Why is the new Sega® [1] game console doomed from its beginnings? To understand this, one must understand some of how the video game industry operates, and how computer hardware and video games are interrelated.
Sega's first offering of note was their Genesis® [2] game console. It featured a 16-bit processor, custom graphics hardware, some effective multimode hand controllers, and cartridge (ROM) based games. Cross-licensing was abundant, games were many, and the unit did extremely well.
There were add-ons available for the [3] Genesis system. One was a CD drive, allowing CD-based games. It never did well in the marketplace, as it was expensive, the games were expensive, and a CD can't compare with a ROM for speed-of-access. Another accessory was an add-on CPU module offering 32-bit games. However, not very many of these were offered or sold, and this hardware and software, facing stiff competition with the “true” 32-bit systems, soon ended up on the remainder bins.
Sega wasn't about to give up, however. They designed the Saturn® [4], an all-new system with a fully-integrated 32-bit CPU, new graphics hardware, a CD-ROM drive allowing cheaply produced games, and enough memory to buffer the CD-ROM's still-limited transfer speed. Technically, it was quite a nifty system. However, the competing systems were being marketed heavily, and with extremely attractive agreements with third-party game manufacturers, had more titles too (this being a huge market force). Sega neglected to pass on the cost savings of the CD-based games, and the new venture was quickly perceived as a failure, which, in the video game industry, is a self-fulfilling prophecy.
The third wave of video game consoles is out now, with extremely cost-effective consoles (they are sold at a loss, in order to quickly amass an installed base sufficient to attract the game manufacturers, who ultimately decide the market worthiness of any new system). Sega, never to give up, is preparing, with much marketing fanfare, to release their new third-generation system. These guys are smart, they know their marketing, hopes were high.
However, in some sordid backroom, a huge company with massive monetary resources, in what was apparently a poorly-thought-out marketing ploy, induced Sega to design in a product that was not only inappropriate to video games, but is a stripped-down derivative of one of the most unreliable pieces of software ever to be mass-marketed.
The dark truth is that the new Sega Dreamcast® [5] is running microsoft windows CE under the hood [6]. Why is this bad? First and foremost, video games are among the most demanding pieces of software in existence. Absolute reliability is required — do you think an avid gamer will deal well with having a game freeze unexpectedly just as they're about to pull off a major win? Think about the video games you've played. Do they crash? Slow down randomly? Lose data? Of course not. Such bad behaviour simply would not be tolerated. Such games are instantly panned by video game reviewers, orders are not made, and they are quickly shunned by all and sundry. Microsoft software, however, is known far and wide for its quirks, unreliability, and general user hostility [7]. This is accepted in the cost-is-everything peecee world, but is simply not an option for a video game system. Even microsoft has finally admitted publically that their software is not server-class. And windows CE was never intended for high-demand activities, it was designed to operate PDAs!
Worse, this software was written specifically for the intel CPU family, a group of processors that have stolidly inherited all the poor design decisions of their ancestors, and therefore force some downright strange designs on the software that runs on them. However, the Dreamcast doesn't have an intel CPU. No, the designers inexplicably chose a Harvard architecture [8], CPU, the Hitachi SH-4 [9], as the main engine. Granted, this is a good thing, the intel CPUs are abysmal as far as real-time performance goes (An 8-bit, 2MHz Z-80 can utterly trounce a 100MHz 80486 for video gaming). For video game programming, which is sort of stream-of-consciousness software, the natural CPUs are those with a von Neumann [10], architecture, where the instructions and data are stored together, and accessed linearly. The Harvard CPUs have a different internal philosophy, however. Their world is divided distinctly into two halves, “instructions” and “data”, and the two are carefully segregated. For running the same instructions in a tight loop on an assortment of unrelated data (i.e. DSP applications), this arrangement is appropriate and efficient. For video games, it is rather nonsensical. The only advantage I can imagine is that such CPUs often have separate buses for “instructions” and “data”, allowing more effective bus bandwidth by essentially having two buses. However, arranging storage and layout for such designs is high art, and currently is not done well automatically. Either Sega has decreed a fixed layout that all games must use (wasting either instruction or data storage or bandwidth in almost every case), or they leave the task to the game programmers (who don't want the additional effort, and adds non-trivial complexity to the hardware to allow variable splits; another point of failure).
Another requirement for video games is real-time response. When enemy ships are flying at you en masse, at high speed, it is crucial for playability for the fire button to react effectively instantaneously. All the time. Every time. Again, if this requirement isn't met, the game becomes unpopular, unplayable, and won't sell many copies.
If a single game system becomes associated with more than a few games that aren't near-perfect, console sales will quickly dwindle, game manufacturers won't expend the effort to support another console, and the whole venture will perish, leaving behind a few owners with orphaned consoles. It may seem like hyperbole to require that game systems and software be “near-perfect”, but the simple fact is that video game players are an extremely demanding lot, and are accustomed to high-performance, robust, interesting games. Ironically, video game software is among the most well-written, reliable software currently being produced.
Sega is attempting to improve their position by pre-selling consoles before they're available, giving themselves and the third-party game manufacturers an effectively guaranteed market. This is an extremely good idea, especially if early sales are impacted by reports of unreliable or unplayable games. There will also be some sales to diehard Sega fans [11].
If you are among the unfortunate folks that pre-bought a console, oops. There is no requirement that you actually follow through — then your losses are only the pre-pay amount (generally US$10). You can always console (excuse the pun) yourself with an Atari Jaguar [12], commonly and cheaply available, along with a few games. This system is a guaranteed orphan, Atari is gone, no more games are ever being produced. But you know this in advance, and it won't set you back much. And you can buy that system today!
The final nail in this sad tale is that once a system is fielded, it's effectively cast in stone. No major changes to the hardware or even system software can be made, for risk that existing games won't work any more. Since the games are being written by a variety of third parties, it's impossible to ensure that they don't depend on certain configurations. So, if the initial offering is broken in some way, all the succeeding systems will have to be broken in exactly the same way.
It's a tough row to hoe, but the video game industry tempers the appeal of huge amounts of revenue overnight with the absolute requirement that even early shipments of hardware and software, from disparate manufacturers, work right, the first time, and every time. So, sadly, this disaster will probably spell the end of the Sega saga (sorry). What a pity. Another worthwhile company, with good products, good people, and a potential bright future, will be destroyed by microsoft's utter incompetance and complete disregard for anybody else.