FPGA VAX update
ajp166 at bellatlantic.net
Sun Oct 23 17:28:07 CDT 2005
>Subject: Re: FPGA VAX update
> From: shoppa_classiccmp at trailing-edge.com (Tim Shoppa)
> Date: Sun, 23 Oct 2005 18:03:34 -0400
> To: cctalk at classiccmp.org
>Paul Koning <pkoning at equallogic.com> wrote:
>> >>>>> "Michael" == Michael Sokolov <msokolov at ivan.harhan.org> writes:
>> Michael> Chuck Guzis <cclist at sydex.com> wrote:
>> >> If PC-based emulators are fast enough, why bother with the IC
>> >> version?
>> Michael> 2. A pee sea based emulator requires a pee sea. I refuse to
>> Michael> contaminate my house with a pee sea.
>> SIMH would run on a Mac, wouldn't it?
>SIMH will run on a VAX, in fact :-).
>Getting back to FPGA's, I know of a couple FPGA implementations of
>PDP-11's. (They are mostly KDJ11 clones, but they differ in a couple
>of tiny respects.) With extreme effort in the late 90's, they managed
>to make 4 FPGA's be about a factor of 2 factor on most benchmarks than
>By the time the FPGA implementations made it to market the PC-based
>emulators were so much more cost-effective for most applications
>(despite their warts of running under a host OS...)
>I suspect that a FPGA implementation of a VAX would have a
>performance about equal to a 11/780 if done by an average Joe.
>Someone with much experience in caching/pipelining could probably
>eek out a factor of 2x or 3x by pulling out all the tricks in the
Considering that 2901C (750ALU) was 100ns (1982 timeframe) I'd think
by now the latest FPGAs are easily able to beat that by a factor of
25 or more.
So I think that's on the low side. Without pipelining to any great
extent and the availability of far faster memory than the mid 80s.
I'd think 10x is shooting low.
Seriously the more you put on the chip the faster it will go as busses
are still working at ~1ns/ft and them chips are at best less than .2".
The cray XMP was the tour de force around those limits back in the
I'd think the real factor is that the VAX was CISC to the extreme and
the microVAX (320780 chip) was on the order of 1-2million transistor
equivilent gives one an idea of the task size though.
More information about the cctalk