Benchmarking ADTPRO 2.0.1

[Keatah]

I had some extra time while running my mid-month backups so I figured I'd run some informal benchmarks on ADTPRO 2.0.1.

 

Configuration:

Pentium III 1.4GHz

Legacy 16550 ISA serial port on BX6-R2 mainboard

Apple //e 128K

Super Serial Card

Standard Disk II with standard controller

Disk for both send and receive tests was the standard ADTPRO-2.0.1.DSK image.

PC server 2.0.1 - 115200 baud

 

key:

r = send-to-pc from apple,

s = receive-to-apple from pc.

the number immediately after r/s is the block size

time is in seconds (avg of 3 runs, and ignoring the 1st)

 

2.0.1 ProDos client

r1 41.07
r2 39.92
r5 39.75
r10 39.41
r20 44.21
r40 44.10

s1 27.70
s2 27.85
s5 28.39
s10 27.35
s20 28.00
s40 27.46

2.4 DOS client
r 42.30
s 35.85

 

Observations:

The 2.4 client has mid-range performance and the cool-looking LockSmith-like Fastcopy screen. There's not much to be done in this older version that affects transfer times. Turning on/off the CRC has little or no effect on time. If there is, it got lost in the noise.

 

In client 2.0.1 I got the best overall performance with blocksize set at 10. Note that r1, r2, r5, r10 are steadily decreasing in time, albeit a small amount. This I would guess is from less and less checking/ack and communicating with the server.

 

But with r20 and r40 transfers take longer. This is because the motor doesn't start up in a timely fashion. r1, r2, r5, and r10 have enough granularity to where the disk motor is turned on prior to writing. With r20 and r40, the motor isn't turned on till the end of the transfer thus the added delay.

 

But still note that r40 is faster than r20 by a 11/100ths of a second! I ran this test 5x to be sure.

 

In r1 time is lost due to much checking & acknowledging.

in r40 we have a motor delay.

 

The sending-to-pc times are all about the same and the observed variances are most likely due to disk rotational positioning and other not-easily-measured system delays.

 

In any case, all the tests worked out where the Disk II heads were always resting at the innermost track at the start of each iteration, easiest way to get consistency across the board.

 

I'm BORED!!!!