Hubbry Logo
Gravis UltraSoundGravis UltraSoundMain
Open search
Gravis UltraSound
Community hub
Gravis UltraSound
logo
8 pages, 0 posts
0 subscribers
Be the first to start a discussion here.
Be the first to start a discussion here.
Gravis UltraSound
Gravis UltraSound
from Wikipedia
The GF1 chip

The Gravis UltraSound or GUS is a sound card for the IBM PC compatible system platform, made by Canada-based Advanced Gravis Computer Technology Ltd. It was very popular in the demoscene during the 1990s.[citation needed]

The Gravis UltraSound was notable at the time of its 1992 launch for providing the IBM PC platform with sample-based music synthesis technology (marketed as "wavetable"), that is the ability to use real-world sound recordings rather than artificial computer-generated waveforms as the basis of a musical instrument. Samples of pianos or trumpets, for example, sound more like their real respective instruments.[citation needed] With up to 32 hardware audio channels, the GUS was notable[citation needed] for MIDI playback quality with a large set of instrument patches that could be stored in its own RAM.

The cards were all manufactured on red PCBs, similar to fellow Canadian company ATI. They were only a little more expensive than Creative cards, undercutting many equivalent professional cards aimed at musicians by a huge margin.

Versions

[edit]
Gravis UltraSound (Classic)
16-bit recording daughterboard

UltraSound (Classic)

[edit]

The first UltraSound was released in early October 1992, along with the Gravis PC GamePad. The Ultrasound was one of the first PC soundcards to feature 16-bit, 44.1 kHz stereo. The final revision (v3.74) of the GUS Classic features 256 kB of onboard RAM (upgradeable to 1024 kB through DIP sockets), hardware analog mixer, and support for 16-bit recording through a separate daughterboard based on the Crystal Semiconductor CS4231 audio codec.

Reception

[edit]

Computer Gaming World in 1993 criticized the UltraSound's Sound Blaster emulation and lack of native support in games, stating that "it is hard to recommend this card to anyone other than a Windows MIDI musician".[1]

UltraSound MAX

[edit]
Gravis UltraSound MAX

Released in 1994, UltraSound Max is a version of the GUS with a CS4231 codec on board, 512 kB of onboard RAM (upgradeable to 1024 kB with a single SOJ chip), and Panasonic/Sony/Mitsumi CD-ROM interface slots. CS4231 provides support for Windows Sound System specs, although the IO port range doesn't match the WSS hardware, and can be used for SoundBlaster emulation. The software CD includes a demo that featured "3D holographic sound" through the use of software HRTF filters.

Gravis UltraSound PnP Pro

UltraSound Plug & Play (PnP)

[edit]

Released in 1995, the Ultrasound Plug & Play was a new card based on AMD InterWave technology with a completely different sound set. Supposedly Synergy acted as the ODM-producer for it (as evidenced by their logo on the rear side of the card, although early and now very rare GUS PnP cards did not have the Synergy logo). The card features 1 MB of sound ROM, no onboard RAM (although it can be expanded to 8 MB with two 30-pin SIMMs), and an ATAPI CD-ROM interface. A 'Pro' version adds 512 kB of on-board RAM required for compatibility with the GUS Classic. In 2014, a RAM adapter for the 72-pin SIMM was produced by retro-computer enthusiasts that made it possible to install 16 MB of RAM on the 'Pro' version without any modifications to the card.

UltraSound ACE (Audio Card Enhancer)

[edit]
Gravis UltraSound ACE

Released in 1995, this budget version of UltraSound Classic has 512 kB of RAM (upgradable to 1024 kB, just as is the MAX), and has no game port or recording ability. Marketed as a competitor to Wave Blaster-compatible cards, it is supposed to be installed alongside a SoundBlaster Pro/16 card as a sample-based synthesis (marketed as 'wavetable synthesis') upgrade. A prototype of this card was named "Sound Buddy".

UltraSound CD3

[edit]

An OEM version of UltraSound Classic produced by Synergy, with 512 – 1024 kB of RAM. It features AT-BUS CD-ROM interfaces following Sony, Mitsumi and MKE/Panasonic standards. This is the only Gravis sound card with a green circuit board It is similar to a few card clones, including the Primax SoundStorm Wave (model Sound M-16B) and the AltraSound.

