GSI and the CUBI 7  EPIC  STORY
THE 3D COMPUTER GRAPHICS HISTORY IN FRANCE
1970 - 1990

FrançaisFrancais

THE GSI MACHINE : "GENERATEUR DE SYNTHESE D'IMAGES" 1979: (Synthetic Image Generator) 

First  machine in the world le to display 400 shaded, colored polygnas in real time.

The GSI machine was based on a depth priority algorithm. with complex space partitionning rules. Each part of space was separated by solid angles and planes, where  polygon priority was computed. I managed the hardware parts design, which was developped by Inforel company, by years 1979. The software part was developped under my direction by a Supelec resercher. I wrote myself the microcode assembly which was running on an AMD 2910, 128 bits slice computer. I wrote this software in APL language, a very powerfull mathematical software interpretor.
I used also later APL for the design of  a modelling and animation authoring tool, running on VAX machines.

GSI was dedicated to flight training simulators. I presented my system in PHOENIX (Arizona) by year 1980 at "IMAGE II conference". At that time, flight simulators was a very promising research topic, for civil spatial and military flight training. GSI was a great success : Boeing, Général Electric and Lookhed proposed immediately several jobs to me in the USA !

1979: GSI: 400 colored and Gouraud shaded polygons in real time !  (Henry Gouraud has same diploma as me : ECOLE CENTRALE of PARIS)                 
About  400000 Pixels/s! Here is a Russian Suckoy 28. A very interesting aircraft  for french military people at that time !
Le GSI (Generateur Synthetique d'Images) with author dseigner at work ! To the left, one can watch 2 GSI hardware machines, containing the 128 bits slice processor. To the right, a CROMEMCOM Z80 computer, with a joystick. I received lots of visitors, including french companies, Délégué général de l'Armement, and even the French Minister of Defence, Mr Jean-Pierre Chevènement.

After the GSI American success, I received job opportunities from FRANCE TELECOM                
                                    

I designed after main schemes of the GSI 10000, a 10000 polygon per frame Z-BUFFER machine
AWARD: Royalties from the french Defense Nationale.  
Until 1979, as I understood algorithmic limitations of priority-based algorithms for hidden surfca removal, I designed a more powerfull, real time machine, based on the Z-BUFFER algorithm, described by Evans & Sutherland (1971), but never really implemented in hardware before). At that time, it was the first machine entirely based and wired with an HARDWARE based Z-BUFFER. Even US Companies, such as GENERAL ELECTRIC were using very limited priority machines. Very large (more than 6 electronic racks). In France, THOMSON simulator company in TRAPPES and after CERGY and TOULOUSE was using reduced models and video camcorders. Their priority-based machines (VISA models) were very expensive and they had many difficulties to design their data bases (Airports for landing training, landscapes, and so on).SIEMENS also with David Niles tried to develop in Paris a priority based machine, but without success. These facts reinforced my will and confidence in the Z-BUFFER algorithm. At that time, it was a promising research & developpement approach. It was high time for me to design a real time 3D machine based on the Z-BUFFER algorithm. 

Following my example, the best INFOREL employees, teached by me, were employed by another company, called SOGITEC, which caused INFOREL bankruptcy. They developped a Z-BUFFER machine called GI-10000 by copying my own designs of my Z-BUFFER machine schemas, but they stayed exclusively simulation-oriented. This market domain was at that time the only one which was able to support the huge necessary  financial R&D costs development.e

So at that time, I understood that 3D workstations (a brand new idea !) would be needed not only for flight simulator companies, but also for audio-visual, CAD-CAM, medical and architectural & urbanism companies.

But SOGITEC continued his development only in the field of flight simulators, neglecting other markets they completely ununderstood. Finally they were absorbed by DASSAULT SYSTEMS. They were very proud to sell 2 GI10000 machines in the US, which were immediatly copied by US companies, and finally they made also bankruptcy.

