A Look Inside Bitboys October 28, 2002 Dave Barron

Introduction

Page:: ( 1 / 6 )

Bitboys has long been a mystery within the graphics industry. Many consider them a joke, few are loyal followers. Yet behind all of this a real company exists. They are, without question, a company of truly dedicated people who strongly desire to do something good. How do I know this? I used to work for them. For nearly a year I was employed at their Dallas office, until shortly before it was ultimately closed down. This experience has provided me with a sufficient amount of insight in the company and its history to relate the real story behind Bitboys.

Bitboys history goes back to the days of a group known as Future Crew. Future Crew made a name for themselves in the 3D technology demo development, notorious for pushing computer graphics capabilities to the max. From this group, a number of small companies developed, and amongst them Bitboys was founded in 1991. Much would happen over their lifetime, bringing them from the original two founders to what they are today.

The Early Days

Bitboys’ first few years of existence were comprised of traditional software development for local companies, sustaining themselves for growth. This proved extremely boring for them, to say the least, but doing so provided the necessary funds for company growth. During this time, Bitboys continued work in the background on their true love, 3D graphics.

During 1993 Future Crew developed Second Reality, a 3D technology demo. This demo brought them to the realization that existing graphics hardware could not deliver the necessary functionality and performance. For example, one issue they saw was the lack of functionality in hardware’s blitting and line drawing. Thus, Bitboys set out their own development of a graphics processor.

During ’93 and ’94, Bitboys began extensive research into 3D hardware. By early 1995, they had teamed up with Finnish silicon company VLSI Solutions to develop a complete graphics processor. VLSI, in turn, introduced Bitboys to Tritech. Thus, VLSI and Bitboys worked together in developing TR25201 for Tritech. This brought up successive chips, TR25202, TR25203 and TR25204.

TR25201 Specifications

50 Megapixels/sec
32-bit color
24-bit Z + 1-bit stencil
32-bit SD/SG/EDO DRAM
800 MB/sec Memory Bandwidth
Bilinear/Trilinear Filtering
Loopback for single pass multi-texturing
1 Million Vertex/sec T&L Engine
Hardware Bump Mapping
200 MHz RAMDAC


SIDEBAR: At their peak, Bitboys had between 40 and 50 employes and contractors.

Pyramid 3D Probilems

Page:: ( 2 / 6 )

Problems Arise

Initial silicon spins of TR25201 were riddled with bugs, causing a highly distorted image. Timing issues resulted in multiple silicon respins, in an attempt to resolve these problems. When the final version of the chip was developed, TR25204, the issues were solved and a few additional features were included, including a built-in VGA core. Yet it was at this point the real problems developed.

While Pyramid3D was being developed, Tritech was stuck in the middle of ongoing litigation, having been charged with violating an audio patent. With the suitor demanding back royalties, an eventual loss in the case proved fatal for Tritech. Their doors were closed and Pyramid 3D never saw the release that they had worked for years towards.

With this major occurrence, Bitboys again found themselves at square one. They no longer had a chip to release, nor a partner to design it with.

Looking Towards the Future

Late in the development of Pyramid3D, Bitboys had begun working with Tritech on a next generation architecture that would make heavy use of embedded memory. With Tritech’s closure, their partner was gone, yet the concept certainly was not. Thus, on their own, they began the development of their Xtreme Bandwidth Architecture (XBA), and their next-generation chip, Glaze3D.

During this time, Bitboys was actively searching for a partner for synthesis and back-end, contacting the likes of Real3D, Rendition, Creative Labs, ATI, NVIDIA and Diamond Multimedia. Unfortunately, as is often the case, they found the “not created here” attitude that is all too common in the graphics industry. The result of this was a rethinking of Bitboys future partnership and design plans.

Originally, Glaze3D was designed as a high-end chip with 9 MB of embedded DRAM on a 512-bit internal bus. Running at 150 MHz, the original Glaze3D would have provided a fill-rate of 600 Mpixels/sec with dual texturing. Having developed environmental bump mapping, this was an included feature, as was the support for multi-chip solutions.

