Here’s a heads up I am quite happy to be giving: today is going to be an Amiga/BeOS/Atari day on OSNews. Let’s start with this story about converting an Amiga 600 to a FPGS-based emulation machine.
That said, a couple of months ago I ran across the MiSTer FPGA project spearheaded by sorgelig. This project is based on the Terasic DE10-Nano board which has a decent sized Altera Cyclone FPGA paired with a dual-core ARM Cortex-A9 CPU. Sorgelig has designed a number of add-on boards that allow the DE10 to interface with additional devices. He also has ported (and improved) many cores for this board, including the Minimig-AGA core which provides a very nice recreation of the Amiga from a 500 to a 1200.
After buying a DE10 and getting the Minimig-AGA core running on it, I was immediately infatuated with the quality of “emulation” on this thing. It felt much more complete than the UAE4ARM and Amiberry emulators and the video quality looked much nicer. Not to mention, the near-instant power on (and off) felt more like a real Amiga. Following in the footsteps of my previous Raspberry Pi conversions, I decided to convert an Amiga 600 to FPGA as the 600 case fits so nicely on my desk.
As the author notes, this is not a simple or straightforward mod, as there’s 3D printing involved. Still, it’s a fascinating process to document.
https://shop.inventronik.de/index.php?id_product=8&controller=produc…
A bit expensive, but dude, what an unlocked achievement.
Kochise,
Yeah, right now it’s too niche. I’ve been saying it for a long time, FPGAs open up a new world of possibilities for things that are bottlenecked running on typical CPUs. I’m waiting for intel, AMD, or even ARM vendors to incorporate a real FPGA to their CPUs (either on die, or a co-processor).
Obviously not many people care about simulating old processors, but the beauty of FPGAs is that there are tons of applications like compression, crypto, AI, voice recognition, audio/video synthesis, real time processing, etc. And even things that we haven’t even thought of yet. But until FPGA tech is built into commodity hardware, the masses won’t get to experience it. Surely this comes up in engineering meetings at intel/amd/etc, but I don’t know if they can mass produce these at price points that ordinary consumers would be willing to pay.
Yeah, talked a lot about possible inclusion of FPGA into Atari-like industrial computers : http://www.yaronet.com/topics/45659-atari-68kpowerpccoldfireemulation
I’d guess that fuller integration of GPU might happen sooner, as a coprocessor to CPU (that’s what AMD buying ATI was ultimately about, right?) in the vein of x87…
PS. @ssokolow, here we have (emphasis mine) “including the Minimig-AGA core which provides a very nice recreation of the Amiga from a 500 to a 1200″ so the word usage in favour of which I spoke recently ( http://www.osnews.com/comments/30528 ) gains some traction already
(also, a bit further down, “I was immediately infatuated with the quality of “emulation” on this thing” – emulation in quotes)
Alfman: Regarding the previous MISTer thread about 8/16 and 32-bit cores. Here is the 486 CPU core for MISTer (missing the FPU module): https://github.com/MiSTer-devel/ao486_MiSTer I would love to see an old SGI MIPS system implemented on an FPGA.
Darkmage,
This highlights why MIST FPGA wasn’t designed for 32bit CPUs. Despite the significantly higher clock speed for the FPGA, the 16bit limitations within the MIST configuration mean more passes are needed, reducing performance.
Having an entire 486 system (minus floating point) on this relatively “cheap” FPGA is a cool accomplishment, but to match the performance of the original they’d probably have to upgrade the FPGA/memory to eliminate the 16bit bottlenecks.
On the other hand, theoretically they might be able to implement a 486 CPU using a significantly more complex superscalar pipeline that attempts to make up for the 16bit limits by starting the next instruction even before the previous instruction is finished (somewhat like pentium’s U and V pipelines). This adds tons of complexity that the original 486 didn’t have, which raises a new issue: the cyclone FPGA used by MIST is already at 80% logic capacity, it may not be possible to add a second parallel pipeline without upgrading anyways.
That Suska board is very expensive and there doesn’t seem to be a large community around it.
If you want an Atari ST FPGA go with the MIST:
https://www.lotharek.pl/
About $250 and can run many cores.
I have a MiSTer and find it to be a lot of fun. I have an Amiga 1200 shell/keyboard with a Keyrah and use it with the MiSTer. More expensive, but way better than a Raspberry PI. (not as versatile as a full fledged computer running WinUAE/other emulators though).
The FPGAs are great but development of new cores and updating old cores tends to be difficult because you often need expensive tools like Quartus to compile.
I highly recommend the MiSTer if you want to get started with an FPGA system that has a lot of cores. Sorgelig is a one man army here as he is doing most of the core porting/creation and hardware design work (there are a few others that are now getting involved).
You will need the Terasic board (about $130) and at a minimum the SD-RAM board (about $35). See links below for more information.
http://www.atari-forum.com/viewforum.php?f=119
https://github.com/MiSTer-devel/Main_MiSTer/wiki
Edited 2018-07-09 16:37 UTC