UltraSound Extreme

[edit]
Gravis UltraSound Extreme

Released in 1996, the UltraSound Extreme is a 3rd party OEM system combining the UltraSound Classic with an ESS AudioDrive ES1688 sound chip for Sound Blaster Pro and AdLib emulation. It was produced by Synergy as was the ViperMAX. It has 1 MB RAM by default and cannot be upgraded any further.

UltraSound Clones and OEM cards

[edit]

All clones use the original Gravis GF1 or the AMD InterWave soundchip.

  • Primax SoundStorm Wave (GF1) – there are two variants of cards from the well known scanner and mouse producer. Re-labeled Altrasound as Sound M-16B and different Sound M-16C with 4x CD-ROM Interfaces.
  • D&B UltraWave (GF1) – this card has 512 kB onboard RAM by default, upgradeable to 1024 kB RAM using a 512 kB SOJ-40 DRAM chip. Has a DIP socket for a 16 kB boot ROM (optional) and an IDE interface (secondary).
  • Synergy ViperMAX (GF1) – same card later repacked as UltraSound Extreme, but with only 512 kB RAM on board.
  • Expertcolor MED3201 (InterWave LC) – probably the only card with cut-down variant of GFA1 chip - AM78C200 InterWave LC. First series was with standard Am78C201KC.
  • Compaq Ultra-Sound 32 (InterWave) – one of the last InterWave cards was designed for Compaq Presario desktops. Newer "C" revision of InterWave - AM78C201AKC and TEA6330T fader. Produced by STB Systems.
  • STB Systems Soundrage 32 (InterWave) – standard InterWave card missing SIMM slots and IDE interface. There was "Pro" variant with 512 kB RAM. AM78C201KC chip.
  • Core Dynamics DYNASonix 3D/PRO (InterWave) – features additional DSP chip that offered a graphic equalizer and additional sound FX presets.
  • Philips PCA761AW (InterWave) – card design closely resembles the "AMD InterWave OEM Design" prototype. Has a footprint for 512kb RAM, often left unpopulated. AM78C201KC chip.
  • Reveal WAVExtreme 32 (InterWave) – AM78C201KC based design. Comes without RAM and has no sockets/footprints to add any.
  • As of February 2015 there have been efforts by hobbyists to produce a new InterWave based UltraSound compatible soundcard named ARGUS.
  • PicoGUS (open source) – Pico Pi (RP2040) based design started in 2022 by Ian Scott. Community support built up over the subsequent years. As of 2024 its emulation has similar compatibility to an original GUS. The PicoGUS also includes support for Tandy 3 Voice like the Tandy 1000, Game Blaster / Sound Blaster 2.0 (Sound Blaster), and intelligent mode MPU-401 support. [2][3]
  • MK1869 Xtreme – A sound card with both AMD InterWave and ESS ES1869 sound chip made by Marmes and Keropi of pcmidi.eu in 2025. The card features 1MB of sample ROM (standard InterWave bank) and 8MB of RAM soldered on board. At release time initially 4MB cards were also available, and provided soldering pads on the back side to add another 4MB of RAM. The card also features a wavetable daughterboard header, AC97 front panel audio header and connectors for PC Speaker and CD audio.

GF1

[edit]

The GF1 was co-developed by Advanced Gravis and Forte Technologies (creator of the VFX1 Headgear virtual reality helmet) and produced by Integrated Circuit Systems under the ICS11614 moniker. The chip was derived from the Ensoniq OTTO (ES5506) chip, a next-generation version of the music-synthesizer chip found in the Ensoniq VFX and its successors.

The GF1 is purely a sample-based synthesis chip with the polyphony of 32 oscillators, so it can mix up to 32 mono PCM samples or 16 stereo samples entirely in hardware. The chip has no built-in codec, so the sounds must be downloaded to onboard RAM prior to playback. Sound compression algorithms such as IMA ADPCM are not supported, so compressed samples must be decompressed prior to loading.

The sound quality of the GF1 is not constant and depends on the selected level of polyphony. A CD-quality 44.1 kHz sample rate is maintainable with up to 14-voice polyphony; the sample rate progressively deteriorates until 19.2 kHz at the maximum of 32-voice polyphony. The polyphony level is software-programmable, so the programmer can choose the appropriate value to best match the application. Advanced sound effects such as reverberation and chorus are not supported in hardware. However, software simulation is possible; a basic "echo" effect can be simulated with additional tracks, and some trackers can program effects using additional hardware voices as accumulators.

