Hardware races

From: Richard Erlacher <edick_at_idcomm.com>
Date: Tue Sep 18 08:39:25 2001

Something tells me you'll be disappointed if you expect a PDP-whatever to keep
up with a 486-66. If you stop and think about it, it's the rate at which a CPU
gets in and out of memory that determines how fast it performs, and the memory
for which the PDP-series processors were designed was quite a bit different from
what was commonly used with a 486DX2.

The typical '70's memory had an access time on the order of half-a-microsecond,
while the typical '486-motherboard had 10-15 ns CACHE on it, together with 60-70
ns (OK, maybe as slow as 80ns) main memory. That suggests that data can be
moved in and out of main memory at a rate of 12.5 MHz x 32bits, or at 50 MB/sec,
and into and out of cache at nominally 246 MB/sec, in bursts.

I don't know what the PDP's themselves can do, but the fastest memories I
remember seeing in quantity during the late '70's were of 45-55 ns speed, though
I'm sure there were some smaller faster ones. The popular types were 2147's and
2148's, organized as 4k x1 or 1k x4, respectively, though later in the '80's
there were some larger ones of comparable speed avaialble.

Were the PDP-types cached? What sorts of caches did they use? The Q-Bus cycle
timing, the only real contact I had with the PDP-types, though my experience was
with a micro-Vax, certainly didn't suggest it would operate at speeds
approaching those cache speeds, so the cache would have had to be on the CPU
cards.

I'd submit that a fair test would be, say, to calculate the value of PI, or some
other task which involves only integer computations, to, say, 250 places, using
code written for each processor in assembly language, thereby eliminating the OS
or a compiler as an influencing factor. Be sure system functions are turned off
while it runs, so only the test process is running, and use system calls only
for display and input functions. I'd say a more reasonable test would be a
PDP-type against, say, a comparably equipped '286. The last one I used ran at
25 MHz. The most common ones late in their market life ran at 16, and the first
ones ran at 6 or 8, depending on what you wanted to spend.

In the comparisons I saw run while I was in the aerospace industry, the
MicroVax-II compared more or less, with a PC/AT-clone at 12.5 MHz, just for
reference, but the PC disk subsystem was much slower than the uVax's so the test
was skewed. I've no idea how the uVax-II compared with the 11/34 you have in
mind.

Dick

----- Original Message -----
From: "Mike Kenzie" <KenzieM_at_sympatico.ca>
To: <classiccmp_at_classiccmp.org>
Sent: Monday, September 17, 2001 9:33 PM
Subject: Hardware races


> On another list someone mentioned that a PDP11 ran at about the same
> speed as a 486-66.
>
> I don't have a 486-66 but do have a PDP11/34 in the garage so I
> suggested that if a 486-66 could be found we could have a race.
>
> I tested the capacitors on the weekend and they were OK, still need
> to test the RL01 drives and get them hooked up.
>
> From the cables in the cabinet and the terminator on one of the drives
> I'm guessing that they are daisy chained.
>
>
> A few questions remain.
>
> What would be the best way to test the Power supply it is currently
> off the chassis, does it need to be connected for testing?
>
> Any suggestions for a fair race?
>
>
>
> Recent additions: Unisys PW2 (Unisys ICON NETWORK), Microcom (apple
> clone), SHARP PC4500
>
>
>
> Collector of Vintage Computers (www.ncf.ca/~ba600)
>
>
Received on Tue Sep 18 2001 - 08:39:25 BST