Opinion: PPC, 68080 and other

[amifan]

I had a similar situation, I wanted to use an Amiga, but do not have the desk space to fit a real Amiga anymore due to moving to a smaller flat. Emulation was not something that I was interested in, the A500 Mini is nice but it does not feel quite like a real Amiga and the A600GS I am not sure about. However, I had been thinking for a long time about a Vampire V4 Standalone and now after over half a year I am happy I got one. It feels like a real Amiga but has awesome new features such as the SAGA and 16 BIT audio but it still feels very much like a real Amiga. The SD Card slot acts like a floppy disk and with ApolloBoot you can run all the various Workbench versions on one machine. Plus it is a tiny little box which helps if you are limited in space. Admittedly, it is not cheap, but you get a lot of Amiga for the price and improvements are constantly coming.

[Thomas Richter]

Originally Posted by pix

Coming across the 68080 FPGA rekindled my interest, but as I read about the performance I was a bit disappointed and it doesn't run OS4.1 correct? So I have to choose between OS4 and 3.

Os 4 is for PPC, not for 68K, and the 68EC080 FPGA core does not emulate the PPCs, only the 68Ks to some degree.

Originally Posted by pix

I'm not interested in ARM processors at all really.

It's just another processor architecture that is, however, unlike PPC still available in quantities and for budget prices. However, there is no native AmigaOs port for it. It doesn't work any better or worse than the PPC, except that it is in active use.

Originally Posted by pix

I also don't really understand why the PPC and 68080 are bounded by clock speed.

For the PPC, this is simply a matter of supply & demand (as there is none), so they are none made up to decent clock speeds, and even if they would, they would be too expensive to fit into an exotic system. For the 68EC080 softcore - this is an FPGA, and the CPU emulation is loaded at boot time into it. The chip itself is a generic gate array that can do anything you like, though that means higher complexity, less clock speed. FPGAs don't clock as high as specialized chips that were actually build for a particular task.

Originally Posted by pix

Would really hope the Amiga will one day get standardized on OS 4 with either Moto or PPC redesigns at higher power and its own chipsets to really offer Amiga developers a way to flex the differences of an Amiga against anything else.

The PPC Amigas are rather pretty generic PPC machines with an operating system running on them that pretty much reassembles in many aspects AmigaOs, including its problems. There is nothing left of the chipset if that is important for you. The 68EC080 includes an emulation of the chipset, and a partial emulation of the CPU, so that's certainly closer to the original setup, though also limited in other aspects.

Originally Posted by pix

In a world where we have to choose between boring ol x86 and an endless stream of ARM, we need that!

I don't know what "we" is supposed to mean, but different people want different things. There are plenty of choices. Thus, in particular, if you want a really cheap Amiga which is really fast, a software emulator is probably the easiest option - it's also more compatible than existing hardware emulations, certainly a lot more than the PPC Amigas, and faster than existing Amiga hardware.

[Mr-Z]

It's definitely true that emulation is the cheapest&fastest way to run a high end classic Amiga.

But a PiStorm32 with Pi4b/CM4 running at 2.2 Ghz you are already getting WinUAE like speed.
It's safe to say that CPU performance is quite overkill already for most of the existing 68K software base.
WHDload compatibility is extremely good as well, would say not better or worse than on WinUAE.

Yeah it's an ARM processor underneath but who cares, you get a crazy fast A1200 and lots of goodies like DDR4 RAM, SD card up to 25MB/sec, RTG GFX, Wifi and soon FrameThrower for RGB pass through over HDMI.
And not unimportant for an affordable price.

Add support/drivers for 2D&3D acceleration and the hardware h264/h265 video decoder and you have a killer classic Amiga.

[Promilus]

Motorola made 060 last of the classic 68k line long, long time ago. Anyone thinking that it's possible to get something quite like it today is delusional. So what options are there actually?
PowerPC? Not really, it was good idea back then, implementation not so good. Architecture is so far behind and basically only in niche industries which makes it extremely expensive for what it offers.
FPGA? Sure, it is possible and that's exactly what you get with either Vampire or things like MiSTer (obviously V comes ahead with actual CPU performance due to much different softcore).
Emulation on different CPU architecture whatsoever - this is relatively easy to implement and due to big popularity of either x86 or arm makes it pretty affordable as well. Someone doesn't like either of those architectures? Try your luck porting emu68 for RISC-V ...

