Given the recent pricing increases on de10-nano, and the fact its fpga is nearly at capacity with larger cores, has anyone looked at porting the mister cores to something like an Ultra96 or even better the $200 kria k26 kit? The kria has pretty much everything you could want. 4GB (!) of ddr4, hdmi and dp 1.2, 4 usb A and some bits of gpio. It also has 256k xilinx le's, which aren't the same as altera le's but still, it's a lot more.
e: Looking at this it'd be tricky to add a low latency SDRAM/SRAM cache. there's a lot of bram though
Xilinx Zynq port? Kria K26
Re: Xilinx Zynq port? Kria K26
I don't see any GPIO on the K26, so adding I/O and special memory hats or shields looks a bit tricky.
The Ultra96 looks better on paper, but without a direct ASIC gate to ASIC gate comparison (which both companies' marketing departments would object to, as it makes pulling the wool over customers' eyes harder), the only way to know for certain whether this would be a true upgrade would be to use Xilinx tools to port HDL softcores of various computer and console systems and see how much space in the FPGA each takes.
The Ultra96 looks better on paper, but without a direct ASIC gate to ASIC gate comparison (which both companies' marketing departments would object to, as it makes pulling the wool over customers' eyes harder), the only way to know for certain whether this would be a true upgrade would be to use Xilinx tools to port HDL softcores of various computer and console systems and see how much space in the FPGA each takes.
Re: Xilinx Zynq port? Kria K26
Depends. The $199 kit's carrier board is a little lacking though it has PMODs that break out some GPIO. Someone could make a better carrier board tuned to MISTer specifically: display, networking, usb and enough low latency ram to implement all cores that can't handle ddr4 latency.
Ultra96 is more comparable to Kria, being a xilinx product too. It's a smaller and less advanced fpga, and no hdmi, but you could use an active converter. More gpio and there's a lot of 96board accessories for IO already. Kria is just significantly larger than the de10-nano so it might actually be worth the effort.
Perhaps more effort than it's worth. How many cores can run fine on ddr4 anyway?
Ultra96 is more comparable to Kria, being a xilinx product too. It's a smaller and less advanced fpga, and no hdmi, but you could use an active converter. More gpio and there's a lot of 96board accessories for IO already. Kria is just significantly larger than the de10-nano so it might actually be worth the effort.
Perhaps more effort than it's worth. How many cores can run fine on ddr4 anyway?
Re: Xilinx Zynq port? Kria K26
The Kria SOM itself is really interesting, it actually has more GPIO than the DE-10. It also has over 4x the BRAM as the SOC on the DE-10.mxml wrote: ↑Tue May 04, 2021 1:14 am Depends. The $199 kit's carrier board is a little lacking though it has PMODs that break out some GPIO. Someone could make a better carrier board tuned to MISTer specifically: display, networking, usb and enough low latency ram to implement all cores that can't handle ddr4 latency.
Ultra96 is more comparable to Kria, being a xilinx product too. It's a smaller and less advanced fpga, and no hdmi, but you could use an active converter. More gpio and there's a lot of 96board accessories for IO already. Kria is just significantly larger than the de10-nano so it might actually be worth the effort.
Perhaps more effort than it's worth. How many cores can run fine on ddr4 anyway?
And of course, it also has a GPU and extra ARM cores that could be used for menus and such, which is constantly being asked for on the DE-10 at this point.
Writing a framework for this thing would be the same gargantuan effort as it was for the DE-10, so it's dead in the water unless Sorgelig is on board. And I can't imagine him being on board unless the list of things that would run on this SOM and not the DE-10 got long. At this point that list is short, probably better ao486 and Amiga cores. Possibly N64?