Re: Apple II benchmarks

[pausch@saafNOSPAM.se (Paul Schlyter)]

In article <3934BB53.D4508129@pacific.net.au>,
Matt Jenkins  <mdj@pacific.net.au> wrote:
 
> John King wrote:
>> 
>> FWIW regarding speed, I was pretty underwhelmed at the speed differences
>> between a Tandy 1000a (4.7 8088), and a Model One @ 2.4 MHz.
>> 
>> "Roy and/or Janet Miller" <millers@inetnebr.com> wrote in message
>> 3933FF70.BD9334D4@inetnebr.com">news:3933FF70.BD9334D4@inetnebr.com...
>>> I was reading an old Byte last night, and article comparing an 8086-2
>>> based clone against an IBM PC and an Apple IIe running ProDOS. Some
>>> interesting comparisons:
>>>  Disk read/write in BASIC
>>>     A2 - 36/37  PC - 46/56
>>> Sieve test in BASIC
>>>     A2 - 245 secs  PC- 191 secs
>>> Calculations in BASIC
>>>     A2 - 94 secs  PC - 69 secs
>>> Spreadsheet recalc
>>>     A2 - 22 secs  PC - 11 secs.
>>>
>>> Not too bad, pitting a 1 mhz 6502 against a 4.77 8088. Oh, on the last
>>> one, how did the 8086 do? at 4.77 exactly the same as the PC. By the
>>> way, the Calculations was 10,000 multiply and divide operations, the
>>> spreadsheet was recalculating a 25 by 25 cell sheet with each cell 1.001
>>> times larger than the cell to its left. And my IIc+ is about 3.5 times
>>> faster than the IIe.
> 
> The really surprising thing is that the Intel chips have hardware
> multiply, so those machines should have eaten the IIe for breakfast,
> but they didn't.
 
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?
 
> I conducted tests like that years ago trying to prove to a XT equipped
> friend that clock speed isn't everything. Needless to say, a IIe with a
> transwarp card fitted (or any other accelerator) embarrases 4.77 Mhz PC's.
 
The 4.77 MHz PC did indeed not perform particularly well: its clock
speed was slow, 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.
 
> That said, I don't think the tests in BASIC are very fair, since it's
> benchmarking the BASIC interpreter as well as the hardware. Ultimately,
> such tests need to be hand tuned assembly to show which machine is
> actually faster.
 
Depends on what you want to test.  If you want to find out which machine
runs your BASIC programs fastest, it appears to me at least quite reasonable
to include the Basic interpreters (or compilers) in the test, since you
intend to use them on the target machine.  Few people code much in raw
assembly.
 
Also, if you mainly intend to run I/O bound application, the CPU performance
may be less important than the I/O performance.   Et cetera et cetera.
 
-- 
----------------------------------------------------------------
Paul Schlyter,  Swedish Amateur Astronomer's Society (SAAF)
Grev Turegatan 40,  S-114 38 Stockholm,  SWEDEN
e-mail:  pausch at saaf dot se   or    paul.schlyter at ausys dot se
WWW:     http://hotel04.ausys.se/pausch    http://welcome.to/pausch