And if you have lots of money ... make your own damn 68k ASIC.
It actually doesn't f matter what you choose. There won't be sudden revolution which will take Amiga on top and most likely there never will be great new Amiga "NG" platform as well. Just take whatever suits you the best and enjoy without ruining fun for other ppl.

[eXeler0]

Originally Posted by Promilus

8><………
And if you have lots of money ... make your own damn 68k ASIC.

Im guessing that is still the ”secret” goal of the Apollo-core team, to turn the 68080 FPGA into an asic.
Would actually be interesting to see what sort of performance it would deliver at something like 1.5 GHz and also performance/W compared to lets say ARM chips typically found in Raspberry Pi etc.
It would probably show up in accelerators for Amiga, Atari and even some Macs. It could even find usage in some consumer products etc, who knows.

[Bruce Abbott]

Originally Posted by Thomas Richter

The 68EC080 includes an emulation of the chipset, and a partial emulation of the CPU, so that's certainly closer to the original setup, though also limited in other aspects.

Note that 68EC080 is a made up designation created by one of the Vampire's detractors (was it you?).

The 68080 core doesn't include an 'emulation of the chipset', that's done separately in the V4 FPGA.

You say it's a 'partial' emulation of a 68k CPU, but fail to mention that the 68020, 68030, 68040 and 68060 are only 'partial' implementations too.

Doesn't have a (compatible) MMU? Neither does the 68020 (a separate chip is needed for that, and very few Amiga accelerator cards have one). Doesn't have a full FPU? Neither does any other 68k CPU. The 68020 and 68030 have none, and the 68040 and 68060 only partially implement the 68882. The 68060 is also missing some 68020 integer math instructions. Furthermore the instructions relating to CPU functions such as caching are not the same in all of them, so you need to code for the specific CPU in order to be fully compatible.

In reality the 68080 core implements enough of every other 68k CPU used in the Amiga to run 99.9% of games and applications. It obviously doesn't work with utilities or system libraries that use specific features of a particular CPU, but neither do other 68k CPUs. EC and LC versions of 68k CPUs were also commonly used in machines 'back in the day' because they were cheaper, and still are today now that the original chips are getting rare and expensive.

Unlike all those old Motorola 68k chips, the Apollo 68080 has the advantage that 'missing' features can easily be added at no cost to the user. The only issue I can see is that the designer (Gunnar) has his own ideas on what should be included in it. Maybe if people like you had a more positive attitude he might be more accommodating.

[minator]

Originally Posted by eXeler0

Im guessing that is still the ”secret” goal of the Apollo-core team, to turn the 68080 FPGA into an asic.

Interesting idea, but likely to be hideously expensive.
There are relatively cheap ways of doing it but performance will be limited.

Would actually be interesting to see what sort of performance it would deliver at something like 1.5 GHz and also performance/W compared to lets say ARM chips typically found in Raspberry Pi etc.

If they went all out and spent $millions, they might get to RPi 3 level performance. AFAIK it's an in-order design, so I can't see it going anywhere near a Rpi4.

It's a really cool project, but it's difficult to see how they can make the economics work.

[Bruce Abbott]

Originally Posted by Mr-Z

It's definitely true that emulation is the cheapest&fastest way to run a high end classic Amiga.

This assumes you already have a PC (or Mac?) with enough power. If you only have an older PC then it might be slower than some alternatives, and upgrading it might cost more.

The real advantage of emulation is that you don't need any original Amiga hardware. However that does mean that you miss out on a desirable part of the retro experience.

But a PiStorm32 with Pi4b/CM4 running at 2.2 Ghz you are already getting WinUAE like speed.
It's safe to say that CPU performance is quite overkill already for most of the existing 68K software base.

This is true. Even the lower spec Pi's do a very good job.

WHDload compatibility is extremely good as well, would say not better or worse than on WinUAE.

And unlike other hardware solutions, PiStorm can be as compatible as you want simply by adjusting the emulation to suit.

Yeah it's an ARM processor underneath but who cares, you get a crazy fast A1200 and lots of goodies like DDR4 RAM, SD card up to 25MB/sec, RTG GFX, Wifi and soon FrameThrower for RGB pass through over HDMI.

And if you don't want all that stuff it can just act as a faster (or even not faster) CPU.

