@OlafS25

Quote:

OlafS25 wrote: @cdimauro emulation is good but people only take real hardware seriously. That is true for both users and developers. In this sense substantial upgraded hardware would be important even if emulated environments are still more powerful.

That's a personal, emotional, feeling.

Like I stated other times:

y = f(x)

and I care only of y.

@OlafS25

Quote:

OlafS25 wrote: @cdimauro upgraded Roms? With what? Compatibility is pretty good already

The ROMS provided with WinUAE aren't updated from long time. For example, I found some severe issues running it with my RTG configuration (which was based on my real Amiga 1200 system hard drive image).

@OlafS25

Quote:

OlafS25 wrote: @cdimauro "software has to be partially rewritten" who will do it?

The existing developers.

Remember that MorphOS was written from scratch. And its (native) applications too.

The community is alive, and the transition is possible. Of course, it'll require time.

@cdimauro

Quote:

A successor which cannot be called as "Amiga"?

Far more obvious than Zeta.

Quote:

And don't even share the same design and goals, since it's just a PC with a CPU replaced with a PowerPC one

Amiga design failed in test of time (it hurts to write this). Why repeat previous mistakes (when you can make new ones...)?

Quote:

But the story ends here.

I don´t share your opinion (at last!).

@KimmoK

Quote:

KimmoK wrote: @OlafS25 Windows was originally 16bit and single core. Then they made it 32bit and pre-emptive. Then they made it 64bit and multicore. Same is doable with any/every OS with enough work.

Absolutely NOT! Windows provided abstract/opaque APIs and didn't published its internal structures. The Amiga o.s. (and all its "successors") are the exact opposite.

So, you need a lot of work.
Quote:

People tend to end up using the best or latest variant, same for programmers.

Except in the post-Amiga land.
Quote:

Having super fast HW but very RETRO OS is not better than using UAE ... but at the same time ... It would be cool if someone would develop AAA+ GFX with some FPGA. (mainly AAA like features + chunky modes + simple 3D + DDR3/PCIe) (but it would be grazy and very likely unprofitable) It would be cool if someone would develop NG68kCPU with some FPGA. (mainly 1Ghz, 36bit address bus, SoC features) (but it would be grazy and very likely unprofitable) It would be cool if someone could update AmigaLikeOS for that AAA+NG68k combo. (but it would be grazy and very likely extremely unprofitable, and anyway breaks some compatibility as usual)

Please, let the AAA "features" away, except for things which were already provided by other computers (PCs and Macs): packed/chunky graphics, 3D, shaders, multichannel 16 & 24 bit 2D & 3D channels, and whatever I missed to list now.

@cdimauro

Quote:

I saw, but take a look at some other benchmarks, with more common applications.

Interesting, FFT benchmarks not mentioned (well, PA6T "cheats" there, as it has built in accelerators for FFT operations). Not that I dispute your general assessment of PA6T performance.

@KimmoK

Yes having new FPGA based 68k hardware would be cool, it is different so there might be even some people from outside interested, of course niche and of course not competing with modern hardware, just a additional small toy.

I have doubts that amiga developers return even if there would be a modern new Amiga OS. I contacted game developers that were active on amiga in the past. They are now mostly doing smartphone games (if still active at all) and were not interested anymore. Also they are obviously getting old. Young developers never used amiga in the past, API is very different to what most people use today and there are no modern development environments with class libraries so to me winning new developers is difficult expecially if you add that you cannot earn money with it.

@pavlor

Quote:

pavlor wrote: @cdimauro Quote: I saw, but take a look at some other benchmarks, with more common applications. Interesting, FFT benchmarks not mentioned (well, PA6T "cheats" there, as it has built in accelerators for FFT operations). Not that I dispute your general assessment of PA6T performance.

Accelerators are welcome. I think that it's right to use what a processor/SoC provides: it's there. Use it (if possible)!

But the problem with the PA6T is that the core, put aside such accelerators, lags behind even an old G4 (at a lower clock!).

