Linked by Thom Holwerda on Wed 24th May 2017 23:03 UTC
Amiga & AROS

Ars reviews the Amiga X5000, and concludes:

The X5000 is different. It feels like an exotic car: expensive, beautifully engineered, and unique. If you bought one, you'd be one of a proud few, a collector and enthusiast. It practically begs for you to dig in and tinker with the internals - the system comes with an SDK, a C compiler, Python, and a huge amount of documentation for things like MUI, the innovative GUI library. On top of that, there is the mysterious XMOS chip, crying out for someone to create software that leverages its strengths. It feels like a developer’s machine.

Should you buy one? That depends very much on what your needs are. If you are simply after the best price-to-performance ratio for a desktop computer, this is not the machine for you. But if you are interested in something very different, something that is pleasant and fun to use, and yet can still be used for modern desktop workloads, then the X5000 is worth a look. I have had this review unit on my desktop for over a month now, and frankly I don’t want to give it back.

I reviewed the sam440ep with AmigaOS 4 way back in 2009, and came to a relatively similar conclusion - these machines are a ton of fun, but they're just prohibitively expensive, meaning only existing AmigaOS users will really get their hands on these. They really, really need a more accessible machine or board - a few hundred Euros, tops.

Thread beginning with comment 644786
To view parent comment, click here.
To read all comments associated with this story, please click here.
RE[2]: Not an Amiga
by Earl C Pottinger on Fri 26th May 2017 15:52 UTC in reply to "RE: Not an Amiga"
Earl C Pottinger
Member since:
2008-07-12

I have to agree with you. The 68000 series topped out at the 68060 at 75 MHz.

Now I found 68xxx code to be far more efficient than 86x code (often 3 to 4 times faster) but with the Intel/AMD chip now measuring in the GHz the old 68xxx chips are left in the dust.

And if they used the same methods that let the Intel/AMD chips work today they probably would not be any faster than them.

Reply Parent Score: 1

RE[3]: Not an Amiga
by JLF65 on Fri 26th May 2017 18:55 in reply to "RE[2]: Not an Amiga"
JLF65 Member since:
2005-07-06

It's not just the clock rate - they've poured billions into making the internals better at decoding instructions so that those old clunky x86 CISC operations now run at better cycles-per-instruction that most RISC operations. When you can retire from four to a dozen operations per cycle, what do you need RISC for? Crazy... and I still hate the x86 ISA. Too bad they can't apply all that know-how to the 680x0 ISA. People would probably go back to writing in assembly again. ;)

Reply Parent Score: 2

RE[4]: Not an Amiga
by Megol on Fri 26th May 2017 19:55 in reply to "RE[3]: Not an Amiga"
Megol Member since:
2011-04-11

It's not just the clock rate - they've poured billions into making the internals better at decoding instructions so that those old clunky x86 CISC operations now run at better cycles-per-instruction that most RISC operations. When you can retire from four to a dozen operations per cycle, what do you need RISC for? Crazy... and I still hate the x86 ISA. Too bad they can't apply all that know-how to the 680x0 ISA. People would probably go back to writing in assembly again. ;)


There are several computer/processor architects that have worked on x86 processor designs saying that the 68k is harder to make fast in hardware. None of those are x86 fanatics in any way, one of them essentially said that x86 wasn't that bad (that some claim). :-)


While I don't understand exactly why they think this it likely is dependent on where the complications are located in the architecture and whether adding extra hardware makes the complications manageable without slowing down the pipeline in general.

Examples of known (to me) bottlenecks:
. decoding of '020 extended address modes
.. worst case for one instruction is a MOVE with indirection of both source and destination operands
.. punting to microcode is likely to be too expensive

. the flags (condition codes) setting rules
.. requires split renaming of flags for best performance
.. potentially more complex than x86

. MOVE setting flags
.. harder to do move elimination early
.. can be partially handled with instruction fusing (combining a MOVE with the following dependent instruction)

These are from the top of my head - have (or had) a list with many more from when wanting to make a super-scalar 68k clone.

Reply Parent Score: 2

RE[3]: Not an Amiga
by Megol on Fri 26th May 2017 19:39 in reply to "RE[2]: Not an Amiga"
Megol Member since:
2011-04-11

I have to agree with you. The 68000 series topped out at the 68060 at 75 MHz.


Officially yes, unofficially some '060 can run over 100MHz overclocked.

Continuing, the Apollo core a.k.a the 68080 (not a Motorola chip) runs at over 100MHz as designed and have more performance per clock than the 68060.


Now I found 68xxx code to be far more efficient than 86x code (often 3 to 4 times faster) but with the Intel/AMD chip now measuring in the GHz the old 68xxx chips are left in the dust.


The performance difference isn't that large with the same effort spent on both 68k and x86 with a few exceptions.


And if they used the same methods that let the Intel/AMD chips work today they probably would not be any faster than them.


Probably not given the resources Intel spends on processor design and (semiconductor) process design. The AIM alliance (Apple, IBM and Motorola) failed to make the PPC competitive with x86 and the PPC were easier to run fast.

Reply Parent Score: 2