[aeberbach]

Originally Posted by Mr-Z

soon FrameThrower for RGB pass through over HDMI.

Once this appears the A1200 will be my ultimate Amiga. I can't think of much else I would want!

[Bruce Abbott]

Originally Posted by eXeler0

Im guessing that is still the ”secret” goal of the Apollo-core team, to turn the 68080 FPGA into an asic.
...
It would probably show up in accelerators for Amiga, Atari and even some Macs. It could even find usage in some consumer products etc, who knows.

68080 is just the Vampire's CPU. The other functions that make it into a full Amiga are what I think their real 'secret' goal is. IOW, not just a CPU but an SOC (System on Chip). This is like the Raspberry Pi, but oriented towards towards being an Amiga rather than a generic Linux box.

As for other applications, it would be super cool if after all these years the Amiga started to be used in consumer products, even if it didn't have quite the same performance as ARM etc. Not likely to happen though because you would need sales in the millions to justify it. Still you never know - would you buy a microwave oven or TV with 'Amiga Inside'? I would. Just need to get some Chinese manufacturer interested in it...

[grelbfarlk]

68080 is just one of the many names for the code that boots up on the FPGA on the Vampire.

One of the first plans was to buy an off the shelf FPGA board like a Terasic DE-10, but then they realized they'd be just selling adapter boards like for the RPi. Like the AOS4 PPC solutions, they realized if anyone could just assemble some off the shelf components and boot an OS onto it, it's like a $50 adapter.
Why not just make your own crazy boards at five times the cost?
This goes into PPC chips as well, oh they're expensive, slapping an SoC PPC board together with some RAM is going to cost a mint.
A $90 PPC CPU available for a decade is about 4X the speed of the 68K FPGA emulation on a Vampire.
Now, is the FPGA on the Vampire less than $25, YES! Sometimes!
Is the CPU on the RPi loading a different ROM and OS on boot to get to the environment you want to get to less than $25?
Yes!
How fast is it?
About 10X faster than the Vampire, just loading on a different ROM on boot on a cheaper piece of hardware.
Well it must be expensive?
No not really you can get an RPi and an adapter, on a bad day for 1/4 the cost of a Vampire.
Well what's the registration fees to use that stuff? Who do I have to pay for the privilege of booting this thing up?

[Promilus]

68080 is not "name for the code" - it is distinct functional block which can work basically independently on all other additions. This only implements CPU with memory interface. Anything else is optional, like actual 68k external bus interface or SAGA (and within SAGA things like Maggie3D are also distinct functional blocks resembling structure of actual chipset spread over few chips). V2 started with just 080, 68k external interface and iirc rtg. All other stuff was gradually added (yes, FPU support as well, same goes with SAGA which might've existed as blueprints before but only was introduced at some point with V4SA). It basically means while it does have both chipset and CPU inside one IC (FPGA itself) there are distinct entities which can work pretty much independently so you could've upload just SAGA to FPGA and connect it to regular 060 - it might work but rather slowly due to poor memory interface. 080 is separate entity, it should not be mixed with actual products so what V2 or V4 offers. V2/V4 is amalgamate of all the softcores produced by Apollo Team.

As for ASIC Vampire - even if it does materialize it would not be nearly as fast per clock as original design. Because now 080 with relatively low frequency operates with DRAM which is basically much faster. So there are less issues with cache misses and queue reload. With 1.5GHz ASIC (even if they do pass EMC and doesn't suffer from milions of issues along the way) the actual performance will be greatly affected by speed of memory interface and cache. And I am quite certain latency of memory will begin to influence performance greatly. Which is not an issue now in 100MHz range but we're talking about something 10-15x faster. That's why I think it was pretty darn funny when Gunnar was comparing memory performance of DDR3 based 080 and PowerPC working with regular SDRAM... yeah. But even then that PPC was tuned to work well with relatively slow main memory vs core speed. 080 is not. And while I think faster memory is possible to implement for ASIC vampire it might push actual dev team too far or inflate price even further. Because we're basically talking about effective memory speed of well over 3GHz. That's not for some DIY ...

[dreadnought]

With gaming being my strict priority, all these options leave me fairly uninterested. I totally get it how they might be exciting for people who are into Amiga's productivity applications and coding, and it's nice that there are new hardware developments, but in the end - for me - all this talk about moar mHz and super-features means nothing if there are no new games that can use it (I don't count ports here, while technically impressive they aren't exactly "new").

