Done. Results from 11/83+FPP under RSX/Fortran 77
$ @PI.CMD
$ SHOW TIME
10:24:07 20-NOV-2025
$ RUN PI
PLEASE WAIT WHILE i COMPUTE THE VALUE OF PI...
AFTER 262144. ITERATIONS, I HAVE DETERMINED
PI IS APPROXIMATELY 3.141592653589796
$ SHOW TIME
10:24:29 20-NOV-2025
$ @ <EOF>
$
Verdict: Fortran 4 sucks in terms of compile performance. Use Fortran 77
Verdict: The Pro/380 maintains its mantle of the worst computer of all time in terms of performance.
(I was tipped off when the compile of the fortran program took seconds on the 11/83 and a minute or so on the Pro)
Sorry for jumping in a bit late on this thread. I was made aware of it from discord just now.
I'll make a bunch of different comments, but to start with the direct reflection on the above - yes, the Fortran IV compiler sucks compared to the Fortran-77 one. F4 basically generates threaded code with no optimizations, while the F77 compiler actually does proper compilation, and by default also does optimizations on the code. So the difference in the resulting binary can be huge.
I think there have been a little too much focus and misunderstanding of the Qbus in here, though. The cache on some of the CPUs are there because the memories at the time wasn't that fast. It's not that the Qbus can't be that fast. So with newer memories, there is just less gains with the cache. So an 11/53 without cache, but on board memory performing about the same as an 11/73 isn't that surprising. The 11/73 have cache, and will be running roughly at CPU speed most of the time. The 11/53 memory is probably allowing the CPU to also run at full speed, so it comes down to just CPU speed on both, and since they are both running the J11 at 15 Mhz, the performance should be comparable.
PMI memory improves performance some, but I'm not surprised the speed difference between the DEC PMI and Clearpoint isn't that great. It's the occasional memory read that can be faster, but again - you do also have the cache in there, so for most of the time, it isn't that much a factor. Faster booting (if I assume we're using the same controller and disk) can maybe have a little more impact since the cache isn't much affecting things in that scenario.
Of course, with on-board memory, and faster memories, DEC ditched both the PMI and cache on the 11/93. But performance isn't hugely different from an 11/83 anyway. The main benefits are that memory access by the CPU can happen without contention from Qbus (I think it can even be done sortof in parallel), and of course, the speed of main memory is fast enough that it's all about the CPU speed limit.
Any OS should have a pretty marginal overhead with interrupts and other activities going on, unless you are actually running something else actively at the same time.
And the FPJ11 is really just an FPA. It accelerates FP. The J11 does FP instructions with or without the accelerator. It just runs faster with.
As opposed to the FPF11 (was that the name?). Without that, the F11 cannot do FP instructions at all.
And sadly, yes, the Pro-380, while nice in some ways, really is sad in that it runs the CPU at 10 MHz, and also P/OS don't even allow you to use split I/D space, or supervisor mode. A machine with such potential but such a failure.
And obviously, the CPU can run faster, but as I heard things, some support chips can't, which is why the -380 only is running at 10 Mhz.