Sample RAM

[edit]

The UltraSound offers MIDI playback by loading instrument patches into adapter RAM located on the card, not unlike how instruments are stored in ROM on other sample-based cards (marketed as "wavetable" cards). The card comes with a 5.6 MB set of instrument patch (*.PAT) files; most patches are sampled at 16-bit resolution and looped to save space. The patch files can be continuously tweaked and updated in each software release.

The card's various support programs use .INI files to describe what patches should be loaded for each program change event. This architecture allowed Gravis to incorporate a General MIDI-compatible mapping scheme. Windows 95 and 98 drivers use UltraSound.INI to load the patch files on demand. In DOS, the loading of the patches can be handled by UltraMID, a middleware TSR system provided by Gravis that removes the need to handle the hardware directly. Programmers are free to include the static version of the UltraMID library in their applications, eliminating the need for the TSR. The application programmer can choose to preload all patches from disk, resizing as necessary to fit into the UltraSound's on-board RAM, or have the middleware track the patch change events and dynamically load them on demand. This latter strategy, while providing better sound quality, introduces a noticeable delay when loading patches, so most applications just preload a predefined set.

Each application can have its own UltraMID.INI containing a set of patch substitutions for every possible amount of sample RAM (256/512/768/1024 kB), so that similar instruments are used when there is not enough RAM to hold all of the patches needed (even after resampling to smaller sizes). Unused instruments are never loaded. This concept is similar to the handling of sample banks in digital samplers. Some games — including Doom, Doom II and Duke Nukem 3D — come with their own optimized UltraMID.INI.

The UltraSound cards gained great popularity in the PC tracker music community. The tracker format was originally developed on the Commodore Amiga personal computer in 1987, but due to the PC becoming more capable of producing high-quality graphics and sound, the demoscene spilled out onto the platform in droves and took the tracker format with it. Typical tracker formats of the era included MOD, S3M, and later XM. The format stores the notes and instruments digitally in the file instead of relying on a sound card to reproduce the instruments. A tracker module, when saved to disk, typically incorporates all the sequencing data and samples, and typically the composer would incorporate their assumed name into the list of samples. This primitive precursor to the modern sampler opened the way for Gravis to enter the market, because the requirements matched the capabilities of the GF1 chip ideally. The problem with other sound cards playing these formats was that they had to downmix voices into one or both of its output channels in software, further deteriorating the quality of 8-bit samples in the process. An UltraSound card was able to download the samples to its RAM and mix them using fast and high-quality hardware implementation, offloading the CPU from the task. Gravis realized early on that the demo scene support could be a sales booster, and they gave away 6000 cards for free[citation needed] to the most famous scene groups and people in the scene.

Compatibility

[edit]

As the GF1 chip does not contain AdLib-compatible OPL2 circuitry or a codec chip, Sound Blaster compatibility was difficult to achieve at best. Consumers were expected to use the included emulation software to emulate other standards, an activity not necessary with many other cards that emulated the Sound Blaster through their sound hardware. The emulation software ran as a huge TSR that was difficult to manage in the pre-Windows days of complicated DOS extenders.

Although there was native support for many popular games that used middleware sound libraries like HMI (Human Machine Interfaces) Sound Operating System, the Miles Audio Interface Libraries (AIL), the Miles Sound System or others, the user had to patch the games by replacing the existing sound drivers with the UltraSound versions provided on the installation CD. Also, the UltraSound required two DMA channels for full-duplex operation, and 16-bit channels were generally faster, so many users chose to use them, but this led to errors for games that used the DOS/4GW DOS extender, which was common in the UltraSound's era.

The two principal software sound emulators included with software package were:

  • SBOS, Sound Board OS — Sound Blaster Pro 8-bit stereo emulation and AdLib FM synthesis. It was a real-mode software emulator that recreated the AdLib's OPL2 FM synth chip and required that the user have a 286 processor or better. There were special versions for the UltraSound MAX (MAXSBOS) and AMD InterWave-based cards (IWSBOS), which made use of the CS4231 codec chip instead.
  • Mega-Em — advanced emulation software that required at least a 386 processor and EMM manager with DPMI/VCPI support. Mega-Em emulated the 8-bit Sound Blaster circuitry for sound effects and the Roland MT-32/LAPC-I or Roland Sound Canvas/MPU-401 for music. It supported UltraMID TSR functionality.