Of course writing new games utilizing such powerful machines is far from trivial and higly unlikely it will ever happen. But that's the reason why I'm perfectly happy with my OG HW A500's and MiSTer's slightly-faster A1200 core. Admittedly, the latter could be a bit more faster yet, so as to provide snappier WB experience, but I suppose that will eventually happen in the next ~5 years, once a more powerful (but still affordable) FPGA board appears.

[Thomas Richter]

Originally Posted by Bruce Abbott

You say it's a 'partial' emulation of a 68k CPU, but fail to mention that the 68020, 68030, 68040 and 68060 are only 'partial' implementations too.

Nope, the 68EC020 is a partial 68020, in the same vain the 68EC060 is a partial 68060.

Originally Posted by Bruce Abbott

Doesn't have a (compatible) MMU? Neither does the 68020 (a separate chip is needed for that, and very few Amiga accelerator cards have one).
Doesn't have a full FPU? Neither does any other 68k CPU.

The 68030, 68040 and 68060 have one.

Originally Posted by Bruce Abbott

The 68020 and 68030 have none, and the 68040 and 68060 only partially implement the 68882.

Which doesn't matter since software can fix that up. I wouldn't have a problem - and I even suggested - that the 68EC080 had a software-assisted MMU or FPU a which software performs partially the functions of those units, same as on the 68060 or 68040. Thus, for example, if the MMU table walk is too complicated for an FPGA, and this may very well be, the net result of software + softcore could perform everything a full 68060 could do. That is for all practical purposes what matters.



There is quite a difference if you start with something you know you can turn with a little help into a full software product, or you ignore compatibility and create something that breaks on purpose existing software.

Originally Posted by Bruce Abbott

In reality the 68080 core implements enough of every other 68k CPU used in the Amiga to run 99.9% of games and applications. It obviously doesn't work with utilities or system libraries that use specific features of a particular CPU, but neither do other 68k CPUs. EC and LC versions of 68k CPUs were also commonly used in machines 'back in the day' because they were cheaper, and still are today now that the original chips are getting rare and expensive.

..and your point is? Same as the 68EC020 couldn't run everything a 68020 could due to its limited address space, the 68EC080 cannot run all the applications of a 68060 due to lack of design considerations of its maker. It is an EC processor by all means. Restricted intentionally by design, not fixable by software.

Originally Posted by Bruce Abbott

Unlike all those old Motorola 68k chips, the Apollo 68080 has the advantage that 'missing' features can easily be added at no cost to the user.

So why aren't they? There some technical reasons, but the biggest issue is that Gunnar's ego is in the way. The technical reason is that an MMU would be very much a bottleneck between core and bus, and that it would limit the throughput and execution speed of the FPGA significantly. It sits in a critical path of the design that would downgrade the 68EC080 to a speed that is slower than that of the 68060. It would become a less a limitation if an ASIC would ever materialize, but it still means that one should possibly study a possible implementation in an FPGA upfront, before starting an FPGA. Concerning the FPU, I suspect the problem is likewise. You can get FPGA cores that offer units for double precision, my bet is that the selected core has one. Unfortunately, that makes it quite hard to implement extended precision on top, or you would need many additional cells - and have a slower implementation - if you need full 80 bits.

Originally Posted by Bruce Abbott

The only issue I can see is that the designer (Gunnar) has his own ideas on what should be included in it. Maybe if people like you had a more positive attitude he might be more accommodating.

Many people had a more positive attitude towards Gunnar if Gunnar would have a more positive attutide to them. Gunnar outright ignored the written down will of the P96 designers, to name the nail on his coffin for me. I had a more positive attitude if he wouldn't give me the feeling that software is worthless, and should be free and unpaid. Gunnar abuses people for the sake of his only interest, and that's not an attitude to work with in a market that small.

[Thomas Richter]

Originally Posted by Promilus

As for ASIC Vampire - even if it does materialize it would not be nearly as fast per clock as original design. Because now 080 with relatively low frequency operates with DRAM which is basically much faster. So there are less issues with cache misses and queue reload. With 1.5GHz ASIC (even if they do pass EMC and doesn't suffer from milions of issues along the way) the actual performance will be greatly affected by speed of memory interface and cache. And I am quite certain latency of memory will begin to influence performance greatly.

