View Full Version : performance of old pc/xt

July 19th, 2012, 11:51 AM
I got back here acidently
I was looking for the util to change speeds on someones 8 mhz turbo xt.

anyway" the post was about sluggish performance. of a pc/xt/clone.
there is a semi secret: change the mfm drive interleave.
newer systems with ide drives run at 1:1
meaning the system reads data from one sector to the next. without the drive spmnning around until the cpu can "catch it" . the AT controller normally the wd1003 may do 2:1 ? it has been years it could be 3:1 there are enhanced AT 16 bit controllers with a wd1006 chip that will
allow/permit 1:1 operation of a mfm drive such as a st225 or st 251,

But the Xt 8 bit card is even slower
It normally runs at 3:1., meaning every third sector is the NEXT sector to be read.
Actually this is too much for the computer.,
this provides 30k/s performance. feels like the drifve is full of thick grease,
Gibson of Spinrite fam had a util called spintest., that determines the best and most efficient interleave., it could be 4:1 ( typically used by rll drives) or even 5:1.

this is applicable to the western digital controllers. the XEBEC card used in early pc/xt is very slow.
What you do it run a utility to determine the best interleave.
it will increase data transfer from 30k/s to 130k/s and make an old pc xt "snappy"
Since these 8088 machines come in 4.77, 8,h and 10 mhz, and the Juko with a v20 is 12 mhz. You cannot determine with 'slower " interleave is " fastest"
You have to test it. then back up everything and do a fresh low level format.

I just thought I would post this as it will help those trying to wring the best out of old pc's.
BTW using an 8 bit XT controller in a 286 386 system works.
you set the drive to NONE., but because of the mis-mach it is really s l o w.
years ago I could not afford even a decent 286, so I was stuck at the xt-8088 level too long.,
This is something I learned than and I am passing it on so it will help those who
were not around 30 plus years ago.
last time I posted about floppies
it was deleted. as the administrators did not understand.
I left for several years.
This is not theory but based on real experience and knowledge., things many never knew or have forgotten.

Unless someone who was there tells you how will you never find out.

July 19th, 2012, 10:02 PM
Yes Spinrite is a handy utility, Norton's CALIBRAT[e] did the job too :)

the XEBEC card used in early pc/xt is very slow. What you do it run a utility to determine the best interleave. it will increase data transfer from 30k/s to 130k/s and make an old pc xt "snappy"

Feels even snappier at 400KB/s :D

July 19th, 2012, 10:12 PM
There are a few XT controllers that claim 2:1 interleave capabilities.

In reality, particularly in the world of the 4.77MHz 8088, it matters how fast the application can process data. I could see varying the interleave between, say, FAT and root directory tracks and application data areas.

July 20th, 2012, 11:57 AM
My xt controllers were either Western digital
or older wd with updated bios
and two dtc? that had an actual drive table
but the llformat was incompatible with the wd cards
when I tested things
the fastest I ever saw with an 8988 machine was 130.

this is a physical / mechanical thing. the software does not matter.
if the exe or com file was larger,m it took longer to load
certain programs like 123 used the 8087 co-processor and seemed to run better when recalculating.
but few programs took fulkl use of an 8087.

So under any circumstances using a bigger but more efficient interleave always helped.

July 21st, 2012, 11:47 AM
spinrite : later versiosn would fit on a boot floppy
the first version required that the 360k floppy be rebooted to "activate" the program ??
V 5 was jst a file
spintest did not do a low level format
or test for errors
it simply tested data transfer rates at differing interleaves
then debug was used to ll format a mfm or ( heaven forbid)
a rll drive. You had to remeber the best and most efficient interleave

BTW a st-225 was 615 4 616 616 11 5 9think that is right)
changing that 5
to 7 as suggested in some old instructions
made the driev very slow
also the short bios chip on the 2/3 length cards could be replaces with the chip from a newer card
the eaqrly chip skipped right by the option to enter numebers
if it was set for a 210m with 2 heads and you had a 20m with 4 heads., you got 10m. the manual, I have a typed in copy explains all.

July 31st, 2012, 06:56 PM
a few more things that might be obvious

I had several different drives
a 20m miniscriev ran VERY slowly on one wd 1003 controller
ither identical drives were slower tha n seagate -st-225 but were acceprible another controller even with swapped bios chips did not help
it was the combo of that drive and that card. If it happens to you just mark that drive and card.

the st-251 were really in three categories 40mc for the 252 sand either 28 or 40 for the st-251-1. You habe to test to know. They said it was possible some of the early drives were really 28ms drives.
The very best mfm driev was the mitsubishi mr-535
it could either be mfm 40 or rll 60.

August 1st, 2012, 08:36 AM
The very best mfm driev was the mitsubishi mr-535
it could either be mfm 40 or rll 60.

I'll agree with that......I still have the Mitsubishi MR-535 I purchased new in 1988 ($250) and it still works fine with the controller
I purchased at the same time. The controller is an AMS-1100 RLL.

I have a similar vintage Microscience HH-1060 RLL drive 65MB which I also run off the same controller.