Amiga 68030/40?

[ericgus09]

No reason? Please do speak for yourself when you express such sweeping qualifications as fact. Apart from big box Amiga, there are Macs and Atari systems that would be brought much closer to a MiSTer implementation with a fully featured 68030 or 68040, preferably at realistic period appropriate speeds.. but that's just my take.

Agree 100%.. and even within the Amiga universe, not everyone only uses their Amigas to play Turrican and Lemmings.. some of us use it for other non-gaming activities .. Honestly gaming is the least of the things I do on my various amigas.

[rhester72]

No reason? Please do speak for yourself when you express such sweeping qualifications as fact. Apart from big box Amiga, there are Macs and Atari systems that would be brought much closer to a MiSTer implementation with a fully featured 68030 or 68040, preferably at realistic period appropriate speeds.. but that's just my take.

I said what I said. Based on what you've said, what you REALLY want is a MMU...and that is a shade of a very difficult color (in terms of complexity).

[Caldor]

the Vampire cards does not have an ARM component, I am not sure it will ever be easy to swap out.

No need for ARM, there’s the FPGA and it can run another 68k core for the Minimig firmware, take care of I/O, menus etc. This is how it works on the TC64 already

The problem I pointed out remains the same though... the Minimig and TC64 cores are not designed to work with real Amiga hardware. They would need to be modified quite a bit to work as a Vampire replacement. It would be possible, but it would require the Vampire developers to want to do it and currently that does not seem to be the case. The TC64 core does do what you say, but it is also designed around the TC64 specifically, which is VGA output and such. It does also support actually swapping cores on the TC64 just from the SD card. With the Vampire 2 cards, the native Amiga output is still used for non-RTG stuff, and I do not think it would be possible to have an FPGA menu there and such. It should be possible to have such a menu on the RTG output, and the Vampire 4 does have all the video out on the HDMI of the vampire, and the audio output is there as well. That should give some more freedom to do stuff like an overlay FPGA menu and such.

But again, while its possible, it needs someone to develop it.

[Caldor]

No reason? Please do speak for yourself when you express such sweeping qualifications as fact. Apart from big box Amiga, there are Macs and Atari systems that would be brought much closer to a MiSTer implementation with a fully featured 68030 or 68040, preferably at realistic period appropriate speeds.. but that's just my take.

Why realistic period appropriate speeds? Have you tried testing in WinUAE? I doubt any 040 software relies on the 040 CPU being slow. That should only have been an issue for 68000 and maybe some 020 software and games.

It takes extra effort to limit the speed and make it some exact speed in these cores. Which is why f.ex. the PSX core is still a bit faster than the original PSX rather than the exact same speed, and the argument for that not being a priority is because PSX software was developed after it had become common practice to not assume the CPU to remain at a static speed.

As for whether 030 or 040 instructions are required... I do not especially know. But the difference in architecture seems to be required to get more speed out of the MiSTer in this case I think. Making it faster with an 020 might not be possible, and even if it is, I do not think the gain would be significant. Stuff like MMU and FPU would probably make more of a difference, but pretty sure MMU needs more than an 020 CPU.

[ericgus09]

the Vampire 4 does have all the video out on the HDMI of the vampire, and the audio output is there as well. That should give some more freedom to do stuff like an overlay FPGA menu and such.

But again, while its possible, it needs someone to develop it.

When pressed on this idea of doing a mister like overlay to configure the vampire, Gunnar seems to be very much against this as "its not how the amiga does things" .. if you notice any of the vampire configuration/utilities etc are all amiga command line executables .. I think they are purposefully NOT doing on screen type overlays ala Mister or other popular FPGA platforms do.

[Bas]

I would prefer historic accuracy over maximum speed because software emulators will outrun the FPGA anyway. The fastest Amiga ever would be WinUAE anyway. MiSTer would be the place for accuracy.. kind of what Jotego does with his arcade cores. I can imagine that priority for this kind of development is low, but the interest surely isn't zero.

[akeley]

I hope one day it will be possible to have all the cores as close to the real thing as possible, even if not 1:1 - warts'n all. This would assure maximum compatibility and also contribute greatly to the "using real hardware feel". Not just for Amiga, but many different micros as well.

