Re: Accelerator //e

[pubpc1@library.ucla.edu]

Supertimer wrote,
in response to Mitchell Spector who wrote,
in response to my reply:
> >>Mitchell Spector wrote:
> >>
> >>>     Probably, but your still dealing with the limitations of the
> >>> 65C816 processor (even an ancient 386SX does math better, and in
> >>> a lot of cases a 286 or 8086) and the bus speed of the C64/C128.
> >>
> >>(1) When running my Zip GS set to 10Mhz, it runs everything, math
> >>    or not, faster than my PC Transporter (equivalent to a 186
> >>    10Mhz).
> >
> >    Are you sure about that? I recall several people mentioning
> >they were able to decompress ZIP archives faster through the
> >PC Transporter running MS-DOS unzippers, than with PMPUnZip.
> >The NEC V30, incidentally, is more similar to the 8086 than
> >the 80186.
> 
> Since I have an XT and a GS, I thought I would help settle this
> debate.
> 
> Experiment time!  The 8086 XT is older than the GS, so comparisons
> are fair game.  I dusted off my Turbo XT, which has a genuine
> Intel 8086 going at 8Mhz.  It has no math chip.  The XT is packed
> with 640K of RAM and has an MFM hard disk.  In the other corner,
> I have the accelerator on the GS set to 8Mhz.  GS has an SCSI
> hard disk and 4MB of RAM.  Let the games begin!
> 
> First test: decode a JPEG file.
> 
> The GS used a GS/OS JPEG decoder while the XT used JPG2GIF, a
> DOS JPEG decoder.  And they're off...both computers take forever
> to do their thing.  The winner is...the GS.  Part of the blame
> can be attributed to the XT having to use disk caching to get around
> the 640K limitation and running up against the slow MFM hard disk
> access, but I think the GS processor is genuinely faster than the
> 8086 in this task.
> 
> Test 2: unpack a .zip archive
> 
> The GS used Angel, an 8-bit decoder but considerably faster
> than PMPunzip.  The XT used pkZip's official .zip unpacker.
> And they're off.  The winner is again the GS.  This time, the
> XT has no excuse.  It has 640K of RAM while Angel is using
> just the 128K that it is accessing in IIe mode (even though
> the GS actually has 4MB).
> 
> I conclude from these tests that the 65816 is faster than the
> 8086 when running at the same speed in terms of Mhz.  It is,
> in my opinion, most comparable to the 286.
> 
> Does the PC Transporter use the additional memory of the GS
> as extended memory?  Does the PC Transporter have a math chip?
> If these two conditions are true and the PC Transporter has
> access to the faster SCSI drive of the GS, then that emulator
> may be faster than a regular Turbo XT.  I remember AE's claim
> that the PC Transporter was faster than a regular XT, but am
> skeptical that it was significantly faster than a real
> Turbo XT.  I can't believe that it would make much of a
> difference in the above tests.

Actually, the PCT is slower than a real Turbo XT.  It can't
use the GS' memory at all, so you are stuck with 640K.  And
you can forget about the faster speed of SCSI disk access.
The overhead caused by getting data from the drive to the
GS and then to the PCT negates any speed advantage.  In fact,
the disk access speed is horribly slow.  It is much slower
than even a PC/XT hard disk.

The PCT is a neat hack, but anyone claiming that a PCT can
outperform an accelerated GS has got his head screwed on
wrong.  :)

As for the math coprocessor, the PCT has a socket for it,
but it is not standard equipment.  There was talk about
the possibility of using a match coprocessor installed in
this socket in Apple mode.  Now _that_ would make the PCT
useful.

-Scott G.