Re: Apple II benchmarks

[Roy and/or Janet Miller <millers@inetnebr.com>]

Supertimer wrote:

> pausch@saafNOSPAM.se (Paul Schlyter) wrote:
>
> >Roy and/or Janet Miller  <millers@inetnebr.com> wrote:
> >
> >> Paul Schlyter wrote:
> >>
> >>> That's because the 8088 didn't have hardware multiply/divide.  It was
> >>> implemented in microcode, and one mul or div took some 100-150 cycles
> >>> to execute on the 8088.  On the 80188 this had been cut down to about
> >>> 30 cycles -- but you didn't compare the 6502 to the 80188, did you?
> >>
> >> As stated above Paul, Byte magazine did the comparison in 1985, and they
> >> were dealing with a product review for an 8086-2 based computer. That's
> >> why there was no comparison to an 80186
>

It's too bad that Paul hadn't noticed that *I* hadn't done the tests. Besides
that, the tests were between real machines that one could buy, not fantasy
machines. To my knowledge, the only computer sold in the US in any real numbers
that used the 80186 was the Tandy 2000, and it wasn't all that compatible with
the IBM - different disk drives, different video display, a not too compatible
BIOS, and I believe the slots weren't all that compatible either.

>
> >Which means they were comparing to an 80x86 with no hardware multiply....
>

Does the 6502 have hardware multiply? Anyone know? If not, then it certainly
was a fair test wasn't it? And I don't think it would have been a fair test
using a math co, when very few PCs or XTs ever had them. Usually only those
doing full time heavy duty spreadsheet work would have them, and the few who
used them for engineering or scientific work, and if that was all the market
that IBM had, the Apple II would still be the dominate machine today.

>
> Interestingly, I had a PC Transporter card in my IIGS and I
> found it slower doing almost all tasks than the IIGS (with
> accelerator turned off).  The PC Transporter is a coprocessor
> card that uses an NEC V30 processor, a clone of the 80186.

Actually the V20/V30 are optimized versions of the 8088/8086 with the addition
of 8080 microcode.

>
> It had 640k of memory and a socket for an 8087 math
> coprocessor.  I never populated this socket, but the PCT
> was horribly slow even though its CPU clock speed was
> turbo speed and the V30 was supposed to be faster than
> the real 80186.
>

The V20/V30 was between 5% and 40% faster than an 8088/8086, depending on what
kind of code was being run. It was at the high end when doing floating point
multiply and divide operations. Which is really interesting in your comparison
of the PCTransporter and the IIgs. The turbo version of the V30 should be much
faster than the PC times given above; how does it compare to the IIgs at math?
And since the V30/8086 were full 16 bit chips (i.e. 16 bit data bus, instead of
the the 8 bit data bus that the V20/8088 had) they should not only be much
faster than the V20/8088, but even the 65816, unless, of course, the 65816 is a
more efficient chip design.

>
> Now in 8-bit emulation mode, the IIGS should be equal to
> an 8-bit "Apple II" clock for clock right (since in 8-bit mode
> the 16-bit functions of the CPU are turned off)?  But the
> IIGS' clock speed is faster just like the PC Transporter's
> clock speed is faster than the real 80186.
>

2.5 times faster compared to about .5 times faster, plus the wider data bus of
course.

>
> To make the story short, I used an 8-bit unzipper (Angel)
> and compared the IIGS to the PC Transporter in unzip
> file speed.  Also, I compared the speed of converting a
> JPEG file in both.  The IIGS won both tests AND the
> PC Transporter was painfully slow running an early
> version of the GeoWorks GUI while the IIGS ran GS/OS
> much faster.
>
> My conclusion:  sorry, the PC didn't get acceptable until
> the '486 and didn't REALLY take off until the Pentium
> (and competitors like the AMD K6 and Cyrix MII series)
> came on the scene.  It wasn't until I saw the 200Mhz
> models of these sixth generation processors come on
> the scene did I really get excited about the PC.
>
> >>> The 4.77 MHz PC did indeed not perform particularly well: its clock
> >>> speed was slow,
> >>
> >> but faster than most Z-80 computers (most as in the millions of
> >> Kaypros, TRS-80 Models, 1, 3, 4, Osborne, and hordes of other CP/M
> >> machines)
> >
> >Slightly faster, yes, since the standard clock speed of CP/M computers
> >at those days were 4 MHz.  Which of course means those 4 MHz Z80
> >machines didn't perform very well either, particularly compared
> >to today's standards... :-)
>
> I am convinced now that we see history through our
> own experiences.  After all, why is every time
> someone mentions "Apple II"  you think of an
> Apple II+ in CP/M mode while to me I think of a IIGS?

Was this directed to me? Or to Paul? When someone says Apple II, I think of
6502/65C02 based personal computers.

>
> And why is it that every time someone mentions "PC"
> I think of a PC Transporter card in a IIGS in Turbo XT
> emulation (80186 compatible) and you think of a PC
> with a math coprocessor chip?
>

When someone says PC, I think of my Portable PC, or the XT that my wife used at
work when I was in grad school. Neither one had a math co, and both were mostly
used for word processing.

