Friends, here’s a very deep dive (to be continued!) in the Decmate II : hardware, software, compatibility (or lack thereof) with the PDP-8 : more info than I ever had about this series of machines:
6 Likes
Oh this is a little surprising:
Intersil’s 1975 development of a pure CMOS PDP-8 CPU was completed independently of DEC, intending not only to create a microprocessor of their own but also one with a presumably guaranteed market … its features included moving the Extended Arithmetic Element (EAE) option on-chip for multiplication and division. Intersil sampled and sold the IM6100 openly, seeing its first use in Pacific Cyber/Metrix’s long-lived PCM-12 computer line, but nevertheless offered it to DEC as well in the fall of 1976. Harris Semiconductor came on as a second source…
DEC management was reportedly surprised to learn of the IM6100’s existence but found its implementation (based on the PDP-8/E) sufficient, and Olsen permitted what DEC called the “CMOS-8” to ship in the 1977 DECstation VT78, a trial low-end machine.
Edit: and this is a bit of an oddity, about the floppy drives:
RX50 sectors are 8-bit (we’ll get to the exact disk format later), but because the DECmates are 12-bit, the controller can also perform disk access directly in 12-bit mode for faster operation. However, changing the bit width does not change the disk’s storage density, and 12-bit mode is stored effectively as 16-bit with four bits wasted, cutting capacity in half.
4 Likes
256 words vs 512 bytes - I guess that’s ‘cutting capacity in half’ but… you do get 75% of the number of bits…
PDP bits are 25% BIGGER over the average micro. 
12 bits do the work of 16.
I bet there is a lot of nasty tricks needed to have a floppy disk,as you need it to
act like a hard disk interface wise with DMA.. SPARE TIME GIZMOS did a lot of work to interface a modern disk. Interfacing was not a problem, finding room to fit the IO software was.
1 Like
Their bits are the same size. Their bytes, on the other hand…
Yes, they’re quite chunky that way. The STG solution for RAM was to store one word in 16 bits, wasting 4 of them.
This article, plus some additional discussion about it on lobst.rs, makes me wonder if there was a specific OS/8 distribution for the HD-6120 in the DECMates. I didn’t realize it was different from the Intersil 61xx series, and it might explain why I was having difficulty with double precision in Fortran.
1 Like
That’ll be this discussion, I think.
1 Like