geoffk@belfry.sandiego.NCR.COM (Geoffrey Kim) (05/21/87)
Having been a previous PeeCee owner (pitiful, aren't I?) I was just curious how the Amiga rates in terms of performance? Concern here is that Lattice compile/execution times will be N times slower/faster than running it on a PC? What would be the PC equivalent of the Amiga's 68000 @8Mhz? Would it be a (an?) 8088 @6/8Mhz or a 80286 @6/8/10/12Mhz or ? Anyone? Thanks in advance. Geoff K.
dillon@CORY.BERKELEY.EDU (Matt Dillon) (05/21/87)
>N times slower/faster than running it on a PC? What would >be the PC equivalent of the Amiga's 68000 @8Mhz? Would it be >a (an?) 8088 @6/8Mhz or a 80286 @6/8/10/12Mhz or ? An 8088 at around 10-16 Mhz, I would guess. That is, if you can stand the @#$@ segmentation model and lack of registers (and this applies to 68020/80286 comparisons as well). Although many would disagree, my personal opinion is that machine architecture is extremely important and high level languages only partially hide it. Frankly, the only reason Intel is still living is due to Big Blue's blind support of it. -Matt
daveh@cbmvax.cbm.UUCP (Dave Haynie) (05/21/87)
in article <127@belfry.sandiego.NCR.COM>, geoffk@belfry.sandiego.NCR.COM (Geoffrey Kim) says: > Keywords: Amiga, IBM, comparison > > Having been a previous PeeCee owner (pitiful, aren't I?) I > was just curious how the Amiga rates in terms of performance? > Concern here is that Lattice compile/execution times will be > N times slower/faster than running it on a PC? What would > be the PC equivalent of the Amiga's 68000 @8Mhz? Would it be > a (an?) 8088 @6/8Mhz or a 80286 @6/8/10/12Mhz or ? > > Anyone? How 'bout me? From a straight hardware point of view, a 68000 running at 8MHz is probably closest to an 80286 running at 8MHz (without too much knowledge of the '286 specifics on my part). Both processors have 16 bit address busses, both take 4 cycles to access memory, etc. The Amiga's processor is actually running at 7.16MHz, so it will be a bit slower than the 8MHz processor. Looking closer at the architecture, you find that the 68000 is internally a 32 bit machine in many respects, having 32 bit registers and many 32 bit instructions. It has a logical address space of 32 bits, with a linear physical address space of 24 bits. As I recall, the chip has 3 16 bit ALUs, a pair will do the 32 bit stuff. The 80286 is internally a 16 bit machine, with an extended segmented architecture and emulation of the original 8088 segmentation. It also has a 24 bit address space, though of course not a linear space. I believe the 68000 has some more powerful instructions, but the '286 benefits from smaller instruction lengths and the efficiency that its 64K segments give you on smaller programs (though, via PC relative coding, the 68000 can also act more efficiently on small programs). Anyway, what this means is that 68000 programs are usually longer for the same operation than 80286 programs. If you're dealing with operations of less than 64K in length, the 80286 stuff may run faster than an equivalent 68000. The 80286, however, takes a small performance hit when running programs larger than 64K, and a big performance hit when manipulating data objects greater than 64K in size. The 68000 has no problem with these. Also, most current 80286 systems are running in 8088 emulation mode, which is another big performance hit. Because of this, the Amiga will sometimes run up to 3x the speed of an AT on CPU bound tasks. Finally, the Amiga has a few devices designed to offload work from the 68000. The Bimmer (Blitter) is a device optimized for graphic maniplulation that moves images (with barrel shifting), draws lines, and fills areas faster than the 68000 could. The Copper (co-processor) is a tiny processor that accepts 3 types of commands used for altering graphic configurations. A few other items, too, make the Amiga's overall performance in its graphic environment usually exceed that of similar machines. > Thanks in advance. Geoff K. > > -- Dave Haynie Commodore-Amiga Usenet: {ihnp4|caip|rutgers}!cbmvax!daveh "The A2000 Guy" BIX : hazy "These are the days of miracle and wonder" -P. Simon
richard@pnet02.CTS.COM (Richard Sexton) (05/23/87)
Is this why Lattice 'C' manual for the PC says something
to the effect that "... because of the complex nature
of the 8086, no registers are left over to use as
register variables..."
