Comparison with the VFP/NEON featues would be nice.

I am mostly interested it the performance/mhz and estimated power consumption.

In other words, is this product anywhere near to an alternative to the ARM cores?

Even if it has half the performance that would still prove to be very interesting indeed.

Lets assume the new 68K Cores will outperform the ARM clock by block - what does this change?

It makes it a great candidate processor for usage scenarios outside of the Natami product. I am thinking of other uses to leverage the huge time investment in this project.

68050's got good performance/ease of use and that's the main advantage.

I hope the 68050 will keep true to the ISA while making a leap on the performance scale.

I'm a 68k nostalgic but I beleive this is almost completely irrelevant with modern compilers & in the modern world.

Also the load-store machines have been designed in part to be easily programmed *by* compilers :)

writing code in C and improving the compilers has been the way to go for a long time.

I plan to write in C and C++.  But I also plan on having a new compiler based on LLVM someday.  I don't know how long that will take.  It will have to be based on an x86-style backend modified for 68k use.

Most ARM and Atom processors are still single core. If more performance is needed, the x86 multi-core processors offer a better price to performance but battery life will suffer a little. I think the superscaler N68070 could do well in the single core arena. There are plenty of other ways to boost performance without going multi-core. I think multi-core is a bit overrated too. In theory, 2 cores can be up to twice as fast but that's if the software is perfectly written and the hardware is not sharing the same resources. It makes sense when more power is needed and it's not practical to increase the clock or add bigger caches anymore. The trend for mobile processors is to integrate more multimedia features like SIMD and GPU into the processor as this uses less electricity. That's kind of along the same lines of what Natami is trying to achieve.
 

I agree. I think the 68k is simpler than x86 though. I would expect the 68k to fall 1/2 way between modern ARM and a single core x86. That's not a bad trade off for ease of use and smaller code.

68020+ was a pain in the neck due to the very complex addressing modes (and some instructions), which brought instructions up to 22 bytes in length.

At least the command part of the instruction fit in the first 4 bytes. Everything else is data extensions and we know the offset of where to find the data from the command.

That was the major issue for the 68020+ decoder.

How fast do you think UAE runs Amiga software. I acts like it can run it prettty fast and I hear that it does. Is there something about silicon that makes decoding 68k slower as compared to an emulator. Do you think a processor with a modern design could emulate 68K as to similar performance of a Phenom cpu?
I hope I am not talking in circles.
Thanks.
 

Now where did I put thost optical processors? :)

Remember that now x86 means AMD64, which have a simpler ISA (many legacy instructions were removed, and segmentation also; there's a flat pagination model), and have 16 general purpose and (almost) ortogonal registers (you can use them as data or address registers).

So I think that a modern 68K cannot have similar performances to a Phenom, but can have good ones.

This is true.
This gave us headaches for quite a while.

But we solved this with the 68050 design.