----- Original Message ----- From: "Phoebus Dokos" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Monday, July 14, 2003 2:04 AM Subject: Re: [ql-developers] K68 Core
> Or as Zeljko suggested a Crusoe (which is not crap) :-) > However because of their NDAs, binding agreements etc. it's out of the > question too. I don't like Transmeta. Very closed design. > Another VERY good idea is the PowerPC (especially the new offerings from > IBM) which may not be that low consumption as a say a MIPS but it's > definitely the best CPU around in terms of price/performance for your buck. > Not to mention that it is fully supported. 68K-to-PowerPC translators > already exist and it could be the stuff of dreams for all hobby-computing > afficionados (our QLers included). Unfortunately that too would require a > huge investment, the market just doesn't have... > As I see it, PowerPC is not so vastly different from MIPS. I have searched for a nice model, but there is not such a thing. Everything is either in BGA, obsolete or otherwise problematic. > Maybe (I say maybe) a good idea would be (given the QLs 20th anniversary) > to gather all die-hard Amiga, Atari, QL (and other currently regarded-as- > obscure) platform users to provide a viable alternative to the > mainstream... maybe a sourceforge project even? Not a bad idea. But IMHO problem with these things is: -closed design with finite resources. Whoever does the design, tries to bake a buck (or ten) from selling it -they try to be 100% compatible and end up being uninteresting I am contemplating doing good old QL with new chips as a dead simple few_chips_design, preferably with old 68000. But problems that arise from that keep me spinning in an endless loop. With 68000, it's hard to ensure bandwidth needed for generation of a decent picture (like 800x600 with 16bp or such). If design is beefed up for this, it stops being simple. If I consider alternatives, it causes chain of design changes which end with whole thing being awfuly degeneric and mutated. Only few_chip_design that would capture true QL spirit while still being simple that I can envision is to do whole thing with Analog Devices ADSP-2192 DSP, emulating 68000 with DSP/blitter/FPU extensions. Everything else could be packed in a simple CPLD, along with a couple of SDRAMs. Unfortunately, ADSP-2192 is not cheap, until recently it was unavailable and there is no free development software for it. Besides, it could probably emulate 68000 at some 30/50 MHz. With incredible math performance, but still quite dated integer performance. So, after all this it seems to me that using MIPS (or similar machine) with Linux and optinally emultaing native QL seems like a optimal solution... Regards, Branko
