[comp.unix.xenix] Simple disk performance test.

larry@tapa.UUCP (Larry Pajakowski) (03/09/88)

We just got a Compaq-386/20 with the Compaq 130mb. disk.
I did a crude measure of disk performance on the Comaq under SCO Xenix-286
2.2.1 (Xenix 386 is waiting on some Excelan drivers).  I also did the
measurements on an IBM PC/AT at 8mhz with the stock CMI drive.

Yes I know that more than just the machine is being test here but that is the
whole point.

The measurement consisted of:

   time dd if=/large_file of=/dev/null bs=30k		#for file reads

   time dd if=/dev/rhdxx of=/dev/null bs=30k		# for raw reads


               Raw Reads             File Reads

Compaq         550K/sec.             220K/sec.

PC/AT          120K/sec.              26K/sec.

All the factors (1:1 interleave, ESDI, 20mhz. processor, fast seek times)
really add up to impresive performance.  This generally agrees with the
overall feel of the machine.

Perhaps others could post the results of like tests on other machines and
software combinations.  It would make an interesting comparison.

det@hawkmoon.MN.ORG (Derek E. Terveer) (03/12/88)

In article <579@tapa.UUCP>, larry@tapa.UUCP (Larry Pajakowski) writes:
> We just got a Compaq-386/20 with the Compaq 130mb. disk.
> I did a crude measure of disk performance on the Comaq under SCO Xenix-286
> 2.2.1 [..]
>    time dd if=/large_file of=/dev/null bs=30k		#for file reads
> 
>    time dd if=/dev/rhdxx of=/dev/null bs=30k		# for raw reads
> 
> Perhaps others could post the results of like tests on other machines and
> software combinations.  It would make an interesting comparison.

Well i just did the crude test, mentioned above, on my Acer-1100 running uPort
386 with a 6085 miniscribe drive (72M), 3:1 interleave (i think):

time dd if=156k_file of=/dev/null bs=30k	/* file */

  156k
--------  ==>  96k/sec
user+sys

time dd if=1243k_file of=/dev/null bs=30k	/* file */

  1243k
--------  ==>  73.9k/sec
user+sys

time dd if=/dev/rdsk/0s0 of=/dev/null bs=30k	/* raw */

 71,680k
--------  ==>  340.9k/sec
user+sys
-- 
Derek Terveer	det@hawkmoon.MN.ORG	uunet!rosevax!elric!hawkmoon!det

terry@wsccs.UUCP (terry) (03/16/88)

In article <579@tapa.UUCP>, larry@tapa.UUCP (Larry Pajakowski) writes:
> We just got a Compaq-386/20 with the Compaq 130mb. disk.
                ^^^^^^^^^^^^^
> All the factors (1:1 interleave, ESDI, 20mhz. processor, fast seek times)
> really add up to impresive performance.  This generally agrees with the
> overall feel of the machine.

	There are a great number of complaints I have with the 386/20, most
of which stem from the use of standard AT-style bus (Edge-triggered) hardware.

Before I get flamed, let me point out that I am not trying to brow-beat you
into a PS/2-80 or anything.

IBM did one thing right: They went level-triggered.  This is _not_ better
because it is faster (it isn't), it is better because it allows you to add
boards with a slower-than-20-meg onboard clock and still have them work.

Anybody who tries to hook up an intellicom or computone board to a 386/20
running Xenix (any version that'll run on the '20) can tell you horror
stories.  No, I am not talking about the "phantom math processor"; that was
fixed in 2.2.2 rev k.

There are some basic design incompatabilities with the cheaper boards here,
kids.  A fast machine is nice, but a reliable one is nicer.  If you absolutely
_MUST_ have a '20 for the speed, choose you 2nd party boards _very_ carefully
and expect to spend more for them.


| Terry Lambert           UUCP: ...{ decvax, ihnp4 }                          |
| @ Century Software          : ...utah-cs!uplherc!sp7040!obie!wsccs!terry    |
| SLC, Utah                                                                   |
|                   These opinions are not my companies, but if you find them |
|                   useful, send a $20.00 donation to Brisbane Australia...   |
| 'There are monkey boys in the facility.  Do not be alarmed; you are secure' |