>
> By the way, while the PC Transporter was slow, I did
> compare it to a stock IBM XT and yes, it was faster
> than a stock XT.  After all, the PCT was a TURBO XT.
> That's not saying much, however.
>
> >>> and it still had an 8-bit external bus.  I remember
> >>> quite well how my 6 MHz Z80 card in the Apple II often outperformed
> >>> the 4.77 MHz PC, except on floating-point intensive code, where the
> >>> 8087 in the PC did give a speed advantage the Apple II was unable to
> >>> catch up on.
> >>
> >> Of course that isn't a comparison between an Apple II and a PC, but
> >> between the CP/M computer in the Apple and the PC.
> >
> >Sorry, but it was still an Apple II: it used the Apple II's bus and
> >memory, and *all* I/O operations were done using the 6502.  No standard
> >CP/M disk format could be read -- all CP/M disks had to be in Apple II
> >disk format.
>
> Hmm.  Then, Paul, all your comparisons are invalid.  You
> know why?  All you said about the PC = Apple II.  Why?
> Because I can plug a PC Transporter into your Apple II+
> and it would still be an Apple II but an Apple II using an
> 80186 compatible processor!  ;-)
>
> Oh, an in the PCT, I/O operations except for video and
> optionally drives are also done using the 6502, 65C02
> or 65C816 depending on the Apple II model.  The PCT's
> hard drive is the Apple II's hard drive if you had one.
> The disk could be an Apple II disk drive if your MS-DOS
> programs were in Apple II format OR you can connect
> your Apple drive directly to the PCT and it would read
> PC format disks.  The video was CGA, directly from
> the PC although a video mixer allows both videos to
> display on the IIGS.  On an 8-bit Apple II, the video goes
> through the Apple II motherboard out the same
> composite output as your Apple II video.
>
> >> But take an Apple II that had a numeric coprocessor board installed
> >> (and there were such available) and it might have blown the doors off
> >> the Intel 8087 equipped machine too.
> >
> >Sorry, but the Apple II didn't "blow the door off" the 4.77 MHz PC
> >without coprocessor -- rather it was able to sometimes keep up with
> >the PC.  Equipping the Apple II with some coprocessor would hardly
> >"blow the door off" an 8087 equipped 4.77 MHz PC either, since you
> >still had that 8-bit data bus, on the Apple II as on the 4.77 MHz PC.
>
> I'm sorry, but this may be true if all you intend to do is
> crunch numbers for a specific purpose (astronomy?)
> but for normally application users, the Apple II was
> very often faster.  I know this because I remember
> being turned off by 4.77Mhz PC speed but found 1Mhz
> Apple IIe quite acceptable.
>
> >Also: the only coporcessors I ever saw for the Apple II did only
> >single-precision FP arithmetic (4-byte IEEE format), and to me it was
> >essential to have greater precision than that.  To me, the 5-byte FP
> >format of Applesoft was barely enough, the 6-byte FP format of early
> >versions of Turbo Pascal was OK, and the 8-byte IEEE double precision
> >format is fine.  But the 4-byte IEEE single-precision format was just
> >not enough for me - therefore I frowned at those coprocessors for the
> >Apple II that I saw.
>
> Maybe if you waited a year or two.  After that Mac
> was released, Apple II coprocessor cards using the
> 68881 chip appeared.  These Motorolla based FP
> cards had what you wanted.
>
> >> I found it interesting that the spreadsheet test was floating point
> >> math, and the Apple at one fifth the clock speed was crunching numbers
> >> in only twice the time of the PC.  I wonder what spreadsheet was used?
> >> Multiplan?
> >
> >If it was an early spreadsheet, it may have been written in assembly
> >language which had been machine translated from 8080 assembly to
> >8086/8088 assembly.  Such programs were quite common in the early PC
> >days: while it allowed a quick migration of software from CP/M to the
> >PC, performance did indeed suffer.
>
> Early PC programs were SO slow and that PCT is
> constantly reminding me.  I am sure it has to do with
> both hardware and software problems and the fact that
> the IBM name was enough to sell the computers and
> competitive speed was not really necessary.
>
> For the Apple II, programmers have a history of doing
> the best work they can.  Optimization is an obsessive
> trait in the average Apple II programmer.  That's why
> AppleWorks (the 8-bit version) has always been a very
> quick program.
>
> WordStar, WordPerfect, Lotus, Multiplan?  I shudder
> to think of the speed on 4.77Mhz PCs.  But I remember
> using 8-bit AppleWorks on a 1MB Apple IIe quite
> fondly.

Actually, WordPerfect 4.x was very fast on a 4.77 mhz PC. 5.1 was a different
story. I did notice that WordPerfect on a 128K //e is slow, esp in two areas -
you could type faster than it could display on the screen, esp at the end of a
line, and in disk access. Use of a RAM disk speeds that up considerably, and
using WP 1.1 on my TransWarped //e is not only acceptable, it's FAST!

>
>
> Thus, assuming you are right that CPU wise the
> 65C02 can only often keep up with the 4.77Mhz 8086.
> Well, that is more than enough to make almost ALL
> Apple IIe applications faster than almost ALL 4.77Mhz
> PC applications BECAUSE of the better programmers
> in the Apple II world and the better programming styles.
> And that was what happened.
>
> Not surprisingly, when AppleWorks was released, it
> beat out Lotus 1-2-3, the top selling PC program, as
> the top selling program across any platform in the
> United States.  ;-)

Roy