Ikonas Graphics Systems - The world's first GPGPU

by Nick England - send e-mail

2005 News - Still useful after 25 years! We introduced the Ikonas RDS-3000 in 1980 and as of July 2005 there were at least 5 systems still in daily use (driving cockpit displays in the USAF B-1 Cockpit Procedures Trainer).  Update 2007 - I understand they have finally replaced the Ikonas RDS-3000 systems with PC-based graphics.- That's 30 years since I got the prototype system running at NCSU.

July 1998 Demo- Thanks to Garret Dunn (former Ikonas engineer) and Bill Cowan (University of Waterloo - "Graveyard of the Ikonas boxes"), Mary and I exhibited a working Ikonas system at the 25th anniversary SIGGRAPH conference in Orlando in July 1998. We ran the "accept" microcode demo, hosted by our original Heathkit LSI-11 with an 8" floppy disk drive. It worked perfectly the whole show. We  essentially recreated our SIGGRAPH '81 booth with some signs, posters, and demo video tape. We also showed the NYIT Ikonas demo reel thanks to Paul Heckbert. Lots of old Ikonas customers came by to say hello, including one who still uses his 15 year old system regularly.

Customer List - We put together an Ikonas customer list for the SIGGRAPH retrospective. Please e-mail me with any additions or corrections.

History - Ikonas Graphics Systems Inc. was founded in 1978 by Mary Whitton and me, based on work I did as a grad student at North Carolina State University in the Computer Graphics Lab headed by Prof. John Staudhammer.  As part of a project funded by NASA Langley Research Center, I had developed a programmable raster display system for cockpit instrumentation display. (See my paper in SIGGRAPH 78). Several people expressed interest in acquiring something similar so Mary and I started Ikonas in the back room of our house to make a commercial version. Robert Whitton was a partner in the business from the beginning and later Pete Evans and Henry Rich became partners also. At one point Ikonas had around 50 employees in Raleigh NC. We showed the first production  Ikonas Raster Display System (RDS-2000) at SIGGRAPH '79 and introduced an improved compatible system as the Ikonas RDS-3000 in 1980. (Retroactively, we named the prototype system the RDS-1000.) Ikonas was acquired by Adage Inc in 1982 and the Ikonas system continued to be marketed as the Adage 3000 Raster Display System. I'd guess almost 400 systems were sold before Adage finally pulled the plug around 1987. As of 7/05 at least five were still in daily use, a minor miracle in the computer business.  There were lots of great people who were part of Ikonas and Adage and we had lots of great (and patient) customers.  Thanks for the wonderful experience. Nick England Further Info - For some more info on the system, see an article I wrote in the January 1986 IEEE Computer Graphics and Applications magazine. Also see Tim Van Hook's SIGGRAPH 86 paper covering some amazing real-time solid modeling microcode he wrote for the Ikonas. June 1985 -  Tim Van Hook's ray tracing code rendering bi-cubic B-spline surfaces entirely in the Ikonas Graphics Processor (world's first GPU ray-tracing)
MARKETING LITERATURE Download 1978-82 handouts Download Systems Specs & Prices 1978-82 - part 1, part 2, part 3 Download Application Notes Download User Group Newsletter
1978 - SIGGRAPH '78 handout (no exhibit) - first purchase order (thanks to OSU)
1978 mailings - RDS-1000 system
1979 - SIGGRAPH '79 exhibitor - RDS-2000 introduction
			RDS-2000 "Image Memory & Video Control" RDS-2000 "Ikonas Processor & Sequencer" RDS-2000 "System Configuration Guide"   more info and photos to be added some day
1980-82 - RDS-3000 Introduction - See "Summary of RDS-2000/RDS-3000 Differences".
1980 magazine ad	1981-82 magazine ad	1982 magazine ad		1985 Poster - Ray-tracing in Ikonas GPU
1984 - pdf download of brochure shown below - renamed Adage 3000

Architecture - Download Ikonas User's Manual

Ikonas developed flexible, programmable, high end graphics and imaging hardware for research and advanced application groups. The system was based around a 32 bit data, 24 bit address bus into which various boards could be plugged. Everything within the system was memory mapped into this 24 bit address space. A host interface was provided with address registers to access anything on the bus. Thus, unlike other systems at that time, the host computer could transfer data directly to/from image memory, or could communicate with the optional programmable graphics processor via shared memory.

Major elements of the Ikonas system were a flexible frame buffer for display and a user-programmable graphics processor for performing graphics and imaging functions. The frame buffer resolution and timing could be set via control registers. The graphics processor included 32 bit integer ALU and 16x16 bit integer multiplier along with address counters, loop counters and the like, all controlled by a 64 bit instruction word. Initially we just had a microcode assembly language compiler, but later Preston Gurd at University of Waterloo developed a C compiler as did Gary Bishop at UNC-CH.