Having a separate core for every model would be great too, but I don't really mind having a unified one with loadable presets (like Minimig now).

[Bas]

I can imagine people getting sick of the "yes but conservation!" yadda yadda.. But have any of you ever tried to find an Amiga 3000 in any sort of viable state that won't break the bank? Most of the ones that I see go for more than EUR 1000 in questionable state. Jotego has proven time and again that painstakingly accurate FPGA cores are possible for arcade systems that would run rings around the Amiga specs-wise so an A3000 or 4000 should fit comfortably. As a niche-within-a-niche, I still very much think that MiSTer is the platform to faithfully simulate vintage hardware and the big box Amigas are certainly worthy of the honour. If not on their own merit, then at the very least as a gateway to the Atari TT and Falcon, NextCube, or bigger Macs like the Centris or Quadra lines, Sun 3/80 workstations or even some early SGI gear. Not everything needs to be about gaming all the time.

[Hodor]

[...]If not on their own merit, then at the very least as a gateway to the Atari TT and Falcon, NextCube, or bigger Macs like the Centris or Quadra lines, Sun 3/80 workstations or even some early SGI gear. Not everything needs to be about gaming all the time.

Getting an accurate core of Atari TT, Falcon, 040 Macs or NextCube would be a major breakthrough to Mister comparable to PSX, Neogeo, CPS-II or similar cores. Unfortunately, the goal is very hard to reach.

[Bas]

Agreed, but the 68030/40 implementation is one essential building block they all have in common. A block that could be relatively easily put to productive use for the Amiga platform, as most of the other work is already in Minimig.

[djhardrich]

The entire time everyone has been arguing/theorizing, these have been sitting quietly: https://github.com/AmicableComputers/wf68k30L and https://github.com/svenschnelle/hp300_MiSTer , the second project is a core that uses wf68k30L. Cache, FPU, and MMU are currently not implemented, possibly worth taking a look at implementing CPU cache even if SoftMMU is still used... If integrated, it should definitely be an -option- though, fx68k is very solid.

[bazza_12]

i wouldn't mind seeing an 040.. for an A4000 core.. i had an A1200 in the early 90's but dreamed of having an A4000/040.. so far out of my reach money wise.

[throAU]

I guess the question is: what will a 68040 MiniMig (or hybrid arm minimig for that matter) give me that WinUAE will not?

Speed? Doubtful. Compatibility? Nope. Cycle accuracy? Nope.

I get it. “It’s a native cpu”. But really it isn’t. Meanwhile with WinUAE I’d be getting snapshots, shared file system with the host pc that can provide high speed networking , high speed SSD storage, etc.

Maybe a cycle accurate a3000 or a4000 would be nice for hardware preservation sake. But for performance? Pointless

[ericgus09]

I guess the question is: what will a 68040 MiniMig (or hybrid arm minimig for that matter) give me that WinUAE will not?

you dont need a PC to run it.

I get it. “It’s a native cpu”. But really it isn’t.

Actually it is, the CPU is created within the logic gates of the FPGA and it is, for all purposes a CPU .. its just in reprogrammable silicon rather than "etched" permanently.

[Bas]

I don't want a host PC with an OS I need to pay for, patch and pamper. The usability of the MiSTer is awesome: pick core, it reconfigures itself into the spitting image of that particular machine and you can have at it.

[throAU]

I guess the question is: what will a 68040 MiniMig (or hybrid arm minimig for that matter) give me that WinUAE will not?

you dont need a PC to run it.

I get it. “It’s a native cpu”. But really it isn’t.

Actually it is, the CPU is created within the logic gates of the FPGA and it is, for all purposes a CPU .. its just in reprogrammable silicon rather than "etched" permanently.

1. you still need a pc or mac to maintain mister with.
2. it’s not a native cpu as some of the spec proposed never existed as real hardware.

if you care about speed, emulation is getting faster as new pcs come,and the FPGA won’t help your io keep up.

[ericgus09]

if you care about speed, emulation is getting faster as new pcs come,and the FPGA won’t help your io keep up.

Well thats just it though.. it really depends on whats important for you .. and specifically you .. (it will vary depending on who you ask)..

