Christmas day. Torn paper encircles the newly-unwrapped CD-ROM game that should have kept the kids occupied for months. Except it doesn’t because the game is too advanced for the computer in the study. This is what happened when Disney launched its Lion King disk in 1994 and since then little has been done to ensure that every multimedia application runs on every computer. Much has been talked about, but there are still few concrete rules to fall back on. Wiser software houses develop for the lowest common denominator but with demand high among generations that grew up watching color television, there is tremendous pressure to incorporate the latest multimedia technology as soon as possible. More than half of the Lion King disks were returned last year.

Dominating

In response to this, and many other problems of getting multimedia equipment to work in harmony, Intel Corp, the world’s dominating processor designer, has been drafting the Native Signal Processing program, making the first results public in March. Both Intel and Microsoft know the current state of the multimedia personal computer is not acceptable. The first shortcoming was found in the audio subsystem. Beeping when it comes to an error is no problem for a computer, but when games developers tried to use audio creatively they quickly found bottlenecks. For a start, there are no genuine real-time services in an MS-DOS and Windows environment – no way of guaranteeing that a piece of sound will arrive at the speakers on time. The best that software developers could hope for when generating complex sounds was that there would be enough spare capacity to meet their demands. Secondly, to get sound into and out of the computer, games authors had to rely on a variety of sound boards and, as there were no fixed ports within Windows, this usually meant by-passing Windows and creating sound drivers under MS-DOS, often meaning that users had to exit Windows before being able to use them. As developers played with tricks and short-cuts, there grew a tangle of different applications trying to muscle in on the sound board. Vendors moaned about their overburdened customer support desks and high return rates. There are two issues in developing for the multimedia system: the interfaces between the software and the multimedia hardware (such as sound boards or modems), and the basic computer on which they run. Sending a ‘squelch’ noise to the speakers every time an alien is killed involves decompressing a digital sound file, converting it to an analog signal suitable for speakers, and pumping it out the back of the computer. This is usually done on an audio codec coder-decoder that has been specially made to handle high quality audio.

By Morgan Holt

To give the software – in this example, a game – access to the codec, the operating system must be modified so that it understands how to play a noise. This is done with small pieces of software called drivers, which, when told what to do by the application, tell the hardware to do it. When an application tells a sound board to ‘squelch’ it is important to get a squelch every from every application, on every computer. The Intel Architecture Labs was set up at the beginning of the 1990s to steer applications written for the iAPX-86 system in the direction of unification. Its intention was to provide a base-line multimedia standard that would ensure every iAPX-86-based computer had some sort of multimedia functionality – sound, video and communications – that could be accessed by the software writers. The result: Native Signal Processing. Native Signal Processing is the general term used to describe the manipulation and processing of signal processing functions associated with real-time media and communications data on the Pentium-derived microprocessor. Native Signal Processing enables the Pentium to process audio, video and communications data using software only, rather than off-loading the data to fixed-function dedicated signal processing hardware. Some functions, such as the data modem, are already making their way into software and Intel foresees a simple telephone plug on the back of personal computers with most of the modem functions being done in software and a cheap commodity modem chip on the motherboard. Power-hungry applications, such as computing video graphics in real-time, will divide their load between Native Signal Processing and whatever hardware is available. Intensive users of multimedia will find the Native Signal Processing base-line inadequate, and so go for the hardware accelerators, but if you buy the Toy Story CD-ROM, at least you can guarantee it will work. That’s the theory. Intel wanted to pull all things multimedia into a single application programming interface that would provide for all the real-time signal processing.

Entry-level

Native Signal Processing was to become part software, part hardware, and in the latest, faster Pentium derivative, the P55c, some of the Native Signal Processing functions were built into the chip itself. In entry-level machines, the technology will supply all the multimedia functions while power users would supplement it with add-in hardware. To achieve this, Intel had a fair scope to experiment. With the P55c, Intel had the opportunity to rebuild much of its hardware to incorporate the Native Signal Processing strategy. While the hardware has been enhanced with some signal processing-like instructions to speed up multimedia functions, Intel made the controversial move of developing most of the Native Signal Processing multimedia functi ons as software only instead of hardware. Intel hoped this would mean a drastic drop in the cost of multimedia: updating software as technology advances is much cheaper than replacing the whole board, and reducing the hardware in the box cuts the price of the whole system. This is desperately needed. Margins for the personal computer have steadily fallen over the past seven years