Date: 21 May 98 16:21:49 +0000 From: "Paul May" Subject: Dr ED's interview Message-ID: <12233.445T2497T9816303@mcmail.com> Newsgroups: comp.sys.amiga.misc X-Newsreader: THOR 2.5a (Amiga;TCP/IP) NNTP-Posting-Host: 195.44.5.217 X-Trace: 21 May 1998 18:13:50 GMT, 195.44.5.217 for the benifit of the fido peeps: An Interview with Dr. Ed Hepler Amiga Custom Chip-set Designer. By Vidar Langberget, foreword by DarkBlue. The majority of you will know of and remember the promise of the AAA chip-set from Commodore. This "killer" sub-system was to follow on from AGA (AA) and was to be the natural progression from that chip-set. Dave Haynie has often remarked on just how good AAA was going to be... Perhaps it could have even saved Commodore, had they not faltered and produced the ill-fated CDTV, A600 and CD32 systems. However, the inevitable occured and Commodore dropped the ball, leaving us with a chip-set which was already losing its edge. Had Commodore been a little more intuitive, then we might have been running machines which would have still seemed admirable today, despite the big C's demise. Unfortunately that was not to be and AAA has now been relegated to the scrapheap of unsung technologies, along with the likes of the Memotech and Sinclair QL breed of machine. What many of you might be unaware of, is that AAA was actually complete and ready to go to the production stage (see Dave Haynie's "Death Bed Vigil" video for evidence of this). Not only that, but Commodore were actually quite advanced down the development path of AAA's successor, the near-mythical "Hombre" chip-set. Hombre was the brain-child of Dr. Ed Hepler and, in astonishing twist of fate, it seems that Dr. Hepler might be willing to complete the project and may even approach Amiga International with it... Where this leaves the Amiga is anyone's guess, but one thing's for sure - following Amiga news sure is interesting these days! The following text is copyright (©1997) Vidar Langberget. The AGA custom chip-set, firmly entrenched on an A1200 motherboard. Vidar: Could you tell us something about yourself, your education and your employment history, etc.? Dr. Hepler: I have a BS, MS, and PhD, all in Electrical Engineering and all from Drexel University in Philadelphia, PA. I joined Bell Lab's after school and spent about 7 years in their Naperville, Ill. (Indian Hill) Facility. That facility was used to design all of the electronic switching systems for AT&T (before it was split up) and the fault tolerant computers which drove them. I was involved in the design of the 3B20, 3B5 and 3B2 computers. A couple of them became commercial products of AT&T after the split. I was also involved with the specification and design of part of the BellMAC-32, an early 32-bit microprocessor developed at Bell Labs. From Bell Labs, I moved on to General Electric where I spent about 5 years. This was in their SpaceCraft Digital Processor department. GE made communications satellites and we did processors of various types for them... I next moved to Commodore where I worked on the AAA architecture and was responsible for the Andrea chip. Andrea was the Agnes/Alice equivalent for AAA. Andrea was far more sophisticated than Agnes/Alice. It had a microcoded processor (very RISC like) built in to do coordinate arithmetic, etc. Much of the line draw arithmetic, clipping, etc., was done in hardware on Andrea... The last couple of years, I reported to the VP of Engineering and was responsible for the architecture of next generation Amigas. In that role, I performed various studies including one which would have produced a single chip Amiga (Motorola MC680x0 core, plus AA logic), and early versions of Hombre which contained a SIMD processor for graphics, etc. The last study became the Hombre and design was started... [Bloody Hell... Hang on a minute, "single chip Amiga... 680x0... plus AAA logic" - all one chip! Could it be that all those rumours were true and that Commodore were seriously considering a portable AGA Amiga? - Ed] I now have my own company, VLSI Concepts, which does Architecture (System and VLSI), design, and CAD work... Vidar: What is the Hombre, and what was your involvement with it? Dr. Hepler: Hombre was to be the next generation Amiga. I did the architecture, did a set of C-level cycle-correct models to perform performance simulations, then did M models (a Mentor Graphics proprietary hardware description language) to synthesize the control logic. Other M models described the data paths and I had a couple of engineers starting to do transistor level design of them. Hombre was to be a complete system in two chips. The first chip was the CPU chip and contained a RISC integer core, a blitter, a 3D colour rendering engine, audio, a CD interface, a peripheral interface, a bus interface and controllers for display and system memories. The second chip was the video data path and contained line buffers, a colour lookup table, etc. The chip set was designed so that with minimal external chips (ROM, DRAM) a complete low-end system could be built. This low end system would be adequate for a game console like CD-32 (CD-64) or a set top box. With the addition of an ASIC (which connected to the peripheral interface), a low-end computer (like the A1200) was produced. When an external processor was added, the Hombre chip set became a peripheral processor and off-loaded the main processor so it could be more useful doing OS and user tasks. The chip set could also be configured to be "slaves" on a peripheral card and therefore could become a graphics plug-in for any platform... Internally, it had 64-bit datapaths. Externally, it could be configured to use 32-bit memory (in low cost (and lower performance)) configurations, or 64-bit memory in higher performance configurations. Both used the same chip, but bonded into different packages. Vidar: What 3D features did it