@cdimauro

if you look at the MorphOS specific software uploads there is not a huge activity. The OS is of course in development but OS without software is more or less useless

the activity on all NG platforms is low compared to 68k and 90s and if the platforms split from NG to NGNG I do not see how this changes

@OlafS25

Quote:

OlafS25 wrote: @cdimauro if you look at the MorphOS specific software uploads there is not a huge activity. The OS is of course in development but OS without software is more or less useless the activity on all NG platforms is low compared to 68k and 90s and if the platforms split from NG to NGNG I do not see how this changes

But the 68K activity is very low on the hardware side, whereas x86, x64, and ARM are... well, you know.

Last but not least, software is MUCH easier, faster, cheaper, and flexible to develop than hardware solutions.

So, I think that we should be a bit more realistic: a new Amithlon is a much better investment. With multicore support as a premium feature.

@cdimauro

Quote:

But the problem with the PA6T is that the core, put aside such accelerators, lags behind even an old G4 (at a lower clock!).

Not in memory intensive tasks (bottlenecks of slow memory interface of non-SoC G4 CPUs are visible eg. in MPlayer and FFMPEG benchmarks).

As PA6T has 2 cores, integrated memory controller and rich expansion capabilities, its perfromance/watt is clearly greater than classic G4 or G5 system. MPC8641D may offer similar or greater performance, but is more limited in expansion.

@cdimauro

Quote:

So, I think that we should be a bit more realistic: a new Amithlon is a much better investment. With multicore support as a premium feature.

Exactly!

@cdimauro

but it would be still seen as emulation (I know that it is not). Might be emotional but it is about humans.

Amithlon would be nice but the same problem as always... who does it? There are 3rd party 68k libraries (f.e. specific for games) who were developed in years. Who would start that today without any realistic hope to earn money with it?

@cdimauro

Yes X64 or ARM are the new platforms. Aros tries to support them but there you have the problem to develop drivers for all the hardware. And Aros tries not to break too much compatiblity while adding features. What you and others wish would completely break with the past.

@pavlor

Quote:

pavlor wrote: @cdimauro Quote: But the problem with the PA6T is that the core, put aside such accelerators, lags behind even an old G4 (at a lower clock!). Not in memory intensive tasks (bottlenecks of slow memory interface of non-SoC G4 CPUs are visible eg. in MPlayer and FFMPEG benchmarks).

Only by a little factor, and with an higher clock frequency. But, yes, it performed better.
Quote:

As PA6T has 2 cores, integrated memory controller and rich expansion capabilities, its perfromance/watt is clearly greater than classic G4 or G5 system. MPC8641D may offer similar or greater performance, but is more limited in expansion.

I don't know how limited it's.

However the latest G5 had similar features like PA6T, but with better performances.

@OlafS25

Quote:

OlafS25 wrote: @cdimauro but it would be still seen as emulation (I know that it is not). Might be emotional but it is about humans.

It is Olaf, it is. Please, be more rational: you already have seen the value of emulation.

A suboptimal emulation, BTW, because 68K apps don't need cycle-exact emulation neither the full chipset emulated. And we are talking about a single-core (used)emulation.

So, if 68K hardware isn't fully exploited, the same is regarding emulation, which can do way better (and in less development time).
Quote:

Amithlon would be nice but the same problem as always... who does it? There are 3rd party 68k libraries (f.e. specific for games) who were developed in years. Who would start that today without any realistic hope to earn money with it?

Passionate people, like the ones which work on 68K hardware (and chipset) too. Someone is trying this way, if you've followed this site.

@OlafS25

Quote:

OlafS25 wrote: @cdimauro Yes X64 or ARM are the new platforms. Aros tries to support them but there you have the problem to develop drivers for all the hardware. And Aros tries not to break too much compatiblity while adding features. What you and others wish would completely break with the past.

What do you think that happens with a 68K hardware solution? NEW features require new drivers (code, in general).

@cdimauro

Quote:

However the latest G5 had similar features like PA6T, but with better performances.

You still need memory and PCIe controller(s).

@cdimauro

Quote:

cdimauro wrote: However the latest G5 had similar features like PA6T, but with better performances.

Latest G5?