There are more issues coming up with faster cores. Currently, the 68EC080 has a write-through cache. That's doable because memory is relatively fast compared to the relatively slow core (by today's means), but this type of cache will become an issue with the core being faster, as you say. You then need a copyback cache, which means that the data in the memory can become inconsistent with the memory in the CPU. At this point, you have the issue that DMA might access stale data, because DMA cycles access memory, not the CPU cache, and at this point, you again need the MMU if you want to allow to support DMA devices on the Zorro bus. Oh well...

[jbenam]

Vampire’s “new chipsets” are mainly made to create vendor lock-in. Their entire business plan verged on incompatible instructions that would only work on Vampire with the logic that “exclusive software” making use of it would then assure the users’ need for a Vampire.

Too bad for them that alienating developers isn’t the best way of getting “exclusive software” done and then the final nail in the coffin (hehe) came when PiStorm became mainstream and utterly destroyed their chances of achieving their plans.

The entire market for Vampires evaporated in mere months. Why purchase an expensive Vampire when you can just purchase a PiStorm for 1/4 of the price?

You’re even using the *original* chipset of your Amiga for maximum compatibility instead of a rough approximation of the behaviours of the original chipset (that excluding the weird-ass extensions they wanted to push on the entire community).

So no “HURRR BUT PISTORM IS EMULATIOONNNNN”, since Vampire is “emulation” as well. They’re just happening differently (one is emulated in software and the other in hardware) - the only difference is that PiStorm’s emulation is for a matter of fact more accurate to the original hardware because it doesn’t try to reimplement everything. So you’re still using your Amiga chipset instead of turning it into an enormous I/O + PSU board.

If someone thinks that Apollo-core Team isn’t in this just for the money and/or a lot of ego on someone’s part they’ve been drinking too much of their Kool-Aid. The vendor lock-in part is very explicit about that. They never wanted to “save the Amiga”, because if they did, they wouldn’t do all that stupid crap the way they did in their implementation.

The ASIC will never materialise since PiStorm took their market and they would never recoup the costs of a production run. Heck, I have always thought that the entire “Vampire ASIC” thing was just baloney to get people to invest in their ecosystem in the hopes of an ASIC version ever happening, even if they knew very well (as Thomas just explained) that it wouldn’t be a very realistic possibility.

Very curious to see if the Apollo-core Team will stick around after their sales plummet to nothingness.

That said, back to the thread at hand - PPC has always been a dead-end (Amiga should’ve dropped it as soon as Apple switched to Intel) and won’t be making a comeback. The original 68k line is deader than dead, so the only way “forward” to keep producing hardware is emulated 68k on ARM.

I am pretty sure I advocated for an ARM version of AmigaOS back in 2009-2010 on this very same forum. Oh, the very different state the Amiga market would be in now if someone of the “decision makers” caught a glimpse of that Glad to see someone going in that direction (I am referring to the A600GS with their mixed-code approach, even if I don’t approve of everything AmigaKit has done with it) - one can only hope (and keep dreaming) that one day an ARM-native version of AmigaOS 3.3 will appear

[hammer]

Originally Posted by pix

Not wanting to start a flame war with my first post on this board, but I'm hoping to get some solid opinion on the best Amiga hardware direction to take.

For where I am coming from, after experimenting with various Amiga offerings over the years, emulation options are a given. Love digital preservation and it does the job + more, but its still emulation.

Pimiga is a great combo of emulator and hardware and on a Pi400 it almost captures the feel of having an Amiga very well, so I'm happy with that end of things.

If I was going to opt to use an OS on off the shelf hardware, while I liked the look of MorphOS and tried it on an iBook the idea of owning Apple hardware to run Amiga makes me feel a bit.... dirty.

I was enthused about AROS. If I must use off the shelf, then x86 is as good as any and AROS does feel very Amiga. Unfortunately it was going well but now seems (to me) to have stalled for some time.

Of course I realize you can take an original Amiga and transform it with accelerators and I'm glad some people do, but I don't think its for me.

Which means if I am picking something that runs on consumer grade hardware then I am really drawn to OS + alternative hardware. Coming across the 68080 FPGA rekindled my interest, but as I read about the performance I was a bit disappointed and it doesn't run OS4.1 correct? So I have to choose between OS4 and 3. I love the look of OS4 and I'm not against PPC hardware as a way forward, but its a shame the Amiga is so sharded by CPU and OS and so on.