Intel and IBM deserve each other.
Whatever happened to IBM's scientific division 68K computer.
One popular conspiricy theory down here is that IBM evaluated
Z8000, 68000 and 8086 and picked the lowest performence
machine, to prevent the embaressing event where one
of their 68020 machines would run rings around their
minis, and by using an 8088 would ... etc
UUCP: {akgua!crash, hplabs!hp-sdd!crash}!gryphon!pnet02!richard
INET: richard@pnet02.CTS.COMwtm@neoucom.UUCP (05/26/87)
Last year I taught an introductory computer class for Biology grad students. The main point of the class was to teach the studetns to use personal computers as convience tools; to that end we learned how to use BASIC (I know, I know...) to quickly knock off simple problems. One of the Problems I gave them was to program the Shell Sort algorithm. One of the students ran the same BASIC code on both the 4.77 MHz XT and the Amiga in Microsoft BASIC for each respective machine. Sorting the same 1000 single-precision floating point numbers took about 90 sec. on the Amiga and about 120 sec. on the XT. Both series were indentical and in the same order on input to both machines. The Amiga was running BASIC alone from Workbench. Given that the BASIC interpreter for the Amiga is more complex than the XT, this seems like a pretty credible real-worldish performace versus the XT for the Amiga. Of course, we're comparing apples and oranges here, and there is no simple way for an XT to do the neat graphics things that the Amiga is capable of. And IBM still doesn't have a blit chip as standard equipment in the new models. Re: the mention of the IBM 9000 "Scientific" computer. Gaaak! The operating system was a mix of CP/M and Intel's ISIS-- or so it looked. Handling of real time interrupts was terrible, as they expected you to do so in Microsoft BASIC (which had unbelievably numerous bugs). You could get the assembler, if you begged on your knees long enough at a Blue Alter (I have too much pride, so we never got it). We were trying to run an HPLC with the darn thing using IBM's own LC software. Never did get it to work because the computer would never stay running for more than a few days at a time. IBM has since sold their lab instrument division to Nicolet who is well on the way to killing off the remaining products, or so it seems. The integal Data Products manufactured printer mechanism also smoked with predictable regularity. This is about the worst personal computer that I've had the occasion to work with. --Bill (wtm@neoucom.UUCP ...!cbatt!neoucom!wtm)
chips@usfvax2.UUCP (05/26/87)
In article <1913@cbmvax.cbmvax.cbm.UUCP>, daveh@cbmvax.cbm.UUCP (Dave Haynie) writes: > in article <127@belfry.sandiego.NCR.COM>, geoffk@belfry.sandiego.NCR.COM (Geoffrey Kim) says: > > Having been a previous PeeCee owner (pitiful, aren't I?) I > > was just curious how the Amiga rates in terms of performance? > The 80286, however, takes a small performance hit when running > programs larger than 64K, and a big performance hit when manipulating > data objects greater than 64K in size. All too true. > Also, most current 80286 systems are running in 8088 emulation mode, which > is another big performance hit. Because of this, the Amiga will sometimes > run up to 3x the speed of an AT on CPU bound tasks. > -- > Dave Haynie Commodore-Amiga Usenet: {ihnp4|caip|rutgers}!cbmvax!daveh > "The A2000 Guy" BIX : hazy The reports of reduced performance are greatly exaggerated... :-) The '286 runs faster in real (8086) mode than in protected ('286) mode. In protected mode, the '286 does several memory reads on every segment register load. Real (8086) mode does not suffer this penalty, but then, neither does it provide memory management. The real (:-) problem with real mode is the real-mode OS -- MS-Doesn't. I believe that too many people would throw out the baby ('286) with the bathwater (MS-DOS). Try Xenix on a fast AT and then judge. -- Chip Salzenberg Address: "{gatech,cbatt,akgua}!usfvax2!ateng!chip" AT Engineering, Tampa, FL Redress: "chips@usfvax2.UUCP" "Use the Source, Luke!" My opinions do not necessarily agree with anything.