29 Sep
2005
29 Sep
'05
12:33 p.m.
From: "Michael B. Brutman" <mbbrutman at
brutman.com>
I don't think that the lack of a DMA controller on the PCjr really
slowed it down any. The old double density drives have a data rate of
250,000 bps. If the data is presented one byte at a time, that is just
31250 bytes per second. Even with multiple I/O clock cycles per byte to
get it to the processor and then to memory, a 4.77Mhz processor can keep
up with that easily.
Where the lack of DMA would hurt you is on a much faster interface, such
as a hard disk. Then you wouldn't have any cycles to spare, the the
difference would show up.
Hi
I don't think the early PC's used DMA for the hard disk.
The HD controllers usually had sector or track buffers.
The program would wait until the buffer was full and
then just move it by software to memory. The floppies
needed DMA because the controllers didn't buffer more
than one byte. The processor would have had to dedicate
it self to the one task without interrupts.
Dwight
29 Sep
29 Sep
3:06 p.m.
On 9/29/2005 at 9:33 AM Dwight K. Elvey wrote:
I don't think the early PC's used DMA for the
hard disk.
The HD controllers usually had sector or track buffers.
The program would wait until the buffer was full and
then just move it by software to memory. The floppies
needed DMA because the controllers didn't buffer more
than one byte. The processor would have had to dedicate
it self to the one task without interrupts.
PC's went away from DMA in the AT, but the XT used DMA channel 3. The XT implemented
only a two-sector buffer, so DMA for multiple sector reads made sense (on the controller
board itself, double-buffering was used).
The XT controllers were (the 10 MB controller was different from the 20MB one) a peculiar
beast with a bunch of proprietary LSI. When the AT version was implemented, the Western
Digital type (WD1000) of controller design was used, with (faster than DMA) programmed I/O
transfers.
Cheers,
Chuck
30 Sep
30 Sep
6:51 p.m.
The XT controllers were (the 10 MB controller was different from the
20MB one) a peculiar beast with a bunch of proprietary LSI. When the
IIRC, the TechRef for the 10MByte controller contains a schematic with 3
chips simply labelled LSI A, LSI B and LSI C. It was fairly obvious from
the piuout that one of this was nothing other than a Z80 processor. I
think one of the others was something standard too.
-tony
9:02 p.m.
On Fri, 30 Sep 2005 23:51:10 +0100 (BST)
ard at p850ug1.demon.co.uk (Tony Duell) wrote:
>
> The XT controllers were (the 10 MB controller was different from the
> 20MB one) a peculiar beast with a bunch of
proprietary LSI. When
the
IIRC, the TechRef for the 10MByte controller contains a schematic with
3
chips simply labelled LSI A, LSI B and LSI C. It was
fairly obvious
from
the piuout that one of this was nothing other than a Z80 processor. I
think one of the others was something standard too.
Is that the Xebec controller? (can't remember if that's who IBM got the
original controller for the XT from or not)
-tony
1 Oct
1 Oct
4:36 p.m.
IIRC, the
TechRef for the 10MByte controller contains a schematic with
3
chips simply labelled LSI A, LSI B and LSI C. It
was fairly obvious
from
the piuout that one of this was nothing other
than a Z80 processor. I
think one of the others was something standard too.
Is that the Xebec controller? (can't remember if that's who IBM got the
30 Sep
30 Sep
11:55 p.m.
On 9/30/2005 at 11:51 PM ard at p850ug1.demon.co.uk wrote:
IIRC, the TechRef for the 10MByte controller contains a
schematic with 3
chips simply labelled LSI A, LSI B and LSI C. It was fairly obvious from
the piuout that one of this was nothing other than a Z80 processor. I
think one of the others was something standard too.
Probably so, but I never substituted a "stock" Z80 to see if there was something
different about the IBM version. OTOH, the "clone" controllers like Xebec
certainly used the Z80.
Cheers,
Chuck
1 Oct
1 Oct
1 p.m.
>IIRC, the TechRef for the 10MByte controller
contains a schematic with 3
>chips simply labelled LSI A, LSI B and LSI C. It was fairly obvious from
>the piuout that one of this was nothing other than a Z80 processor. I
>think one of the others was something standard too.
On Fri, 30 Sep 2005, Chuck Guzis wrote:
Probably so, but I never substituted a
"stock" Z80 to see if there was
something different about the IBM version. OTOH, the "clone"
controllers like Xebec certainly used the Z80.
The XT HD controller was a relabelled Xebec.
AT switched to WD.
3:31 p.m.
On 10/1/2005 at 10:00 AM Fred Cisin wrote:
The XT HD controller was a relabelled Xebec.
I've no doubt that Xebec made the 10 and 20 MB adapters for IBM (even the silk screen
legends correspond to Xebec convention). But does the IBM 10 MB adapter correspond to a
standard Xebec part, or were there changes incorporated into it (beyond the BIOS
signatures, that is)? My 10 MB Xebec adapter built in early '85 has a different
layout from either 10 MB or 20 MB IBM parts.
FWIW, the Z80A on the 20MB adapter is so labeled in the IBM O&A reference manual.
Cheers,
Chuck
29 Sep
29 Sep
8:22 p.m.
On Thu, 29 Sep 2005 09:33:32 -0700 (PDT)
"Dwight K. Elvey" <dwight.elvey at amd.com> wrote:
Yep. The floppy disk interface on the PC Junior is painfully slow
because all data has to pass through the CPUs accumulator.
>From: "Michael B. Brutman" <mbbrutman
at brutman.com>
>
>
>I don't think that the lack of a DMA controller on the PCjr really
>slowed it down any. The old double density drives have a data rate
of
>250,000 bps. If the data is presented one byte at
a time, that is
just
>31250 bytes per second. Even with multiple I/O
clock cycles per byte
to
>get it to the processor and then to memory, a
4.77Mhz processor can
keep
>up with that easily.
>
>Where the lack of DMA would hurt you is on a much faster interface,
such
as a hard disk.
Then you wouldn't have any cycles to spare, the the
difference would show up.
Hi
I don't think the early PC's used DMA for the hard disk.
The HD controllers usually had sector or track buffers.
The program would wait until the buffer was full and
then just move it by software to memory. The floppies
needed DMA because the controllers didn't buffer more
than one byte. The processor would have had to dedicate
it self to the one task without interrupts.
Dwight
6843
days inactive
6845
days old
8 comments
5 participants
participants (5)
-
ard＠p850ug1.demon.co.uk -
cclist＠sydex.com -
chenmel＠earthlink.net -
cisin＠xenosoft.com -
dwight.elvey＠amd.com