smc8516@isc.rit.edu (S.M. Curtin) (12/18/90)
I posted this last week and got no replies or acknowledgement whatsoever, so I'm assuming our server screwed up again. Below are some questions I'd really like to know the answers to. A friend and I are planning to build LUCAS boards for ourselves within the next few months. I seem to remember reading here that one could replace the 68020 in a LUCAS with a 68030 providing he rewires the socket (and changes it?) and ties off the extra 68030 lines. Is my memory correct on this matter. If so, could someone please repost or point me in the direction of the pertinent information. Also, would the MMU in the 68030 be operable in this type of setup. It's the main reason I'm interested in using a 68030. If it's not going to work, I might as well stick with an '020. Which leads me to another question: how difficult would it be to install a 68851 on the lucas board? If it's just a matter of a straight hookup like the 68881, I can deal with it, if not I'd probably need a vivid description of how it's done. (Do the 68851 docs give an example of this?) Which leads me to ANOTHER question: Can the 68851 be hooked up to a 68010 (my current processor) with ease, and if so would something like SetCpu be able to detect and make use of it, despite the fact that an '020 or '030 is not present in the system? Anyone who has read my posts in the past here knows it's sort of an ongoing quest of mine ever since I "upgraded" from a 1000 to a 500 to achieve loadable kickstart on my machine. SetCpu offers that option, but I need the MMU. If I don't do it this way, I'd have to Do SOme cumbersome ram hack that maps memory atop and switches out the $f80000 ROM. I'd much rather go the more elegant route. Any help would be appreciated. Thanks. ..Sean. Oh, by the way, I AM the one who asked about readdressing 501 ram at WOC in Toronto. I think those Q & A sessions are great. I hope you folks at Commodore keep them up.
daveh@cbmvax.commodore.com (Dave Haynie) (12/20/90)
In article <1990Dec17.204644.22516@isc.rit.edu> smc8516@ultb.rit.edu.isc.rit.edu (S.M. Curtin) writes: >A friend and I are planning to build LUCAS boards for ourselves within >the next few months. I seem to remember reading here that one could >replace the 68020 in a LUCAS with a 68030 providing he rewires the >socket (and changes it?) and ties off the extra 68030 lines. Is my >memory correct on this matter. If so, could someone please repost or >point me in the direction of the pertinent information. I don't know where they documented it, but in general, a 68030 with a few lines tied off is a drop in replacement for the 68020. The pinout is, of course, different and must be adjusted for, but the timing is virtually identical. Off the top of my head, you need to pull up the STERM*, CIIN*, CBREQ*, and MMUDIS* lines, all others should have 68020 equivalents or are unused outputs. >Also, would the MMU in the 68030 be operable in this type of setup. The MMU will work, the data cache won't. >Which leads me to another question: how difficult would it be to install a >68851 on the lucas board? If it's just a matter of a straight hookup like >the 68881, I can deal with it, if not I'd probably need a vivid description >of how it's done. (Do the 68851 docs give an example of this?) The 68851 is another story alltogether. It sits between the 68020 and the rest of the world (except the '881/2, which sits on the logical bus), and it's kind of annoying. It changes the 68020 timing ever so slightly, and for the worse, so while your 68020 logic may still work, timing critical stuff like memory has a chance of not working any more. The '851 was designed to be, uh, flexible, and so there are several ways to interface it with the 68020. Motorola wanted to provide for logical bus cache, logical bus mastering by something other than the 68020 without affecting the physical bus, things like that. So it gets a bit gnarly to do anything with it. If you get real serious about it, get the '851 book, and I can probably provide some pointers. But you're much better off with the '030 for most applications. >Which leads me to ANOTHER question: Can the 68851 be hooked up to a 68010 >(my current processor) with ease, and if so would something like SetCpu be >able to detect and make use of it, despite the fact that an '020 or '030 is >not present in the system? Not really. The '851 uses the coprocessor interface and speaks in '020 signal terms. Only, it does its own chip selecting, rather than being externally selected like the '881, so you'd be rather hard pressed to emulate coprocessor cycles. I gather the 68010 does have some kind of coprocessor space cycles, but I don't know if they're anything the '851 would handle (I kind of went straight from 68000 to 68020, so I don't know the 68010 that well, other than the fact it's pretty much a 68000 from the bus protocol point of view). Plus, the bus conversion logic, if even possible, would give you a headache. >...Sean. -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "I can't drive 55" -Sammy Hagar
rbattle@hydra.unm.edu (Ron M Battle STUDENT) (12/20/90)
Last time I wrote to Brad Fowles, he was designing a 68030 daughterboard, available by April 1990. Soooo.... Contact: Brad Fowles The Lucas Project RR #5 Caledon East Ontario, Canada LON1EO
huver@amgraf.UUCP (Huver) (12/22/90)
Putting a 68030 into an adaptor socket, onto LUCAS '020 socket IMMEDIATELY gives you a free MMU (albeit a scaled-down, read on). Other than this, there is nothing to be gained in any form (you must disable 030 data caching, unless of cource you'll add your own cache memory design). The 68851 MMU has built-in chip-select logic, so it is much more straight- forward a hookup than 68881/2. However, as Dave Haynie pointed out, with '851 in place, the system looks like: CPU <-+-> MMU <---> outside world | FPU Even with MMU disabled (become a pass-through), law of physics says you will see some timing difference; it is much worse if MMU is enabled, for address translation obviously takes some time. The MMU in 68030 is a subset of '851. So if you're into MMU hacking and value the capacity (AmigaDOS doesn't), you might want to fool with '851. Make an adaptor board that holds both CPU and MMU, then wire MMU outputs back to LUCAS CPU socket. But prepare to witness malfunctioning due to signal timing skew. If you use FRANCES, you have no need of an MMU. FRANCES provides Kickstart in its space (via copying from WCS) which is exactly what SetCPU does anyway. And with heavy real-time nature of Amiga tasks, I suspect you won't see a virtual memory paging AmigaDOS for quite some time to come. By then you'll probably have saved up for a new 68040 Amiga? >Which leads me to ANOTHER question: Can the 68851 be hooked up to a 68010 >(my current processor) with ease, and if so would something like SetCpu be >able to detect and make use of it, despite the fact that an '020 or '030 is >not present in the system? As said, '851 was designed for 68020. For 68010, Motorola 68451 could be used (don't know if they still make it). But it's a whole different ball game there, you'll have to commit your own heavy hacking to make it work; certainly not worth thinking about if you value your hair. -huver ...!uunet!amgraf!huver