amc4919@cec1.wustl.edu (Adam Michael Costello) (05/03/90)
If you use that socket in the 3000 to upgrade to a 68040, is it still bound to run at 25MHz (or 16MHz, as the case may be)? In any case, if you upgrade to the 68040, does it matter whether the 3000 was 16 or 25? Can the upgrade be done now, or are there things we have to wait for? The 68040 is a 64-bit processor (internally, 32 bit data bus though), right? What else does the '40 have over the '30? Again, does anyone know the limits of the capabilities of the 1950? Thanks much! AMC
murphyd@csuchico.edu (Dave Murphy) (05/03/90)
In article <1990May3.043218.15590@cec1.wustl.edu> amc4919@cec2.UUCP (Adam Michael Costello) writes: >If you use that socket in the 3000 to upgrade to a 68040, is it still bound to >run at 25MHz (or 16MHz, as the case may be)? In any case, if you upgrade to >the 68040, does it matter whether the 3000 was 16 or 25? Can the upgrade be >done now, or are there things we have to wait for? The 68040 is a 64-bit >processor (internally, 32 bit data bus though), right? What else does the >'40 have over the '30? The 68040 is still a 32 bit chip. It out performs the 030 because of it use of the pipeline. By using a pipeline the 040 is able to have multiple instructions being processed at the same time. Each instruction is staged in a different stage of the FETCH, EXECUTE cycle. This allows many a different instruction to complete every clock cycle and there appear to be executing a instruction every clock cycle. This same idea is used by RISC processors and the 80486. The 040 also has a built in FPU. I have heard that the 040 operating at 25 MHz will outperform a 40 MHz 030. ____________________________________________________________ /// __ __ __ __ | \\\ /// /\ __|| || || | | David L. Murphy \\\ \\\ /// /--\miga |__ |__||__||__| | murphyd@csuchico.edu \\\ /// \\\// | murphyd@cscihp.UUCP \\///
daveh@cbmvax.commodore.com (Dave Haynie) (05/04/90)
In article <1990May3.043218.15590@cec1.wustl.edu> amc4919@cec2.UUCP (Adam Michael Costello) writes: >If you use that socket in the 3000 to upgrade to a 68040, is it still bound to >run at 25MHz (or 16MHz, as the case may be)? No. The A3000 motherboard can only be run at either 16MHz or 25MHz, and a Coprocessor slot card can use the motherboard clocks or, alternately, supply its own. If you want an add-in that goes faster than 25MHz, no problem, only it'll have have keep the faster cycles on the Coprocessor slot only. Much like the way the A2630 does things in a 2000. That's really not a big deal; there's not much around these days, other than cache memory, that can even keep up with a 25MHz 68030, much less a faster '030 or an '040. >In any case, if you upgrade to the 68040, does it matter whether the 3000 was >16 or 25? Depends on the design of the 68040 Coprocessor card, but it doesn't have to matter -- such a card could run the A3000 main bus at 25MHz, regardless of the on-board CPU speed. >Can the upgrade be done now, or are there things we have to wait for? I guess you have to wait for a 68040 Coprocessor card. No one's announced one yet. >The 68040 is a 64-bit processor (internally, 32 bit data bus though), right? Yes and no. It's a 32 bit Harvard architecture machine, like the 68030 only better. At any given time you may have two simultaneous 32 bit transfers going on the separate I and D buses. Since 32 + 32 = 64, some call that a 64 bit architecture. I wouldn't call something 64 bit unless it had 64 bit registers and 64 bit operations on such registers. Which I guess does in part exist on the 68040, since math operations and registers are actually 80 bits wide, internally. >What else does the '40 have over the '30? Real big, fast physical caches, very clever pipelining, some hardwired instructions, on-chip math. That's the basic feature list. >AMC -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "I have been given the freedom to do as I see fit" -REM