I'm not interested in ARM processors at all really.

I also don't really understand why the PPC and 68080 are bounded by clock speed. Why can't the PPC push to even the triple core 3GHz of an Xbox 360? Or any is the 68080 clocked at 80-odd MHz? Why not 500? I don't get it.

Would really hope the Amiga will one day get standardized on OS 4 with either Moto or PPC redesigns at higher power and its own chipsets to really offer Amiga developers a way to flex the differences of an Amiga against anything else.

In a world where we have to choose between boring ol x86 and an endless stream of ARM, we need that!

There is a 64-bit PowerPC compatible CPU with >3Ghz i.e. IBM Power9 and it doesn't run AmigaOS 4.1 FE.

For example
https://www.raptorcs.com/content/TL1MB1/intro.html
Talos II Lite Mainboard for $2,062.24 USD.

https://www.raptorcs.com/content/CP9M31/intro.html
IBM POWER9 v2 CPU (4-Core, 16 threads, 3.2Ghz base, 3.8 Ghz turbo) for $944.87 USD.

I can obtain IBM Power9 CPU Processor Module 02AA966 for AUD $326.84 to AUD $586.80. There's $180 USD price Power9 CPU in the USA. A CPU by itself is useless.


QorIQ T2080 Reference Design Board (T2080RDB-PC) with PowerPC e6500 has $1,853.137 USD asking price and it doesn't run AmigaOS 4.1 FE. Don't expect Windows-level interoperability experience with various 3rd party PowerPC boards since the PowerPC camp has very weak standards. This is the problem with the RISC alternative. Many clowns in RISC land wants to boat anchor the customer with anti-interoperability customization.

[Locutus]

Originally Posted by eXeler0

It could even find usage in some consumer products etc, who knows.

Why would you pick an processor with no modern mainline compiler chain, no support for questionable vendor support, no future proofing guarantees, just no modern ecosystem and it's unable to run a modern a modern RTOS' or just a Linux kernel for a new product?


I can go through the catalogues of various ARM vendors and pick a processor that fits my exact product requirements, have proper BSP's, SDK's, and a whole ecosystem around them and not need to gamble my whole project away because when I was a teenager I liked handcoding m68k assembler.


And here's the thing, I can't imagine that Gunnar doesn't realize this as an ex-IBM'r. Which makes the whole ASIC-waffle even more of a disingenuous/scummy/cultish promise (pick your flavor).

[hammer]

Originally Posted by Thomas Richter

There are more issues coming up with faster cores. Currently, the 68EC080 has a write-through cache. That's doable because memory is relatively fast compared to the relatively slow core (by today's means), but this type of cache will become an issue with the core being faster, as you say. You then need a copyback cache, which means that the data in the memory can become inconsistent with the memory in the CPU. At this point, you have the issue that DMA might access stale data, because DMA cycles access memory, not the CPU cache, and at this point, you again need the MMU if you want to allow to support DMA devices on the Zorro bus. Oh well...

The alternative is a modified Amiga chipset that would report memory content change after DMA action to the CPU. Only the relevant cache line needs to be reloaded. CPU cache content may not need to be reloaded if the CPU's focus is somewhere else.

Cache coherence competent chipset matters since classic Pentium.

MAI Teron was labeled cache coherent incompetent by a Linux programmer.

There's a long way to go for AC68080 to reach P5 Pentium's cache coherence competency. A smart and modern Northbridge functions are needed for AC68080.

[Mr-Z]

Originally Posted by Bruce Abbott

This assumes you already have a PC (or Mac?) with enough power. If you only have an older PC then it might be slower than some alternatives, and upgrading it might cost more.

The real advantage of emulation is that you don't need any original Amiga hardware. However that does mean that you miss out on a desirable part of the retro experience.

Every mediocre modern PC should be able to run WinUAE really fast, even my old i7 870 CPU is able to almost reach the same speed as my PiStorm32+CM4 @2.2Ghz.

Emulation like WinUAE does leave out a desirable part of the retro experience indeed, using a physical Amiga is much more fun.
Seeing the Amiga making use of modern/current tech is a kind of magical experience

For me WinUAE is more like a tool to manage my Amiga's like giving me an easy way to setup a new HDD, testing software etc.