AMD InterWave

[edit]
AMD InterWave AM78C201KC

The great potential of the original UltraSound enabled Advanced Gravis to license the new GFA1 chip and software to AMD, who were trying to enter the sound chip market at the time. The chip, released in 1995, was named AMaDeus, with the AMD part number of Am78C201 and was marketed as InterWave. It was enhanced to handle up to 16 MB of onboard memory, IMA ADPCM-compressed samples, have no sample rate drop at full 32 voices, and featured additional logic to support hardware emulation of FM synthesis and simple delay-based digital sound effects such as reverb and chorus. It was compatible with CS4231 codec installed in the UltraSound MAX or 16-bit recording daughterboard for the UltraSound Classic.

The sound "patch set" was reworked from a collection of individual instrument .PAT files to a unified .FFF/.DAT sound bank format, resembling SoundFont, which could be either ROM or RAM based. There were 4 versions of the sound bank: a full 16-bit 4 MB with 8-bit downsampled 2 MB version, and 16-bit 2 MB (different sample looping) with 8-bit downsampled 1 MB version. A converter utility, GIPC, was provided for making .FFF/.DAT banks out of .PAT/.INI collections.

The reference card contained a 1 MB μ-law ADPCM compressed sound ROM, which contained basic General MIDI voices and sound samples to help FM emulation, and 2 slots for RAM expansion through 30-pin SIMMs. The IWSBOS emulator was reworked to include Mega-Em features such as General MIDI emulation, and the SBOS kernel was included in Windows 95 drivers to provide emulation in a DOS Box window.

The process of patching middleware sound 'drivers' was greatly simplified with PREPGAME utility, which could fix most known DOS games automatically either by correctly installing and configuring native InterWave drivers or replacing the binaries for some rare devices like Covox. It could also update DOS/4GW extender to work around its 16-bit DMA bug.

The GFA1 featured a GUS/MAX compatibility mode, but the base card was not compatible with the UltraSound Classic unless some memory was installed.

The InterWave technology was used in the Gravis UltraSound PnP line of cards. It was also licensed to various OEMs such as STB Systems, Reveal, Compaq, Core Dynamics, Philips and ExpertColor. Some high-end OEM variants contained a full-blown 4 MB patch set in ROM[citation needed] and proprietary hardware DSPs to enable features like additional sound effect algorithms and graphic equalizer.

Software drivers for the InterWave were written by eTek Labs, containing the same development team as the earlier Forte Technologies effort. eTek Labs was split off from Forte Technologies just prior to this effort. In August 1999, eTek Labs was acquired by Belkin and is currently their research and development team.

Demise

[edit]

Some game developers of the time noted problems with the software development kit and the product's hardware design. On the user-side, the Sound Blaster emulation was especially hard to get right out of the box, and this resulted in a substantially high number of product returns at the store level and thus soured the retail channel on the product. Bundled software was refined over time, but Gravis could not distribute updates effectively.

The company itself also created its own trouble. When Gravis's list of promised supporting game titles failed to materialize, the company lost credibility with consumers and commercial developers. Several publishers and developers threatened to sue the company over misrepresentation of their products — pointing to outright fabrication of Gravis's list.

The shareware games industry embraced the Gravis more than the retail games industry.[citation needed] Companies which did this in an early stage were publisher Apogee and developers id software and Epic MegaGames. Gravis can also claim victory in the demoscene, which had taken the GUS to its heart, ensuring a dedicated cult following for a number of years. But without the marketing and developer presence of Creative Labs, Gravis could not generate either the sales or support required for the Gravis soundcard to compete in the mainstream market against the de facto standard Soundblaster.

Although the InterWave chip was a substantially improved version of the GF1 chip, this new design was not able to hold up with the Sound Blaster AWE32. More than that, AMD was facing financial troubles at the time so it was forced to close many projects, including the InterWave.

Due to declining sales, Gravis was eventually forced out of the soundcard business, and the UltraSound's failure nearly took the entire company down with it. Advanced Gravis, once one of the dominant players in the PC peripherals marketplace, had bet much of the future of the company on the UltraSound and paid the price for its demise. Shareholders sued the company charging gross incompetence by its management, in regards to the entire UltraSound effort. After significant restructuring, including acquisition by competitor Kensington Technology Group (via its parent, ACCO World Corp), the company retreated to its core market, the one which had made it a success — joysticks and gamepads.