have? Dr. Hepler: It supported Gourand shading and texture mapping in hardware. It could shade or map trapezoids... Pixel sizes ranged from 4-bit to 24-bits... It did most of the rendering steps in hardware and supported non-linear shading and mapping... One of the software engineers did a simulation of the texture mapping algorithm which wrapped an angel-fish picture around the inside and outside of a rotating cylinder... It was pretty impressive and did the perspective mapping quite nicely... Vidar: What kind of sound system did it have? Dr. Hepler: I planned on taking some of the sound circuitry from AAA to produce basic audio. But I planned on changing the output format to that of digital CD (BCLK, LRCLK, DATA) to use commercially available amplifiers. There were two CD channels in and a method of combining the audio... This allowed external audio chips (DSP style) to be placed on the peripheral interface and be merged with the Hombre generated audio. It also allowed audio from MPEG decoders to be merged with Hombre generated audio... Vidar: Would the Hombre be PAL/NTSC compatible? Dr. Hepler: Yes. We were also monitoring the movement of HDTV and were planning on making it as compatible as possible... Of course, in hindsight, we would have had to make some changes as the HDTV standards still seem to be moving. Vidar: Would features like the copper, scrolling and sprite support be included? Dr. Hepler: Yes and no. A copper like structure was there. Scrolling was supported. I talked to enough people to realize that many game developers were unhappy with the sprite hardware. Most said that 8 sprites were not enough. Most also said that the collision detection hardware was inadequate. No matter how many sprites I would have put in hardware, someone would have wanted at least one more. So I put in no sprite hardware, but put in some blitter modes which made generating scenes with software sprites very easy. There was really only one hardware sprite, for a cursor... Vidar: What resolutions would the Hombre offer? For games? Dr. Hepler: It would go up to 1280 x 1024... Perhaps more interesting was that it supported 4 playfields (at lower resolutions)... And each playfield could be a different style of display... For example, one playfield could be HAM (for a very nice background) while others could use the colour lookup tables... Vidar: How long do you think it would take to finish it? Dr. Hepler: The development schedule at the time was 18 months long for both hardware and software. We were part-way through it when things fell apart. The schedule was planned such that Hombre equipped games-machines (CD-64) would have been available for the Christmas 95 season. Vidar: How do you think the Hombre would compare to the current consoles and 3D cards for the PC, not to mention AAA? Dr. Hepler: I believe that it would have been competitive with any of the game consoles currently available. Some of the PC 3D graphics chips have probably pulled ahead, but they are targeting a different market... Other than the lack of compatibility with ECS and AA, it was superior to AAA. (We had decided to scrap AAA in favour of Hombre.) Vidar: Many people are sceptical about custom chip sets, because they say off-the-shelf stuff is faster, easier to upgrade and cheaper. What are your views on this issue? Dr. Hepler: It depends a lot on what the overall strategy is... You have to remember that all chips are custom chips, whether designed by "a big company" or a little company. I think that a lot of the scepticism is due to a lack of knowledge of how chips are designed. With the new synthesizers, etc., it is not nearly as difficult as it was even 5 years ago... Since I make my living doing these kinds of designs... I'm all for them :-) Vidar: Is it possible to upgrade the Hombre, without designing a completely new chip? Dr. Hepler: I had a specific set of products in mind for the Hombre. I would have to discuss product strategy before answering this sort of question... Vidar: It seems like the rest of the computer world are moving towards programmable VLIW chips etc. What do you think about these, and would it be possible to integrate a chip like this into the Hombre to boost performance? Dr. Hepler: Hombre had an integer PA-RISC core on board to act as the system processor in the low-end mode or as a peripheral processor in the high-end mode. I added a few instructions to it to help in graphics and 3D processing, much like the MMX instructions that Intel has added to the Pentium. Dr. Hepler: The VLIW work (like that of the MultiFlow) that I am aware of is still well off into the future at the chip level. I think that it would be premature to try to put something like that into Hombre. It probably would not fit into the cost window... Vidar: What was the target price for the Hombre? Dr. Hepler: I was targetting $45 (USD) for the two chips (total). This would have been our cost to produce... Vidar: Would the PCI bus' bandwidth be problem if it was put on a PCI card? Dr. Hepler: No. The reason for this is that Hombre had its own processor and its own memory. The only thing passing through the PCI interface would have been commands or source data. All graphics fetches would have been from local memory. Most rendering would have also been from local memory. Vidar: If Gateway 2000 asked you to finish the Hombre and provided enough resources, would you be interested? Dr. Hepler: I have tried to contact them, but have been unsuccessful in getting to talk with anyone. I now have my own company and have done some multi-media design work that they may be interested in licensing. I would also be interested in doing some contract work in system or VLSI architecture... Vidar: Thank you very much for taking the time to answer these questions! -- Paul May Team *AMIGA* User Of The small AMIGA OS With A BIG Future -- Clean Up Your OS , Not Your HardDrive --