rehrauer@apollo.HP.COM (Steve Rehrauer) (05/04/90)
In article <1990May03.062713.25299@csuchico.edu> murphyd@cscihp.UUCP (Dave Murphy) writes: >The 68040 is still a 32 bit chip. It out performs the 030 because of it >use of the pipeline. The '030 is also pipelined, but not as well. :-) The '040 also offers much deeper on-chip caches (4k each for instruction and data, versus 256 bytes on the '030). >RISC processors and the 80486. The 040 also has a built in FPU. ^^^^^^^^^^^^^^^^^^^^^^^^^^^ Here's the big win, if you're doing floating-point. Inter-chip handshaking delays between (say) an '030 and '882 can be pretty horrendous. Compared to an '030+'882, the '040 screams; an order of magnitude improvement in execution times for floating-point intensive applications on an '040 isn't totally whacko to hope for. Your mileage will almost certainly vary. >I have heard that the 040 operating at 25 MHz will outperform a 40 MHz 030. It may be a bit premature to talk about '040 performance anyway, since no one is now shipping a Real System(tm) with them, and (which is a good thing, since) the preproduction chips aren't entirely bugless yet. -- >>"Aaiiyeeee! Death from above!"<< | (Steve) rehrauer@apollo.hp.com "Spontaneous human combustion - what luck!"| Apollo Computer (Hewlett-Packard)
bdb@becker.UUCP (Bruce Becker) (05/05/90)
In article <1990May3.043218.15590@cec1.wustl.edu> amc4919@cec2.UUCP (Adam Michael Costello) writes: |If you use that socket in the 3000 to upgrade to a 68040, is it still bound to |run at 25MHz (or 16MHz, as the case may be)? In any case, if you upgrade to |the 68040, does it matter whether the 3000 was 16 or 25? Can the upgrade be |done now, or are there things we have to wait for? The 68040 is a 64-bit |processor (internally, 32 bit data bus though), right? What else does the |'40 have over the '30? The 68040 will run at faster clock speeds, plus it executes machine instructions in fewer clock cycles. Some of the real bottlenecks in the Amiga right now are the lack of 1). bus access faster than 7 MHz 2). 32-bit bus interface by custom chips 3). support for 8 bitplanes on display devices 4). support for higher display resolution (1280 by 960 by 8 would be real nice 8^) When these get sorted out the need for a 68040 (or two) will make more sense... Cheers, -- ,u, Bruce Becker Toronto, Ontario a /i/ Internet: bdb@becker.UUCP, bruce@gpu.utcs.toronto.edu `\o\-e UUCP: ...!uunet!mnetor!becker!bdb _< /_ "I still have my phil-os-o-phy" - Meredith Monk
trebbien@cbmger.UUCP (Uwe Trebbien GERMANY) (05/07/90)
In article <1990May03.062713.25299@csuchico.edu> murphyd@cscihp.UUCP (Dave Murphy) writes: > >I have heard that the 040 operating at 25 MHz will outperform a 40 MHz 030. > It will outperform a 030 at 40 MHz, because a 040 running with external clock at 25 MHz is internally running at 50 MHz. Also the 040-FPU is much faster than a 882. I think I have heard that the 040 completes an FMUL-instruction in 6 clock cycles. UWE
eric@oakhill.UUCP (Eric Quintana) (05/09/90)
In article <149@cbmger.UUCP> trebbien@cbmger.UUCP (Uwe Trebbien GERMANY) writes: > >[The 68040] will outperform a 030 at 40 MHz, because a 040 running with external >clock at 25 MHz is internally running at 50 MHz. Also the 040-FPU is much faster >than a 882. I think I have heard that the 040 completes an FMUL-instruction in >6 clock cycles. Actually, piped FMULs produce results every 5 clocks. But a single FMUL execution time (including latency) is 9 clocks in the FPU. > UWE Eric Quintana ...!cs.utexas.edu!oakhill!eric
martint@altitude.CAM.ORG (Martin Taillefer) (05/09/90)
You guys are giving pretty precise info on the 040. Did Motorola publish the 040 manual? I can't find it in the local computer book store. -- ---------------------------------------- Martin Taillefer INTERNET: martin@pnt.CAM.ORG UUCP: uunet!philmtl!altitude!martint TEL: 514/640-5734 BIX: vertex