At this time Bitboys closed its first round of financing and obtained Infineon as its manufacturing partner, with its embedded DRAM technology.



SIDEBAR: Finnish partnerships run much deeper than those typically found in the US. When two companies become partners, they function very much as a single unit, which can often make it difficult to remember who works for who.

Glaze3D

Page:: ( 3 / 6 )

Work begins on Glaze3D

With few people employed at Bitboys and a relative level of inexperience in chip development, Glaze3D development took considerable time. Being Bitboys’ first complete independent design, it was far from an easy task. While certain aspects were straightforward, including the pixel pipeline and triangle setup, other details proved to be rather complicated. There were the more intricate parts of design, not specifically related to functionality, but general chip operation, and also the more complicated units such as the memory manager.

With a level of foresight, Bitboys decided it would be important to include a geometry processor. This chip, then known as Thor, would function as both a T&L unit and as a bridge for multi-chip boards. Yet Thor would never make its way off the drawing board, for resources were focused on Glaze3D.

Glaze3D development continued for a few years, with the majority of the company focusing on the design, Kaj Tuomi converted C code into VHDL, preparing the chip for synthesis. Certainly, it was a drawn out process with Bitboys limited design resources.

New Tools Bring Changes

Glaze3D synthesis work was done in Infineon’s design center in France. This effort sparked an idea in their minds for a tool set designed to automate much of the product design flow. These tools would put in place all of the necessary features for allowing a C programmer to complete unit designs. The programmer would code the unit in C, then using the built-in tools and automated functionality, take the unit to VHDL and if desired, synthesis. Clock accurate simulation capabilities would be included as well, allowing for testing and debugging. Upon returning to Finland, design work began on this tool set. This would later prove to be one of Bitboys wisest decisions.

Bitboys, being forced into a situation demanding the development of a complete design, required a rethinking of their design strategy. Certain aspects of the original Glaze3D architecture were becoming out of date, with performance and a feature set that were not up to par with where the industry was headed. Would they start from scratch or rough it out? Many points required careful consideration.





SIDEBAR: When the Bitboys' Dallas office was open, it was comprised entirely of former 3dfx employees.

Axe	Page:: ( 4 / 6 )

From Glaze3D to Axe

The new tool set was making steady progress after several months; a workable version was ready just before the end of 2000. With these tools expected to reduce design time by many months and Glaze3D requiring several improvements to make it a marketable part, the chip went partially back to the drawing board. While much of the core architecture remained the same, Axe was created with a tweaking and recoding on the new design system. The total amount of EDRAM was increased from 9 to 12MB, the internal memory bus was expanded to 1024-bits, and Matrix Anti-Aliasing and a programmable vertex shader were added as well. While making a fresh start, the Glaze architecture had provided the necessary insight to allow Axe development to move forward quickly.

Axe Specifications

600-800 Mpixlels/sec
Single Pass Quad Texturing
Single Cycle Dual Texturing
MatrixAA
12 MB EDRAM on 1024-bit Bus
DirectX8 1.1 Vertex Shader
50 Million Vertices/sec Sustained
Dual Chip Support over CBI (Custom Bus Interface)

[image]

Axe development

Bitboys set an extremely aggressive schedule for Axe, anticipating tape out in less than 6 months. The original schedule called for the release of Axe during the 2001 Christmas seasons under the name Avalanche3D. With the new tools in place, this appeared to be a realistic timeframe. This, however, would be the first time that Bitboys would take a chip design to final silicon, so there were still a few important things to learn.

Axe, as with every chip, had bugs early on in development. These bugs were rooted out of course, but only with time. The majority of the bugs were found relatively quickly, though some remained for several months, as additional little issues were discovered. In calculating the original schedule, Bitboys had not set aside a sufficient amount of time for verification. Testing and debugging required several months more than had been originally anticipated.