On the contrary, I was recruited by year 1981as 3D manager by FRANCE TELECOM, and developped a multipurpose, 3D workstation I called CUBI 7 largely before the SGI workstations (Silicon Graphics); 

The GSI first 3D computer graphics machine was a great succes, from the part of FRANCE TELECOM and his subsidiary TDF(telediffusion of France) which offered job to me by year 1981. I was Laboratory Manager and I designed with my team a new, real time Z-BUFFER based computer graphics machine.I called CUBI 7

1980-1985: FIRST  worldwide MACHINE based on  Z-BUFFER algorithm (based on depth - tests) :

Applications médicalesMinitel FRANCE TELECOM 1985 en 3D
The CUBI 7 LOGO which means : CCET as
"Centre Commun d'études de Télédiffusion et Télécommunications
Courtesy: CCETT/TPP Laboratory
Medical applications
Courtesy : GBM SIM biomedical group which used TPP/ CUBI 7 hardware and software facilities.
Courtesy GRAVI, an Audio-visual advertisement 3D startup Company, a 3D startup Company. (no monger in activity now)



CUBI 7 was a 3D real time Z-BUFFER machine
 3D en temps réel , la première machine possédant un tampon de mémoire de profondeur, (technique du Z-BUFFER) associée à une mémoire 24 bits RVB in 1981.
I designed this machine with a researcher team in 1981 for FRANCE  TELECOM R&D, largely before  Silicon Graphics, wat that time less performing than CUBI 7.
CUBI 7 was at the beginning workstation oriented, associated to a VAX (from DIGITAL EQUIPMENT) or an SM90computer, developped by TELMAT company, under FRANCE TELECOM licence.  It's main applicative targets wer CAD/CAM, architecture,  médical applications, and also simulation. 

Each pixel was anti-aliased, shaded with Gouraud shading algorithm, and the wired  Z-buffer depth test was computed in 50nS delay only by year 1984!
For a  512x512 pixels screen, display power was about 20 Millions pixels par second!
that is to say: 200000 polygons per second!

The first host  machine hôte was a VAX 750 from DIGITAL EQUIPMENTORPORATION. (Already in FRANCE, the French national state-owned company was unable to provide reliable computer, in spite of milliards of money given by the socialist French government.
After, CUBU7 received a SM90 from TELMAT, puis a une machine SUN.

This machine had  Z-keying capabilities : by year 1984, it was a great performance ! For exemple, it was possible to insert a character in a synthetic 3D scene in real time.

More than 20 CUBI 7 machines  have been built by TELMAT company in EUROPE:
For Research Centers & Industry :
Société GRAVI, CITROEN, Ecole Polytechnique, Ecole Centrale, Laboratoire Central des Ponts et Chaussées in Paris,
Ecole Nationale Supérieure des Télécoms de Bretagne, University of Compiègne,  SFP (Société Française de Production) in Paris, Michel François de Castelmur Productions,
and for companies involved in audio-visual production & animated cartoons .

IMAGE RECHERCH AWARD in FRANCE by year 1985 given by MINIS et MICROS computer newspaper and the Informatic Agency in Paris.
 
The little story of GSI and CUBI 7: Of course, and as usual in France, other socialist researchers and academic people, little administrative (socialist fonctionnaires) people were jealous against CUBI 7 and my technical skill. In spite of my efforts, they achieved to stop definitively my laboratory by year 1990. Then I focused my research areas on image compression for digital still or animated pictures. I was an MPEG4 contributor, in Paris and Hannover. I wrote several software programs and patents on this subject, in cooperation with Professor Burnod from the University of Paris. Five years after, France Telecom administrative functionaire people discovered the  VRML standard (Virtual Reality Modelling Language) for Internet network. They were very surprised that 3D could be used on the INTERNET !

It was too late, unfortunetely, at that time for CUBI7 which was previously abandonned, but I  achieved launching a new laboratory involded in "New 3D services on Networks". I designed then a 3D viewer based on OPEN GL, running on PC's.  With the help of prominent architects such as Professor Louis Mariani in Paris, we digitized several outstanding patrimonial castles such as "Le petit TRIANON" in Versailles, and the VILLANDRY castle near TOURS.

From 1980 up to 1987, several European projects were launched involving TELMAT/CAPTION  Company, my FRANCE TELECOM Laboratory, The University od Thübingen (Professor Wolfgang Strasser and his laboratory) and CAP SOGETI. The main result was a 3D Computer Graphics board (the SPIRIT project), which was the first Z-BUFFER based PCI board for PC.