63848e@eleazar.dartmouth.edu (Jeffrey Buchsbaum) (10/13/88)
I called Dove Computer today to ask about an '030 board ugrade for the II. They told me that I would still need to buy a 68851 chip(PMMU) with their 030. Is is possible that an 030 can still need a PMMU. I would like an 030, but only if it also allows me to avoid a PMMU purchase. Otherwise, their description of a 30 percent increase in performance for the 1800 dollar 33mhz seems strange from a marketing point of view. I thought the 030 had the PMMU in it. Their 030 board clips into the 020 sockets, to clarify any confusion, and comes with inits and stuff to let the caches work. This is from some documentation that they sent to me. One confused MacII hacker. The goofy, all pupose disclaimer (needed or not): I am in no way associated with any company discussed or mentioned. All opinions are my own, and fully admit to not knowing that much. ========================================= Please send email to Jeffrey Buchsbaum at jeffrey.buchsbaum@eleazar.dartmouth.edu or to Jeffrey Buchsbaum H.B.0559 Dartmouth College Hanover, N.H. 03755
michael@mcdchg.chi.il.us (Michael Bodine) (10/14/88)
Jeffrey Buchsbaum (63848e@eleazar.dartmouth.edu) writes: > I called Dove Computer today to ask about an '030 board ugrade for the II. They > ... > from a marketing point of view. I thought the 030 had the PMMU in it. > > One confused MacII hacker. Somebody's confused alright, but it wouldn't appear to be you! The 030 does have the mmu on-chip. The only way an 030 upgrade would still require an 851 for external memory management is if 1) their software disabled the on-chip mmu and 2) their (or somebody's) software required the full functionality of the 851 -- the 030 mmu is a subset of the 68851; not all instructions supported by an 020/851 pair are supported by the 030. Even so, there would have to be hardware modifications and code changes to support the fact that the off-chip mmu by default only responds to a particular co-processor id (i forget which, not relevant) so that any co-processor functions embedded in object code would operate with the on-chip mmu. To access an off-chip mmu, you'd have to recode all mmu instructions to reference a different co-processor id. Seems like an awful lot of hassle to go thru when not doing anything would most likely just make it work! Anybody in Austin listening? At Dove? -- [ Michael Bodine, michael@mcdchg.UUCP, (312) 576-7840 ] [ Opinions expressed are mine! All mine! ] [ Motorola couldn't have them even if they wanted them! ] [ No one else agrees with me; why should my employer? ]
brian@natinst.UUCP (Brian H. Powell) (10/14/88)
In article <10395@dartvax.Dartmouth.EDU>, 63848e@eleazar.dartmouth.edu (Jeffrey Buchsbaum) writes: > I called Dove Computer today to ask about an '030 board ugrade for the II. > They told me that I would still need to buy a 68851 chip(PMMU) with their > 030. Is is possible that an 030 can still need a PMMU? > Their 030 board clips into the 020 sockets, to clarify any confusion, > and comes with inits and stuff to let the caches work. I'm not much of a hardware authority, so use salt grains when reading this. It sounds from your description that their board has a socket for a 68851 on it. If you don't have a PMMU, you usually have a dummy chip sitting in the slot to pass through the proper logical and physical lines, etc. Unfortunately, the signals that the dummy chip needs to handle differ between a 68020 and a 68030. If you replace the dummy chip with a real 68851, the PMMU part of the 68030 works correctly, even though the 68851 is effectively ignored. Quoting from a 68030 manual (page 12-3): When used in a system originally designed for both an MC68020 and an MC68851, the MC68851 may be left in the system or removed (and replaced with a jumpered header). However, if left in the system, the MC68851 is not accessible to the programmer with the M68000 coprocessor interface. All MMU instructions access the MC68030's on-chip MMU. This is true even if the MC68030's MMUDIS signal is asserted. The benefit in removing the MC68851 is that the minimum asynchronous bus cycle time to the physical bus is reduced from four clock cycles to three. If the MC68851 is removed and replaced with a jumpered header, the following MC68851 signals may need special system-specific consideration: [list omitted]... During translation table searches the MC68851 asserts the CLI signal but not RMC, while the MC68030 asserts RMC but not CIOUT. [etc...] Brian H. Powell National Instruments Corp. brian@natinst.uucp 12109 Technology Blvd. cs.utexas.edu!natinst!brian Austin, Texas 78727-6204 AppleLink:D0351 (512) 250-9119 x832