During the summer and fall of 2001, place and route was occurring by a partner company. Initially, early versions of the netlist were used for this, with newer verses as they became available. This place and route took considerable time, with the end result requiring considerably more time than expected. With the end of 2001 approaching, more bad news was on the way.

More problems

Page:: ( 5 / 6 )

Problems at Infineon

While Axe development was taking place, Infineon, Bitboys primary partner and fab, was continuing to suffer financially. With the low level of DRAM prices, it was impossible for them to stay profitable. After multiple quarters of repeated losses, the decision was made to close the embedded DRAM business unit.

This decision had a major impact on Bitboys, as this business unit was to produce Axe. Without any other fab capable of producing the part, Bitboys was stuck in a corner with no way out. Axe would no longer be able to hit the retail market. A decision was made to produce a limited supply of Axe parts, allowing Bitboys to show off the product to venture capitalists as a technology demonstration.

Hammering Away

With Axe taping out near the end of 2001, work on Hammer quickly became the priority. Hammer was to be a high-end DX9 part, fully supporting the DirectX9 specification. Considerable work was done on Hammer, including securing a new fab partner, laying out the complete product specification and even a substantial amount of coding. Proving to be extremely promising, Hammer offered a variety of innovations in anti-aliasing, occlusion culling and memory management. But still, more troubles were to come.

Hammer Specifications

1200+ Mpixels/sec
Single Cycle Quad Texturing
DX9 Pixel Shader
3 MB EDRAM
Proprietary Occlusion Culling System
MatrixAA2
DX9 Vertex Shader
100 Million Vertices/sec Sustained





By the end of 2001, it had become painfully clear that money was getting short. Several layoffs took place, with work moving towards securing additional venture capital. By the end of February, it was apparent that this would take greater time than had originally been anticipated, so additional layoffs took place. I was included in this round of layoffs, unfortunately. Yet, still more were to come, with the eventual closing of the Dallas office.

Not wanting to place the entirety of their hopes on venture capitalists, Bitboys had been simultaneously pursuing other forms of funding. A set of tool development companies had been contacted in hopes of licensing the tool set. Certain companies expressed a considerable amount of interest, although a final sale never took place.

Mobile chipset developers and PDA manufacturers were also contacted in hopes of using certain Bitboys technologies in that market. Early interest was promising, as an impressive technology demonstration was presented on PDAs. In time, this pursuit would be seen as a very wise move on Bitboys part, though the immediate future would yield little.

A New Direction

Page:: ( 6 / 6 )

Near the time of ending US operations, Bitboys had begun work with a major electronics corporation on a mobile solution for PDAs and cell phones. A primary design requirement would be low power consumption, as well as minimal die size. For a company that had always been allowed considerable freedom in design work, the restrictions were to be quite heavy. But a challenge can certainly make work interesting, and this is exactly what happened here.

Design work was soon begun though no contracts had yet been signed. With a relatively small design effort, a rough design was quickly completed. The tool set allowed them to export the C code directly to a FPGA, by means of conversion. With a few available demos, the architecture was demonstrated on the FPGA, demonstrating a surprising level of performance though not using actual silicon.

[image]

Conclusion

And that is Bitboys. Love them or hate them, they are for real and they are doing something. While not focusing on high-end 3D graphics any longer, they still hope to make a valuable contribution to the industry by entering the mobile market. Whether they will ever return to the high-end market remains to be seen, but it is certainly something that the company founders one day hope to do.

The future success of Bitboys is more likely than ever. With a contract in the bag, hopes are high. As with any company, stumbling blocks can come up and companies falter. Whether this will be the case with Bitboys or not remains to be seen. It is certainly something that only time will tell.


SIDEBAR: Would Axe have had enough to compete with ATI and NVIDIA’s best had it made it to market or would it have not made a difference? Voice your thoughts in the comments!