Tony D. wrote:
> I may be talking nonsense, but you describe the Tektronix 4054 as a > 6800-based system. I > thought the 4051 used that processor, but the 4052 and 4054 used a board of > AM2900-series > bitslice chips that implement a processor with an instruction set similar to > the 6800 > but with no BCD operations and some 16 bit extensions. No nonsense at all, Tony. You are correct. The architecture of the 4052 and 4054 were definitely bitslice microcoded implementations of the 6800 CPU, with the omissions and additions as you mentioned. The memory architecture of these computers was also heavily modified to allow bank switching of RAM and ROM programmatically to create an address space that provided more memory capacity than the 64K address space of the 6800 used in the 4051. The display subsystem was also modified to allow additions of graphics co-processors for doing things like display list processing and refresh vector graphics which were drawn at a lower intensity so that the vectors would not be stored on the screen. Later, a special CRT was made for the 4054 that had a unique storage CRT that had two layers, a green layer and a yellow/orange layer of phosphor that could be triggered by different beam intensities, both in storage and "write-through" mode that allowed three-color (yellow/orange, green, and a mix of the two IIRC), refreshed vector graphics at up to 1000 vectors per second. Still, it was just fast enough for simple animated graphics for things like games and graphical editors that would use refresh graphics for placing an object, then writing it in storage mode once it was placed. The special three-color tube was not available in the 4052, only the 4054, but the add-on boards for providing the vector write-thru display were compatible with the 4052. The microcoded CPU and improved memory architecture of the 4052/4054 made the machines significantly faster at compute-bound tasks. I/O and display were not sped up much by the new architecture, as the speed at which vectors and text could be written to the storage tube display was limited by the tube itself, so graphics intensive stuff wasn't all that much faster than the 4051. If the graphics involved computing the vectors in real-time, that type of graphics would be faster on the 4052/4054 due to the significantly faster computing speed of the 2901-based 6800 "clone/extension". Things like GPIB I/O were more limited by the peripheral devices than the CPU itself, so things like I/O to the GPIB 4907 8" floppy disc drives, wasn't all that much faster. I have a 4051 and a 4052A, both working. Comparing them side-to-side doing compute bound things (like finding prime numbers) clearly shows the speed advantage of the bit-slice architecture in the 4052. Drawing "canned" graphics is slightly faster on the 4052 simply because the interpretation of the BASIC code that does the drawing runs significantly faster, which does make a small, but noticeable difference in the time it takes to render an image, with the 4052A finishing any given drawing a bit sooner than the 4051. Floppy disc access on the 4907 doesn't seem to be much faster other than the faster speed of interpretation of the BASIC program, with the actual speed of reading/writing being about the same due mostly to the speed of GPIB transactions, and the fixed rate that data is read/written to the floppy. The 4050-series computers were quite amazing for their time. Nothing else except hugely expensive graphics systems that ran on minicomputers, such as those made by Evans & Sutherland and others, could exceed the capabilities of the 4050-series machines (especially the 4052 and 4054), and the 4050-series machines fit on a desktop and were (other than being rather heavy) relatively portable, very easy to use/program, and cost dramatically less than anything else. DVST was a great technology at a time when large amounts of high-speed random access memory was very expensive. Magnetic core that was fast enough was quite expensive and complex, and IC-based RAM was just beginning to have reasonable capacity, but still ran somewhat slowly, and was also initially quite expensive. As the price of fast, high-capacity IC-based RAM came down, raster type display systems with bitmapped display memory, and even dedicated blitter hardware for shifting bits around in display memory, made cost-effective machines with at least equivalent (monochrome) display capability in terms of resolution, along with everything (including characters) being refreshed graphics straight out of the framebuffer RAM. Once that occurred, the market for DVST shrunk quite dramatically. Desktop workstations (like Sun, Apollo, Perq, etc.) with graphics capabilities that met or exceeded those of the 4050-series quickly took the place of these watershed machines. -Rick -- Rick Bensene, Curator The Old Calculator (and some computers, too) Museum Beavercreek, Oregon USA
