View Single Post
Old 21 February 2017, 19:14   #8
matthey
Banned
 
Join Date: Jan 2010
Location: Kansas
Posts: 1,284
Quote:
Originally Posted by Emmet View Post
I'm going to go out on a limb here and ask, is there anything that could be done to save the "Amiga" brand, or has what's happened over the year(s) simply damaged it beyond repair?
The Amiga reputation is extremely damaged but not beyond repair. Despite the inept companies and criminals holding the Amiga hostage, there is still an interested Amiga community with a possibility for crowd sourcing and individual investors to raise money. The big problem is the legal situation though. A patent lawyer would likely incur significant costs but is really necessary to figure out who to buy what from and then there is no guarantee they would sell for any reasonable amount. I'm guessing that a financial expert would value all of the (ancient technology) Amiga Intellectual Property (IP) at no more than low tens of thousands of dollars but this would probably only get you laughed at by any one of the companies and criminals.

It would be possible to start over without the Amiga IP (including the "Amiga" name) but customers would be lost. Even the transition to PPC likely lost more than 50% of Amiga potential customers and that was before it was obvious that PPC was on life support. Who wants to switch to PPC now as there is only one company (Freescale->NXP->Qualcomm who is invested heavily in ARMv8) left making (embedded) PPC processors based on '90s PPC designs without SIMD and sometimes without even an FPU as the rest of the world's processors get faster and feature upgrades. Add to that paying a 10 times premium (compare for example the $35 Raspberry Pi 3 to Tabor at $400+). Sadly, the fastest Amigas are emulated and even smart phones will likely be able to emulate the Amiga faster than Tabor within a few years. The Amiga could give up on hardware and do software (emulation) only but that would be the end of any Amiga relevance or innovation.

It would be possible to move the AmigaOS to ARMv8. The logic here is that ARMv8 supports big endian and is the most similar architecture to the PPC. It is also cheap with a large selection of new designs. Unfortunately, moving to a new architecture is likely to lose many more Amiga users and then there would be direct competition with cheap ARMv8 hardware like the Raspberry Pi 3. IMO, the only way to save the Amiga is with very affordable mass produced hardware taking advantage of economies of scale. The Amiga could not beat the Raspberry Pi 3 on price, there would be nothing to differentiate the Amiga hardware wise from the competition and it would be late to the show with less support.

Another hardware option is (vertical integration) to create an FPGA 68k+custom chips SoC which can be turned into an ASIC. Practically all of the current Amiga customer base could be retained, customization could aid in compatibility, standardization and features like SMP, the company destiny could be controlled by owning the sythesizable code and the Amiga could start to innovate in hardware again. The downsides are increased costs and risks where business partners could be helpful. Trevor of A-Eon gave a $10 million dollar figure for relaunching the Amiga recently. I believe a mass produced Amiga Pi like hardware board (enhanced 68k+Amiga custom chips+3D) would be much less risky than trying to launch a low end PPC based board using existing Freescale/NXP/Qualcomm processors. Low end PPC may have been what killed the PPC as processors like the PPC 603(e) could not outperform the 68060 without double the cache sizes, 2x the memory bandwidth and twice the clock speed. Launching Amiga ARMv8 hardware would be cheaper and lower risk to bring to market but that market is competitive and saturated while the loss of Amiga customers from the architecture switch could create a situation where the hardware would not sell.

I believe the 68k CPU has hidden potential locked away. It has one of the best code densities (and encoding room to improve) of any popular 32 bit CPU which allows for smaller ICache and less fetch bandwidth without affecting performance. ARMv7 Thumb 2 is close in code density but lost performance by dropping back to 8 GP registers (10%-15% more cache accesses are needed than the 68k with 16 GP registers). Intel made a (no longer produced) superscalar inline Atom CPU with reduced micro-oping which outperformed an ARMv7 OoO Cortex A9 in adjusted 32 bit performance and energy efficiency despite the Atom only using 8 GP registers also (https://research.cs.wisc.edu/vertica...-struggles.pdf). Compared to the Atom, the inline superscalar 68060 has 8 more GP registers for 10%-15% less ICache accesses, better code density and encoding room for improved code density and smaller SIMD instructions (fetch bottleneck of the Atom), easier decoding with less/no micro-oping, a shorter pipeline and easier superscalar instruction scheduling which all translate to better performance and reduced power requirements. The 68060 also exhibits the same strong single core inline superscalar performance of the Atom/x86 (a CISC trait). The Apollo Core is another inline superscalar 68k design with many similarities to the 68060 including the strong single core performance (especially good for games). The 68k can't compete with PPC because it was never allowed to by Motorola/Freescale after the AIM agreement. Hyperion and A-Eon do not want it either. Perhaps it is the Amiga's last hope but only Amiga makes it impossible.
matthey is offline  
 
Page generated in 0.05725 seconds with 9 queries