FPGA can help you with an accurate re-creation as it can work in true parallel, something software emulation cant .. its just sequential "round robin" .. all be it very fast .. but still sequentially single threaded for the most part.. and even that has its own inherent drawbacks.. That said, I dont think one is "better" than the other.. they both have exceptionally valid use cases, but it depends on what your end goal is and your general parameters in getting there are.. its good we have options, for raw speed, yes software emulation is great, its many many times faster and say for setting up a new HDF file its hard to beat the speed of it on a modern computer .. again, it really depends on what you are trying to accomplish.. its a good thing we have a lot of options and toys to play with to get there.

[kolla]

In context of Amiga, FPGA is great for implementing the chipset, and awesome for software that depend on the native Amiga chips. And that’s it. Once you leave the chipset behind and fully go P96 and (to a degree) AHI, software emulation is just better. Yes, with software emulation, “everything” happens sequential and not in parallel, but that is also true on a real Amiga with RTG/RTA.

[Caldor]

I guess the question is: what will a 68040 MiniMig (or hybrid arm minimig for that matter) give me that WinUAE will not?

Speed? Doubtful. Compatibility? Nope. Cycle accuracy? Nope.

I get it. “It’s a native cpu”. But really it isn’t. Meanwhile with WinUAE I’d be getting snapshots, shared file system with the host pc that can provide high speed networking , high speed SSD storage, etc.

Maybe a cycle accurate a3000 or a4000 would be nice for hardware preservation sake. But for performance? Pointless

The Minimig core does have pretty high speed storage access though, sure its not as fast as WinUAE, but its faster than f.ex. the Vampire, which is quite a bit faster than Amigas originally would be. At least for the SD storage, not completely sure about the IDE storage.

The compatibility should also be good I think? I mean, compared to a real 040 Amiga. It obviously depends on how its implemented. And the speed? I was pretty sure there would not be any speed optimization, until I found that part of the speed increase between the different Amiga CPUs actually comes from the architecture being different... which is why the Vampire 080 core can run as fast as it does, despite having the same FPGA as the MiSTer.

How much faster would the Minimig core be on the MiSTer with an 040 CPU be? No idea, it might even be slower. Hard to know without someone implementing it. A 030 might be more interesting though, now that the CDTV has a new ROM released that supports 030. Then again, CDTV stuff could probably overall be done with the 68000 CPU that the Minimig core already supports. I already got some games running with that.

[Malor]

I guess the question is: what will a 68040 MiniMig (or hybrid arm minimig for that matter) give me that WinUAE will not?

Speed? Doubtful. Compatibility? Nope. Cycle accuracy? Nope.

I get it. “It’s a native cpu”. But really it isn’t. Meanwhile with WinUAE I’d be getting snapshots, shared file system with the host pc that can provide high speed networking , high speed SSD storage, etc.

Maybe a cycle accurate a3000 or a4000 would be nice for hardware preservation sake. But for performance? Pointless

The biggest things are near-zero latency and exact duplication of the original video modes onto a CRT, like a 1084S. The HDMI is pretty low latency, too.

WinUAE has implemented a 'beam-racing' mode that renders each frame in chunks, only a little ahead of the video scanline, so it reduces latency as far as it can, but the OS, video card, and monitor will all add some. It's way better than most emulators, but it can only do so much.

With an FPGA directly driving a CRT, the video output is electrically linked to the scan beam, so it literally can't be any better. You can have controller lag, but that's mostly linked to what controller you buy. The best wired controllers are at about a millisecond.

[throAU]

The near zero latency is great for gaming. Which is where you want a cycle accurate a500 or a1200.

Beyond that? Games from the period were not written for anything higher spec than that. And any games written later for higher spec are better played on pc. Where speed, graphics quality and audio quality will be objectively superior.

Applications generally aren’t cycle accuracy dependent and WinUAE will be faster.

[Bas]

Why is it so difficult to comprehend that WinUAE could be as fast or as fancy as it wants to be, it's just not an alternative if it doesn't live on the MiSTer? It's a completely different animal.