Other plug-in boards developed during the lifetime of the system include: microprogrammed 16 bit graphics processor with 16 pixel-at-once parallel write microprogrammed 16x16 bit matrix multiplier microprogrammed floating point matrix multiplier hardware Z-buffer (frame buffer memory with internal compare and conditional write) real-time alpha-blend hardware for two RGB images 68000 processor with hard disk running UNIX(tm) real-time RGB video frame grabber interfaces to DEC, ModComp, Rolm, Data General, and a bunch more computers

RDS-2000 Front Panel

Hardware
Early Ikonas lab -   RDS-1000 prototype frame buffer (512 x 512 x 8) in rack, Heathkit LSI-11, 8" floppy disks, ADM-3 terminal, Summagraphics BitPad tablet	Early Ikonas manufacturing - RDS-2000 wire-wrap boards and testing - part-time EE student labor	many more photos to be added some day	--
RDS-3000 DR64B frame buffer board - about 15" x 15" with 128 16K DRAM chips Each board displayed  512x512x8 or 1024x1024x2. Twelve boards were needed for a 1024x1024x24 full color system. Later versions (DR256) had 128 64K DRAM chips	DR64B wire-wrap side. This one is machine-wrapped. Earlier ones were hand-wrapped. Later ones were multi-layer printed circuit.	Dave Sieg's collection of RDS-3000 systems left over from Omnibus	PCP Peripheral Control Panel analog & digital interface with Switch Box and Dial Box

Software Download Ikonas Programming Reference Manual We developed a great deal of microcode, covering 3-D graphics, image processing, seismic data display, flight simulation, ray tracing, and interactive solid modeling. Download PDF of data sheets shown below for FSS, Seismic-3000, Solid-3000 packages
Solid-3000 Microcode Package	This package of microcode routines provided anti-aliased lines, z-buffered polygons, Phong shading, texture mapping, bump mapping, transparency, adaptive subdivision of bi-cubic B-spline surfaces, etc.	Teapot display at the Computer Museum in Boston 1985-89? On the left was a monitor case with Martin Newell's famous teapot, lit by various colored lights. On the right was a display driven by an Ikonas RDS-3000 running our Solid-3000 microcode, with adaptive subdivision B-spline patch rendering. Press the "RENDER" button and the Ikonas would re-render with shading to match the physical lights. - Photo courtesy of Steve Baker & his Utah Teapot site
Seismic-3000 Microcode Package	"Wiggle Trace" display - interactive curves drawn directly in microcode from seismic data	This package of microcode routines provided a variety of display techniques useful for 2D and 3D seismic data set interpretation. Sub-surface display using transparent surfaces	Interactive 3D volume slicing of seismic data. Most of the same functions got included in our medical imaging package.
IDL2 Ikonas Display Language Microcode Package A real-time display language including anti-aliased vectors, 3-D transformation, and visible surface sorting. The entire display program would be compiled in the host computer, get downloaded and then run completely in the Ikonas processor, with the host processor updating transformation parameters, etc.	Real time display for cockpit instrumentation - written by RTI for NASA-LRC	--	--
FSS - Fortran-callable graphics routines for general 3-D graphics

Customers

Early customers got a set of interface drivers, Fortran read/write/display control routines, and for the optional programmable graphics processor, a microcode compiler (IKASM) and 3D graphics routines (IDL). Later, several C compilers and application libraries were available for the programmable graphics processor. Customers were strongly encouraged (!) to develop their own code. They came up with all sorts of interesting and imaginative ways to use the system. Ikonas display systems were used for the "Genesis Planet Sequence" in Star Trek:The Wrath of Khan, for the video games in Superman III and for a bunch of TV commercials, as well as for a lot of serious scientific and engineering work.

Customer List - We put together an Ikonas customer list for the SIGGRAPH retrospective. Please e-mail me with any additions or corrections.	Z-buffered polygon rendering microcode written by our first customer - UT Health Science Center-Dallas	Paul Hauge at Phillips Petroleum	more to come
Tufts-New England Medical Center	AT&T Bell Labs Murray Hill	Ship Analytics	Digicon Inc.

From "Life Before the Chips: Simulating Digital Video Interactive Technology" by Douglas F. Dixon. Communications of the ACM, July 1989, pp 824-831.
"From our first Ikonas system delivered in November 1981, this remarkably powerful and flexible machine was a major part of our success in communicating the vision of DVI, and nurturing it until the chips arrived in December 1987. Some of that flexibility and power also lives on in its influence on the design of the DVI chip set. The more expressive the simulation environment for a product concept, the more ideas can be explored and prototyped for the final design, and the more likely that features of the simulation system become incorporated in the product."

From "Hardware Support for Multitasking Graphics" by William Cowan, Christopher Wein, Marcelli Wein, Kellogg S. Booth. Graphics Interface '91 proceedings, June 1991, pp 199-206.
"Finally, it is interesting to note that the interface construction discussed in this paper and the further enhancements discussed in this last section are easy to carry out in the [Ikonas] RDS-3000, a decade old design. Its flexibility, modularity, and openness make it possible to design and substitute components in a way that cannot be done with newer designs offering higher graphics performance at the cost of closed hardware. Systems like the RDS-3000 may be unsuitable for the production graphics that makes up the vast bulk of the market, but they are the life blood of laboratories that conduct research into new techniques for combining computation and graphics."