Looking for Bendix G-15 computer

Learned programming on one at UC Berzerkley, basement of Cory Hall as i remember. Try a query to one of the older instructors in the computing department?

The people who actually used these are mostly retired or passed away. The only ones who remember them are mostly people who got to play with them after the machines were "semi-retired". I'd guess the G-15 went into a very rapid retirement when core memory and discrete transistors came out in the very early 1960's. A PDP-5 could mostly run rings around a G-15. When programmed to the umpteenth, with the instructions optimally scattered around the drum, you could get about 3000 instructions/second out of a G-15. A PDP-5 was a 12-bit parallel machine, and could do at least 50,000 IPS. The memory worked out to about twice the size of the G-15's. I don't know about the reliability of the G-15, but a PDP-5 was a very reliable system, the only thing I remember ever having to fix was worn-out console switches.

Jon

Gees, our EE department had G20s (the semiconductor version) that they got for $.01/lb., including shipping, about that time. They had three, though one didn't leave the crates intact (raised floor queen).

As i vaguely remember, there was a small "scratchpad" core memory which i think was used as a lookup for multiply (??). Darn good for learning especially for organizing a program (read: optimizing).

According to the wiki, multiply (Divide, and square root) was done in a built-in subroutine. The IBM 1620, A.K.A. the CADET (Can't Add, Didn't Even Try), used lookup tables for ADD.

Interesting stuff at:

No, the machine was designed entirely too early for core. I am fairly familiar with it, I tried to help resurrect one about 1971. Jay W Forrester (with the help of my freshman adviser William Papian) got the first coincident current core memory working in 1951. It took a few years for industry to gear up to produce commercial core stacks. The G-15 was adapted from the Turing ACE. Although the G-15 was brought out in 1956, the design of the machine started much earlier. I wouldn't be too surprised if somebody adapted a core stack to the machine for some purpose, but the basic cycle time of the machine was SOOOO slow, that it would not have helped much on speed. It certainly would have helped expand the critically tiny memory on the G-15, though. The G-15 was a serial-arithmetic machine, and optimally executed 3000 instructions/second. But, it was a rare program that could come close to that performance. Worst-case was a program totally un-optimized for next instruction location, it would run at 50 IPS (the delay of the long tracks on the drum).

Jon