NostalTech: The Gravis Ultrasound and the PC Sound Before Sound Blaster Ruled

In the early 1990s, "PC sound" usually meant a small FM chip squeezing approximations of instruments out of a handful of operators. AdLib had set the template, and Sound Blaster added digital audio on top of it. Then a Canadian company called Advanced Gravis shipped a card that took a different route entirely. The Gravis Ultrasound, almost always shortened to GUS, played sampled sounds and wavetable instruments rather than relying on a programmable FM generator like its competitors, and it arrived on PC compatibles in 1992 ^¹.

That distinction sounds technical, but it was audible. The GUS carried samples of real instruments and could therefore offer more than sound cards at similar prices, according to the French-language Wikipedia entry on the card. For anyone who had spent years listening to FM renditions of guitars and drums, the difference was less an upgrade than a change of category.

The card found its first true home in the demoscene. Trackers — the software that had grown up around the Amiga's Paula sound chip — could finally be used to their full extent on a PC, and the GUS gave that community the sampling capability they had envied on the Amiga since 1985, along with timbres that recalled the PPG Wave, a professional synthesizer from 1981. The GUS was also unusual in letting input and output use different interrupts and in offering signal processing for effects such as echo. Those were not bullet points aimed at office buyers; they were features for people building sound itself.

What the GUS did not have was the thing that decided the DOS era: compatibility. A former driver developer recalled that a great many games wrote directly to Sound Blaster hardware and did nothing at all for the GUS, so the card's supporters did an enormous amount of work to support as many titles as they could ^². When the card line became the AMD Interwave, that work turned into a routine: the team would buy PC Gamer, go straight to its top 100 games list, purchase every new entry, and have a month to add support before the next issue reset the cycle. That is a remarkable picture of what "driver support" cost in the days before a common audio API did the translation for you.

The same account describes the practical friction of DOS gaming on a GUS. Players had to hack in driver support for both MIDI and digital sound, run a utility to inspect a game's MIDI and find every program change so the right instruments could be preloaded into RAM, and live with the card's full 1MB of memory as a real constraint on MIDI. The former developer's verdict was that the GUS was "before their time" and would likely have fared better in the Windows 95 era, when much of this would have been less of a problem. That is one person's assessment, not a measured result, but it fits the shape of the story.

The GUS also earned goodwill in a way that anticipated later industry habits. Gravis published complete hardware interface specifications for its cards well before open source became a buzzword, and at least one developer used a GUS as a first serious driver project for Windows NT ^³. Opening the hardware did not by itself win the market, but it explains why the card inspired loyalty out of proportion to its sales.

On sound quality, opinions in the community were not uniform. One long-running forum assessment ranked the GUS roughly third among consumer ISA sound cards, behind the Pro Audio Spectrum 16 and well behind Turtle Beach's Pinnacle and Fiji cards. Set against that, a GUS owner described the cards' output as perfectly clean, with no noise from the PC, while still calling them before their time. Both can be true: a card can measure mid-pack in one listener's ranking and still be prized for a quiet signal.

The upgrade path tells its own story about the era's compromises. The same owner bought a GUS 256M and returned it in under a month in favor of the GUS MAX. The MAX is remembered fondly, but the wish list for DOS was revealing: a DAC that was Sound Blaster Pro compatible and an OPL2 or OPL3 chip onboard would have been ideal, though that would probably have made an already expensive card even less competitive. In other words, the best way to sell a superior sound architecture was to bolt on emulation of the inferior standard everyone already owned.

What remains unconfirmed is the fine detail: the sources here do not give launch prices, sales figures, or a full model timeline, and this article does not invent them. What they do establish is the shape of the GUS's life — a card that pushed PC audio toward samples and wavetables, that was embraced by trackers and demoscene musicians, that demanded extraordinary manual effort to support games, and that lost the mass market to a compatibility standard rather than to a lack of ambition.

For a modern reader, the lesson is not nostalgia for a particular ISA card. It is that the winning platform is rarely the most capable one; it is the one software already targets. The GUS is also a reminder of how much invisible labor — buying games by the hundred, mapping program changes, preloading instruments — sits behind the phrase "it just works." And it is an early example of a hardware maker deciding that publishing complete interface documentation was worth more than secrecy. The Gravis Ultrasound did not become the default, but for a stretch of the 1990s it was the sound that people who cared most about sound wanted to hear.

Sources

1. ¹ Gravis Ultrasound — Wikipédia - Gravis Ultrasound hugo.lv
2. ² Highly reviewed speaker can be hacked over the air to infect connected devices arstechnica.com
3. ³ Anyone remember the Gravis Ultrasound? arstechnica.com