already have a PC that it could be installed in, and beyond that minimal ITX PCs could be specially built to house a FPGA card. The only problem is most of them are very expensive and it's unclear if a purpose-built retrogaming FPGA card could be made cheaper.

Alternatively, it would be interesting to explore the idea of a FPGA daughter board attached to a Raspberry Pi, which could leverage the advantages of the Pi (powerful for its price, lots of developer support, ARM processor and linux to drive the FPGA, HDMI port, small) with the speed and accuracy of FPGA. The only question would be if there's enough I/O bandwidth to make that work. Assuming the daughterboard had plenty of on-board RAM that the FPGA accesses, it might be possible. Such a solution would probably still be significantly more expensive than the De-10 nano, but with the ability to design the FPGA to have all the memory bandwidth it could ever need we would surely be able to make cores well beyond the De-10 nano's ability.

[grizzly]

An idea is Cyclone V 5CGXC9 FPGA (301000 LE) but it doesn't have an ARM Cortex A9 (I don't know how important this is for the project).
Also how much RAM is optimal for this project and what are the peripherals you want to have on the board.

The ARM side of the mister is only doing small parts like the mister OS/GUI/sd card/network/usb/loading roms/and i´m sure i forgetting one or two things but starting to smell burnt sawdust here so i give up for now

RAM some NEOGEO roms need 128MB now, ps1/saturn i do not know if they need 128MB or not but ps1 do not need all of the memory if two 128MB sticks is used.
The one thing with two sticks is to double the RAM speed which helps with very,very minor sound out of sync issues on five games i think.
Bigger/faster cores is another beast completely some will probably need more then 128MB but most will probably need higher speed.
Then comes the problem with speed versus latency, many older cores on the mister needs very low latency which is why the 1gig DDR3 memory on the ARM side can not be used since it is way to high latency but at the same time much faster.
There may very well be a requirement to use two different memory on a new mister one with fast speed one with fast latency.

personally i'd want to see a pci-e card for my pc that allows me to run cores locally on my OS, depending on how well things would work concerning input latency and such.

Yes that would be very cool!
BUT would a PC be fast enough for everything like RAM speed AND latency, and what about USB speed/latency, video/audio latency?
Because windows do add alot of latency to USB and it´s highly depending of how much is connected and used at the same time.
Probably the same is true for audio and video.
We could be closing in on regular emulation latency then and if so would it even be worth it then?

My guess is that for it to be a real FPGA solution and still have ALL the good FPGA goodies left.
It would probably have to be an pcie card with it´s own memory/audio/video/usb ports on it!

[grizzly]

And don´t forget that a mister successor needs to be something that can be bought for years to come after it is released.
So there are two ways to go here.
1, someone builds it and needs guarantees that every part is going to be made for years.
2, Opensource it with very/very common parts that probably will be made for years to come that probably will fail in many ways anyway when parts are stopped being made.

Sadly both are probably 99,999999% impossible to achieve for a project of mister's size as long as no billionaire or big company puts up a looooot of zeros after a one.