kholland@hydra.unm.edu (Kiernan Holland) (03/05/91)
Hey does anybody know if the 68040 speed-up boards for the amiga 3000 ,16mhz version, work at top speed or at the speed of the system. What determines the speed of the bus???? How fast is the 68040 compared to the 3000. How much do the speed-up cards cost with and without the 68040 installed. Is the Math-coprocessor in the 68040 or external? (by the 2nd question, I meant 68040 vs. 68030 at same speed). What special features does the 68040 have. And does the 68030 code work on it. Later.
patrick_meloy@outbound.wimsey.bc.ca (Patrick Meloy) (03/06/91)
I'm not exactly a hardware guru (understatement :) but there are a couple of things I have picked up about the 3000. >Hey does anybody know if the 68040 speed-up boards for the amiga >3000 ,16mhz version, work at top speed or at the speed of the system. >What determines the speed of the bus???? How fast is the 68040 >compared to the 3000. How much do the speed-up cards cost with The 3000 bus goes as fast as the original CPU. The 16Mhz version's bus runs at 16mhz, the 25Mhz at 25Mhz. This is one of the 'unsung' nifties of the 3000 over most other 'speedyboxes' on the market. The Macs run on NuBus which is (as I am told) 10Mhz only. No matter how fast the CPU is. I don't know whether the 16mhz boxes are capable of 25Mhz on the bus or if they are restricted to 16Mhz only. Either way your bus is going faster (and in 32bits) which is far far better than the 8/16 bit 8-10Mhz buses of the competition. The 68030 25Mhz has a mip rating in around 7.5 (we all know how bogus MIPS rating are, but its all we have for comparison). The 68040 of the same speed is presently up around 20-27 MIPS (a 486 is typically 15-18 MIPS) and should be able to hit 40 MIPS in the final production (All chips out to date are really more of less BETA versions). Again, I'm not a hardware GURU and am only repeating what I have picked up from other people. If someone who REALLY knows would care to comment we'd all be better off :) --------------------------------------- | patrick_meloy@outbound.wimsey.bc.ca | | 'The Outbound' BBS Vancouver BC | ---------------------------------------
daveh@cbmvax.commodore.com (Dave Haynie) (03/13/91)
In article <patrick_meloy.1053@outbound.wimsey.bc.ca> patrick_meloy@outbound.wimsey.bc.ca (Patrick Meloy) writes: >I'm not exactly a hardware guru (understatement :) but there are a couple of >things I have picked up about the 3000. Well, I guess I am one. What's a hardware guru anyway, a bus error? >>Hey does anybody know if the 68040 speed-up boards for the amiga >>3000 ,16mhz version, work at top speed or at the speed of the system. >>What determines the speed of the bus???? You have to watch out what you mean when you say "bus". The A3000 has three distinct buses -- the Chip bus, the Local bus, and the Expansion bus (Zorro II and Zorro III expansion cards). The Local bus is pretty much just the 68030's native bus. Anyway, a coprocessor card has the option of using the local bus clock (25MHz or 16MHz) or supplying its own. If it uses the Local bus clock, it must synchronize any bus accesses to that clock. If it supplies its own, it must of course synchronize to that clock. While the Local bus can be driven at either 25MHz or 16MHz, regardless of how the machine was shipped, an on-board 68030/68881 that comes at 16MHz can only be driven at 16MHz. So any coprocessor card that wants to run as an actual coprocessor, rather than taking over completely from the on-board 68030, has to run at the original clock speed, at least effectively, for Local bus accesses. The Chip bus always runs a two clock 7.16 MHz cycle. Only one of these cycles is available to the 68030, though on the A3000, that's a 32 bit cycle. The Zorro II bus always runs a four clock 7.16 MHz cycle. Zorro III is a fully asynchronous bus -- it does not depend on any particular clock. The bus controller uses the Local bus clock to create the various bus strobes when the Local bus master is acting as Zorro III bus master. Using the Local bus clock, rather than some unrelated clock, makes the bus couple better to the Local bus master. >The Macs run on NuBus which is (as I am told) 10Mhz only. No matter how fast >the CPU is. Macs all have the equivalent of our Local bus too. The Mac II and IIx run a 15.8-something MHz local bus, the IIci runs a 25MHz local bus, and the IIfx runs a 20MHz local bus and 40MHz cache bus. NuBus is based on a two clock 10MHz bus cycle. In most of the Mac IIs, that 10MHz clock is totally unrelated to the processor clock, though I suspect in the IIfx it is derived from the processor clock, since they're even multiples. NuBus is strictly defined by this clock speed -- it can't change any more than Zorro II's bus clock can change. That sets the upper limit on NuBus performance. Another reason you find NuBus performance slow is that the processor must sync up/down to/from the NuBus when acting as bus master, since they're generally running from unrelated clocks. The final problem is that, while NuBus supports a form of burst mode, it is incompatible with 68030 type burst mode. Zorro III answers each of these complaints, resulting in a much faster CPU to expansion bus conversion. >| patrick_meloy@outbound.wimsey.bc.ca | -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "What works for me might work for you" -Jimmy Buffett