Software supporting GUS patches

[edit]

Emulators with GUS support:

Software synthesizers which can use GUS patches:

References

[edit]

Sources

[edit]
[edit]
Revisions and contributorsEdit on WikipediaRead on Wikipedia
from Grokipedia
The Gravis UltraSound (often abbreviated as GUS) is a family of 16-bit sound cards produced by Advanced Gravis Computer Technology, first released in late 1992 as a high-fidelity alternative to FM synthesis-based cards like the Sound Blaster, featuring via the proprietary GF1 chip for 32 simultaneous voices, 16-bit stereo playback at up to 44.1 kHz, and onboard sample RAM expandable from 256 KB to 1 MB or more. Developed through a joint venture between Advanced Gravis and Forte Technologies starting in 1991, the original UltraSound card aimed to deliver professional-grade audio for PC gaming, multimedia, and music production by leveraging rather than the less expressive FM tones prevalent at the time, though it launched at a premium price of around $110. Subsequent models addressed early limitations, such as the 1994 GUS MAX with 16-bit recording at 48 kHz, full-duplex operation, and integrated CD-ROM support via a proprietary interface, followed by the 1995 Plug & Play (PnP) variant using the AMD InterWave chipset for easier configuration and the Extreme model incorporating an ESS AudioDrive chip for improved Sound Blaster emulation. Technically, the GF1 processor enabled direct loading of custom samples into onboard DRAM for realistic instrument emulation, with software support for General MIDI via custom patches and offering software-based compatibility with AdLib and Sound Blaster via utilities like SBOS, though native emulation was often unreliable and required additional patches for optimal performance in games. Early models supported 8-bit recording (mono or stereo) at up to 44.1 kHz via a Crystal CS4231 codec, while later iterations added stereo mixing, joystick ports, and up to 4 MB of sample RAM, making it ideal for tracker software and MIDI sequencing but resource-intensive due to the lack of onboard ROM sounds. Despite its superior audio quality—praised for CD-like playback and expressive music in titles like Doom and trackers—the GUS series struggled commercially due to inconsistent software support, complex setup for DOS users, and the dominance of Creative Labs' ecosystem, leading to its discontinuation by 1997 amid the shift to PCI architectures. It remains notable in retro computing for its influence on wavetable technology and enduring popularity among enthusiasts for emulating vintage game soundtracks.

History and Development

Origins and Creation

Advanced Gravis Computer Technology, Ltd. was founded in 1982 in , , , by childhood friends Grant Russell and Scott-Jackson, who shared a passion for computer games and music. The company initially concentrated on producing computer peripherals, with an early focus on joysticks such as the , which became one of the best-selling gamepads for PCs in the late . In 1991, Advanced Gravis formed a with Forte Technologies to create a high-performance alternative to Creative Labs' dominant sound cards. Forte Technologies, known for virtual reality hardware like the , supplied the core GF1 chip design, a wavetable derived from Ensoniq's technology and manufactured by Integrated Circuit Systems (ICS). This collaboration aimed to deliver advanced audio capabilities without relying on the CPU-intensive FM synthesis prevalent in existing cards. The development team sought to leverage for reproducing high-fidelity , using digitized instrument samples stored in onboard RAM to surpass the synthetic tones of FM-based systems while maintaining low CPU usage for real-time audio . Prototypes underwent testing in 1992, with the design specifically appealing to gaming enthusiasts and the community, who required versatile, high-quality sound output for trackers and demos without significant performance overhead. Early production encountered hurdles in achieving full compliance with the Multimedia PC (MPC) standards, as initial models featured only 8-bit digital recording and lacked integrated 16-bit digital audio mixing, necessitating software workarounds for complete multimedia functionality.

