Amiga.org
Amiga computer related discussion => Amiga Hardware Issues and discussion => Topic started by: Karlos on March 01, 2018, 11:37:38 AM
-
The recent buzz around the vampire on the forums got me thinking.
What FPGA accelerator solutions exist for the A1200 today and is there any idea about when or if we'll see any with Vampire levels of performance?
I miss the days when I had an A1200 in a convenient keyboard wedge that I could (and often did) stick in a backpack and take different places. The blizzppc and gfx card put an end to all that.
An FPGA accelerator with some sort of modern video output would be a beautiful thing.
-
The recent buzz around the vampire on the forums got me thinking.
What FPGA accelerator solutions exist for the A1200 today and is there any idea about when or if we'll see any with Vampire levels of performance?
I miss the days when I had an A1200 in a convenient keyboard wedge that I could (and often did) stick in a backpack and take different places. The blizzppc and gfx card put an end to all that.
An FPGA accelerator with some sort of modern video output would be a beautiful thing.
Like it or not, but there is a Vampire for 1200 planned, but will most likely not come out before the standalone Vampire v4 unit.
-
The recent buzz around the vampire on the forums got me thinking.
What FPGA accelerator solutions exist for the A1200 today and is there any idea about when or if we'll see any with Vampire levels of performance?
I miss the days when I had an A1200 in a convenient keyboard wedge that I could (and often did) stick in a backpack and take different places. The blizzppc and gfx card put an end to all that.
An FPGA accelerator with some sort of modern video output would be a beautiful thing.
Currently there is no available a1200 FPGA accelerator, the "whole" world is waiting for V1200...
-
Waiting for the Vampire as well, although the standalone model interests me more since you don't have to rely on 25 years old electronics to power it.
An "accelerator" that basically sucks the life out of the computer and takes over everything (ie. a vampire) is basically a standalone system that uses the A1200 as a needlessly fragile and complicated power supply and I/O board... IMO of course.
-
Waiting for the Vampire as well, although the standalone model interests me more since you don't have to rely on 25 years old electronics to power it.
Yeah, I'm kinda torn both ways. Having it fit an a500 case, use an a500 keyboard and use a 3.5" floppy drive would be quite nice.
-
is there any idea about when or if we'll see any with Vampire levels of performance?
If there is going to be an FPGA accelerator for the A1200 with the same or similar speed as that of the Vampire, it is going to be a Vampire. It's not that FPGAs are magically fast at running 68k code. The other 68k softcores reach low to mid-range 020/030 class speed (without FPU and MMU, of course) in comparable FPGAs.
-
If there is going to be an FPGA accelerator for the A1200 with the same or similar speed as that of the Vampire, it is going to be a Vampire. It's not that FPGAs are magically fast at running 68k code. The other 68k softcores reach low to mid-range 020/030 class speed (without FPU and MMU, of course) in comparable FPGAs.
Or someone else improves the other 68k softcores. It's not like vampire is magically the only fast FPGA 68k emulator.
Apollo likely has a headstart, but you don't know what people are doing in private.
-
Or someone else improves the other 68k softcores. It's not like vampire is magically the only fast FPGA 68k emulator.
Apollo likely has a headstart, but you don't know what people are doing in private.
Its not an Emulator :-)
-
Or someone else improves the other 68k softcores. It's not like vampire is magically the only fast FPGA 68k emulator.
Apollo likely has a headstart, but you don't know what people are doing in private.
It's impossible to improve the existing other 68k softcores to the level of the 68080. Write a new one, yes, that's always a possibility. Albeit a very unlikely one as there are few microprocessor developers around that would spend several years worth of their sparetime for such a project.
-
it could be nice if the vampire core could be in an asic that shure will be fast
-
Waiting for the Vampire as well, although the standalone model interests me more since you don't have to rely on 25 years old electronics to power it.
An "accelerator" that basically sucks the life out of the computer and takes over everything (ie. a vampire) is basically a standalone system that uses the A1200 as a needlessly fragile and complicated power supply and I/O board... IMO of course.
If you're not using Vampire AGA or Vampire PAULA, the difference between a V1200 and a BPPC + BVision is none regarding this (except for the PPC part). :P
-
Or someone else improves the other 68k softcores. It's not like vampire is magically the only fast FPGA 68k emulator.
Apollo likely has a headstart, but you don't know what people are doing in private.
E.g. the MiST is about the level of a 30 - 35 MHz 68030 and it's running on a Cyclone III, same FPGA chip as e.g. the Vampire V2 A600. Of course the MiST recreates more than the 68k core itself, but so does the Vampire V2.
The Minimig-AGA core has been developed since 2014 at least, so it's strange that performance hasn't improved. Maybe all the focus has been on improved compatibility.