[Caldor]

WinUAE is emulation and while it might actually end up with nearly the same latency, there are other advantages of the MiSTer:

Boot time
A fraction of the power needed
Highly portable
Specialized, no booting into some OS to then boot WinUAE, load some configuration...
All in one place, also works for other systems
Simpler settings than WinUAE, always cycle accurate

Debating whether its needed is pretty pointless. People want to have it, so the question is whether someone will develop it.

[LamerDeluxe]

The near zero latency is great for gaming. Which is where you want a cycle accurate a500 or a1200.

Beyond that? Games from the period were not written for anything higher spec than that. And any games written later for higher spec are better played on pc. Where speed, graphics quality and audio quality will be objectively superior.

Applications generally aren’t cycle accuracy dependent and WinUAE will be faster.

Actually filled vector games on the Amiga have always benefited a lot from faster CPU's. Racing games, flight simulators, games like another world. And basically any game strapped for CPU power. Most adventure games played better with a faster CPU. Not all of the later, more demanding games are available (or better) on a PC. Like the excellent Genetic Species game, for instance.

I do agree that exact CPU speed probably doesn't matter with anything faster than a 68020, as I don't expect any games utilizing the extra speed to rely on exact timing. That said, there is a case to be made for accurate replication of hardware timing, for preservation purposes.

I had a turbo card (as they were called) for my Amiga 2000 early on, as I was studying 3D animation at the time. Went from 68030/68882 to an Amiga 4000 with an 68040 and upgraded that to a 68060/PowerPC.

[Malor]

The near zero latency is great for gaming. Which is where you want a cycle accurate a500 or a1200.

Beyond that? Games from the period were not written for anything higher spec than that. And any games written later for higher spec are better played on pc. Where speed, graphics quality and audio quality will be objectively superior.

Applications generally aren’t cycle accuracy dependent and WinUAE will be faster.

In general, telling people they shouldn't want or enjoy the things they want or enjoy is not very productive.

For people who loved the Amiga, there are tons of both games and applications they like to use, and particularly with a CRT, the overall system can end up feeling almost exactly like the real hardware. The 68020 in the Mister, however, is substantially slower than what you could get on, or retrofit to, real Amigas, so some people would like to take it faster.

Yes, WinUAE is available, and I'm glad it is, because I use it. But the FPGA approach is nice too, because it's low latency, feels more authentic, is generally very simple to work with, and generally shares about the same interface across a whole bunch of different cores. Flipping from, say, the Amiga to the Apple II takes literally seconds, and then all the configuration and disk loading works about the same way between the two cores. And then if you want to mess around with the Archimedes, that's all true again. And then the 486, or the Genesis, or the SNES.....

Also, whoever created the MegaAGS pack put a ton of time and effort into it. You can probably use it with WinUAE, but it's primarily intended for the Mister setup, and works extremely well.

[throAU]

For what it’s worth I owned both an a500 and a1200. So I do get the nostalgia thing.

Filled vector games are still best played elsewhere

[Malor]

But with a faster CPU, that would be less true. Perhaps nobody will actually do one, but quite a few software packages would benefit from a faster CPU. For example, one that seemed to benefit from all the CPU you could scare up was Stunt Car Racer, and the Amiga version of that game is the best one.

It seems very weird to me to criticize wanting more CPU on Minimig. It may not happen, but it doesn't seem weird to me to want it.

[LamerDeluxe]

For what it’s worth I owned both an a500 and a1200. So I do get the nostalgia thing.

Filled vector games are still best played elsewhere

Not all of the vector games on the Amiga can be played elsewhere. And with an accelerator a lot of them actually work very well.

[Samurai_Crow]

Would this 4-stage pipelined 68030 help? The open source version is deliberately hobbled by using a repeat shifter instead of a barrel shifter on the instruction fetcher and lacks caches in the open source version as well. The core could also be made smaller by adding a write port to block memory instead of using logic blocks to implement the registers. The closed-source version also implements an MMU. It was originally intended for an Atari Falcon core set on another board.

[rhester72]

@Samurai_Crow tbh the MMU is the most interesting part of the 68030. At an instruction set level, it's pretty much identical to the '020, and from a performance perspective, it's infinitely easier just to ratchet the clock up on an '020 than to try to get more performance out of a more complex implementation.