Initial Release and Reception

The Gravis UltraSound Classic was released in October 1992 as a 16-bit ISA sound card priced at $249 USD, bundled with a set of instrument patches and software including UltraSound Studio for music composition and playback. Marketed by Advanced Gravis Computer Technology Ltd. as a premium option for PC users, it positioned itself as a superior alternative to the dominant Sound Blaster series, highlighting its 32-voice wavetable synthesis capabilities powered by the GF1 chip for enhanced audio in DOS-based gaming and multimedia applications. Contemporary reception was mixed, with strong for its audio fidelity in niche communities but over compatibility hurdles. In the , the card gained rapid popularity for its hardware-mixed playback of MOD and S3M tracker formats, allowing complex music without taxing CPU resources and enabling more elaborate productions than competitors. A review in Computer Gaming World (October 1993) lauded the UltraSound's quality as noticeably improved and superior for , yet faulted its emulation as unreliable and requiring extensive configuration for games lacking native support. This initial scarcity of developer support for the card's unique features posed a significant barrier to broader adoption in mainstream gaming. Early sales were modest, with estimates suggesting fewer than 150,000 units of the produced overall, reflecting its limited compared to Sound Blaster's dominance. In response to feedback, Gravis issued driver updates in 1993 that enhanced compatibility in 386 Enhanced Mode and achieved MPC Level 2 certification, improving multimedia integration and interface standards for better performance in emerging applications.

Technical Architecture

GF1 Chip

The GF1 chip, developed by Forte Technologies in 1991 in collaboration with Advanced Gravis Computer Technology, formed the heart of the Gravis UltraSound sound card as a proprietary 16-bit (DSP) optimized for . This custom-designed chip enabled high-fidelity audio playback by storing and manipulating digital waveform samples from real instruments, allowing for more realistic sound reproduction compared to (FM) synthesis prevalent in contemporary sound cards. Forte Technologies, known for innovations like the VFX1 virtual reality helmet, contributed expertise in to create a solution tailored for PC gaming and applications of the era. At its core, the GF1 supported up to 32 simultaneous voices through wavetable synthesis, with playback at sample rates reaching 44.1 kHz for CD-quality audio, though the effective number of voices varied inversely with the sample rate—typically around 14 voices at the highest resolution to maintain performance. Key integrated components included a DMA controller for low-latency data transfers from system memory, a 20-bit address bus capable of addressing up to 1 MB of sample RAM for storing waveforms and patches, a 16-bit DAC for stereo output, and an 8-bit ADC for mono or stereo recording at rates from 2 kHz to 44.1 kHz. These elements allowed the chip to operate efficiently on the 16-bit ISA bus, utilizing configurable IRQ lines (such as IRQ 5 or 7) and DMA channels (typically 1 and 3 for 8-bit transfers, with 16-bit options on compatible systems) to minimize CPU involvement in audio tasks. The GF1's capabilities centered on hardware-accelerated mixing of all voices directly on the chip, enabling seamless real-time audio without burdening the host processor, which was particularly advantageous for resource-constrained 386 and 486-era PCs. It supported advanced features like looped sample playback for sustained notes and multi-stage volume envelopes (attack, decay, sustain, and release) to simulate instrument dynamics, enhancing expressiveness in music and effects. This power-efficient architecture, drawing minimal current for ISA compatibility, prioritized low-latency performance for interactive applications while offloading complex operations like sample to reduce artifacts such as . Despite its strengths, the GF1 had limitations, notably the lack of an integrated OPL chip for FM synthesis, which meant no native support for AdLib or Pro standards; compatibility required software drivers like SBOS for emulation, often resulting in imperfect performance for . This design choice focused resources on wavetable strengths but necessitated additional TSR programs for broader DOS ecosystem integration.

Sample RAM and Synthesis

The Gravis UltraSound utilized onboard DRAM to store raw samples, known as patches, which served as the foundation for its capabilities. The card featured 256 to 1024 KB of DRAM, expandable via socketed DRAM chips such as SOJ packages, allowing users to increase capacity for more extensive sample libraries. This memory held digitized instrument samples, enabling the card to mix up to 32 polyphonic voices in real-time, with each voice drawing from the stored waveforms to produce audio output. The synthesis process relied on the GF1 chip to fetch samples from the DRAM via (), which efficiently transferred data without burdening the host CPU. Once retrieved, the GF1 applied and volume envelope modulation to the samples, supporting techniques such as , , and multi-stage envelopes for dynamic sound shaping. This approach facilitated multi-sampled instruments, where different waveforms were used across pitch ranges to enhance realism, providing a more natural compared to (FM) synthesis methods like the OPL-3 chip. Hardware looping further optimized playback by seamlessly repeating sample segments, reducing the overall required for sustained notes. Patch management involved loading instrument definitions from .PAT files using utilities such as SBASIC or the included tools, which organized samples into the DRAM for playback. With a maximum of 1 MB DRAM, the card could accommodate approximately 100 to 200 instruments, varying by sample quality and length—for instance, the General MIDI set included 192 patches, but higher-fidelity multi-samples like those for acoustic required more space. This manual loading process was necessary due to the lack of onboard ROM, but it allowed customization for specific applications. One key advantage of this wavetable system was its low CPU overhead on a 386 processor, thanks to the hardware-managed mixing and DMA operations, which offloaded processing from the host system. This made it particularly effective for orchestral sequences, where it delivered superior audio fidelity over FM-based alternatives, though the need for patch installation added setup complexity. Samples were stored in 16-bit signed format () at variable rates up to 44.1 kHz, ensuring compatibility with high-quality audio while maintaining efficient .