So all computers should be designed/manufactured/tested and distributed with CPUs that will come off the assembly line tomorrow.

Nice planet you live on.

_________________
Tinkering with computers.

@OlafS25

Quote:

OlafS25 wrote: @cdimauro but it would be still seen as emulation (I know that it is not). Might be emotional but it is about humans. Amithlon would be nice but the same problem as always... who does it? There are 3rd party 68k libraries (f.e. specific for games) who were developed in years. Who would start that today without any realistic hope to earn money with it?

Perhaps this is considering it from the wrong direction, or rather the less relevant direction in terms of potential users.
There is, after all, an existing company with a relatively large existing customer base which currently uses emulation as its core. A 'more native' amithlon type affair might be an attractive option for AmigaForever/WinUAE users... If it offers a comparable/improved experience.

_________________
Troll - n., A disenfranchised former potential customer who remains interested enough to stay informed and express critical opinions.
opp., the vast majority who voted silently with their feet.

@Deniil715

Quote:

And the almighty copper!! OMG what a master piece!!! not... With modern gfx cards you can do a lot more cool stuff that then copper and blitter ever could dream of. I would be surprised if modern 3D HW could not emulate the copper's visual effects with easy, but also vertically or in any angle, something the copper was way too slow and simple for.

Deniil you are comparing apples with oranges here. The copper works completely different to a framebuiffer device. The copper is a product of its time. for palette based bitmaps, and was good for what it did. Sure, it only got as fast as two low res pixels per instruction, but I think the copper was held back by the chipset. I think it was a mistake to clone the registers into AGA so they needed bank switching and bytes cut into little nibbles. They should have cashed in and built upon an exsting foundation in the chipset: chip ram pointers! Had they done that they could have scrapped back switching and simply added a palette pointer with a long word per entry. If this was done it would have been no problem for the copper to display a 12- or 24-bit truecolour screen with a simple copperlist. But no, they didn't think of my idea back then and neither did I so I forgot to tell them!

Being the way it was the copper wasn't much good for rotating, except for hybrid prerendering and other trickery. I studied it with a view to rotate. But it could be used for some 3d tricks.

But the other thing is, modern hardware, AFAIK, prerenders things. The copper was a realtime live GPU. Of course, when you have a fast GPU, with a 32-bit framebuffer, you can pretty much render anything withour limtations. That's no big deal. But can a modern GPU be programmed to execute instructions for every pixel? Not that I know of. Sure, it doesn't need to, becauee it isn't based on a palette, since it has DCI; direct colour injection. If it can do all a copper can do, but way faster and more powerful, I'd expect it can excecute a stream of instructions with parallel multi-threading, per pixel, in realtime as it sends the RGB to the DVI port!

Last edited by Hypex on 14-Aug-2015 at 06:03 PM.

@cdimauro

Quote:

With Trance you don't have a sandbox, but an emulation layer which let you run 68K apps as "first class citizens" like PowerPC apps, with full integration on the system

Well things must have changed. Because years ago when MorphOS when solo on the Pegasos there were comparasons between the MOS and OS4 JIT. The MOS JIT, Trance was reported to be a sanbox emulation; where as the OS4 JIT was task based emulation.

Quote:

Of course, software has to be partially rewritten, and 68K software will run like with AROS & JanusUAE. IMO.

I hope not! I hope it will run better. I could never get 68K apps to work on AROS and I fiddled with it for ages. I also tried on a friends AROS install which was some years newer and I still couldn't get it to work.

@cdimauro

Quote:

So, I think that we should be a bit more realistic: a new Amithlon is a much better investment. With multicore support as a premium featur

What sort of Amithlon? I mean, doing the same 68K thing or emulating OS4? I don't see the point of doing 68K anymore. It's done to death. And really, you can open up AmiKit on your PC anytime now, the same as being able boot Linux on the desktop. Is it worth the hassle of needng to reobot your computer in order to run an emulator that supports a few native x86 apps?

Some people still like this idea. And want to dedicate a PC to boot into an AmigaOS emulator. But I think we're past that stage. And these days it's limiting.