yet another pdp-11 in fgpa
Walter F.J. Mueller
w.f.j.mueller at gsi.de
Sat Jun 26 17:48:56 CDT 2010
Johnny Billquist <bqt at softjar.se> wrote:
> Good thinking.
> But I'm surprised by some numbers here. The J11 at 20 MHz is
> only slightly faster than an 11/70. In fact, if you can throw
> the 11/70 into running all from cache, it might even be slightly
> faster than an 11/9x.
> Or so I seem to remember from looking at the numbers back when
> I last was digging into this.
> Maybe I'm mixing some numbers up here... What I do remember for
> sure is that the 11/9x machines run at 20 MHz, and that they are
> not more than maybe 1.2 times the speed of an 11/70 in general.
oops, you are right, I was mixing numbers here and forgot about
a factor of four. Here the revised arguments:
See http://www.village.org/pdp11/faq.pages/prfmnc.html , there
is a table comparing 11/70 with various J11 systems:
11/23 11/53 11/73 11/83 11/93
----- ----- ----- ----- -----
CPU F-11 J-11 J-11 J-11 J-11
Microcycle(ns) 300 267 267 222 222
Clock (MHz) ? 15 15 18 18
Performance 0.2 0.5 0.7 1.2 1.4
(11/70 = 1)
Cache no no yes yes no
Floating-Pt opt no no yes yes
My mistake was to forget that the J11 needs 4 clock cycles per
microcycle (MC), that's why 18 MHz clock leads to 222 ns MC period.
In the end the J11 (222ns MC) is faster than the 11/70 (150 ns MC)
because it has a better micro architecture (three stage pipeline
vs. instruction prefetch only). The J11 is roughly a factor two
better compared to 11/70 in terms of MC efficiency.
The w11a on the other side does one microcycle per clock cycle. So
one can expect that is roughly a factor 5.5 faster than a 18 MHz J11
(MC rate is 11 times higher (50/4.5), but J11 has a factor two better
Let's look at the Dhrystone benchmarks again:
w11a 11500 lps
11/53 830 lps
So the w11a is measured to be ~14 faster than a 11/53. The 11/53
has half the performance of a 11/70, so one expects that the w11a
is about a factor 7 faster than a 11/70. Because w11a and 11/70
have a similar micro architecture that should match the MC rate,
and indeed it does (50/6.7). So the numbers are consistent, finally.
Thanks for pointing that out.
More information about the cctalk