If you're not using Vampire AGA or Vampire PAULA, the difference between a V1200 and a BPPC + BVision is none regarding this (except for the PPC part). :P
It's still not a "real" 68k processor though. I get your point, but for me personally, an FPGA emulator strapped to an Amiga does defeat the purpose of using legacy hardware in the first place. If there's a compatibility issue in the 68k softcore, AGA or Paula, it would affect both the standalone system and the "accelerator" equally so I see no real advantage to not going with the standalone board. I guess you could mount it in your A1200 case if you wanted...
-
The Minimig-AGA core has been developed since 2014 at least, so it's strange that performance hasn't improved. Maybe all the focus has been on improved compatibility.
What kind of performance improvements can you expect from something that is intended to be as equal to original as it can? I honestly cannot tell the AGA on MiST apart from "real" AGA, except from the one well known horizontal shift bug on low screen modes. I understand from Chaos, that going from ECS to AGA was not really that much work (which echoes what I recall CBM engineers also have written before about AGA... a quick and cheap upgrade of the ECS chipset)
Improving on AGA is what SAGA is about.
-
I suppose a starting point would be to take Majasta's V1 600 design (http://www.majsta.com/modules.php?name=News&file=article&sid=71) and convert it for the A1200 (or Kippers PCI style adapter http://eab.abime.net/showpost.php?p=1204931&postcount=10).
Then you'd have a version of the TG68 in an FPGA that stops the builtin 680x0 and has a bunch of ram.
-
so I see no real advantage to not going with the standalone board. I guess you could mount it in your A1200 case if you wanted...
AFAIK the standalone won't have PCMCIA or floppy, but the A1200 will still have these. You might not care, but I might.
If the standalone fit in an A1200 case and supported PCMCIA (and why not cardbus too) as well as internal and external floppy drives then it would be hard to resist.
-
I doubt that standalone will be much larger than vampire card, so it should fit the A1200 case easily. Not sure how would you connect the original keyboard and old IO connectors.
Isn't the point of standalone to have its own case?
-
I doubt that standalone will be much larger than vampire card, so it should fit the A1200 case easily. Not sure how would you connect the original keyboard and old IO connectors.
Isn't the point of standalone to have its own case?
You would use a Keyrah
http://wiki.icomp.de/wiki/Keyrah_V2
-
What kind of performance improvements can you expect from something that is intended to be as equal to original as it can? I honestly cannot tell the AGA on MiST apart from "real" AGA, except from the one well known horizontal shift bug on low screen modes. I understand from Chaos, that going from ECS to AGA was not really that much work (which echoes what I recall CBM engineers also have written before about AGA... a quick and cheap upgrade of the ECS chipset)
Improving on AGA is what SAGA is about.
Yeah, I guess improved performance and perfect compatibility are kind of mutually exclusive, due to the need to maintain exact timings and so on. I wouldn't mind a "Turbo" Amiga core for the MiST, with all stops removed, for those few demanding 3D games etc., but that's why I'm interested in the Vampire V4 as a complement to my MiST.
-
Yeah, I guess improved performance and perfect compatibility are kind of mutually exclusive, due to the need to maintain exact timings and so on. I wouldn't mind a "Turbo" Amiga core for the MiST, with all stops removed, for those few demanding 3D games etc., but that's why I'm interested in the Vampire V4 as a complement to my MiST.
Compatibility has very little to do with timing. If a program is tied to the timing of a CPU, then it's going to have a bad time running on any accelerator. If it's written to use the Amiga's timers, then exact timings in an FPGA are unnecessary.
Only very poorly written software would be affected, so you'd probably boot off of the m68020 for those. Or, as you wrote, having a "compatible" slow mode and a normal, fast mode would be a good thing.
-
Compatibility has very little to do with timing. If a program is tied to the timing of a CPU, then it's going to have a bad time running on any accelerator.
That still counts as compatibility, the only way to really solve it is to allow the speed to be dropped.
Back in the 80's you could buy PC's that had a turbo switch, if you turn it on then the cpu runs at normal speed or turn it off and it slows down (to as slow as 4.77mhz iirc).
You can fix the software and a lot of it has been fixed when using whdload.
-
That still counts as compatibility, the only way to really solve it is to allow the speed to be dropped.
Back in the 80's you could buy PC's that had a turbo switch, if you turn it on then the cpu runs at normal speed or turn it off and it slows down (to as slow as 4.77mhz iirc).
You can fix the software and a lot of it has been fixed when using whdload.
Dell PCs had a "compatible" stetting in the BIOS that did the same thing but well into the pentium era.
-
If the standalone fit in an A1200 case and supported PCMCIA (and why not cardbus too) as well as internal and external floppy drives then it would be hard to resist.
I don't think I'd hold my breath on PCMCIA.
Seems like everything that could be done with it could be done just as well (or perhaps even better) with a built in contemporary like SDCARD or ethernet or externally via USB.
The A1200/600 PCMCIA ports are not really internal anyway.. they stick out almost the length of the card :)
-
Seems like everything that could be done with it could be done just as well (or perhaps even better) with a built in contemporary like SDCARD or ethernet or externally via USB.
Everything you can do on an Amiga can be done just as well on a PC, but we're still here. Your argument doesn't really work.