Audio Processing and Compatibility

The Gravis UltraSound's audio pipeline begins with the GF1 chip, which handles wavetable synthesis and mixes up to 32 voices before outputting digital audio signals to a 16-bit digital-to-analog converter (DAC) supporting stereo playback at 44.1 kHz. Early models of the UltraSound, such as the Classic variant, include a Crystal CS4231 codec for direct audio input, enabling digitization of sound sources from line-in or microphone ports and mixing with synthesized audio from sample RAM for unified stereo output. This setup allows the synthesized audio from sample RAM to be mixed with digitized inputs during playback, providing a unified output stream without native hardware mixing for FM synthesis. For compatibility with existing software, the offers partial emulation of 2.0 and Pro standards through the GFOS driver and SBOS software layer, enabling mono playback but without hardware acceleration for FM synthesis via OPL-2 or OPL-3 chips. This emulation focuses on and wave table strengths, as the absence of native OPL support limits FM-based audio to software rendering, which was a common workaround in early drivers. Initial support also extends to the Windows Sound System and MPC Level 1 standards, facilitating basic multimedia compatibility in DOS and early Windows environments. The card operates via 8-bit ISA slots, utilizing DMA channels in modes 0, 1, and 3 for efficient data transfer during playback and recording, with configurable IRQs (such as 2, 3, 5, 7, 11, 12, or 15) to minimize system conflicts. Recording capabilities support 8-bit mono or stereo audio at sample rates from 2 kHz to 44.1 kHz, captured through external or line-in ports, after which the digitized samples can be mixed with ongoing synthesized playback. Known limitations include resource conflicts in multi-card configurations, where shared DMA or IRQ settings could cause instability, particularly in systems with other ISA peripherals; these were partially addressed in driver updates, such as version 3.55 released in , which improved overall stability and emulation reliability.

Product Line

Classic and MAX Variants

The , introduced in 1992, utilized the proprietary GF1 chip for , providing up to 32 simultaneous channels without an onboard for processing. It featured 256 KB of base sample RAM, expandable to 1 MB via additional memory chips, and was designed as an ISA card compatible with both 8-bit and 16-bit slots. The card included a joystick port configurable through and bundled a 4 MB patch set for instrument playback, enabling high-fidelity sound reproduction in supported applications. Priced at approximately $110 upon release, the emphasized sample-based synthesis over FM or digitized audio emulation, though its compatibility exhibited limitations such as incomplete support for certain effects. Revisions of the extended to version 1.9, incorporating minor hardware refinements for stability in DOS and early Windows environments. These updates maintained the core architecture while addressing early production issues, ensuring broad compatibility with ISA bus systems and DMA transfers for efficient audio handling. The UltraSound MAX, released in 1994 as an enhanced variant, integrated a Crystal CS4231 16-bit to support full features, including stereo recording and playback at up to 48 kHz, thereby filling gaps in the Classic's audio input and mixing capabilities. It shipped with 512 KB of sample RAM, expandable to 1 MB, and retained the GF1 chip for while adding support for MPC Level 2 standards. Like the Classic, the MAX included a joystick port and the 4 MB patch set, but introduced an optional interface via daughterboard compatibility for , Mitsumi, and drives. Priced similarly at $249, the MAX preserved the Sound Blaster emulation from prior models, inheriting its known shortcomings in digital effect rendering. The MAX underwent revisions up to version 2.1, which featured improved capacitors for enhanced reliability and reduced noise in analog sections. Both the Classic and MAX variants operated on the ISA bus with configurable ports, IRQs, and DMA channels, prioritizing and wavetable performance in gaming and applications.

Plug & Play and Later Models

The Gravis UltraSound Plug & Play, released in 1995, marked a significant redesign of the UltraSound series to incorporate / (PnP) standards, enabling automatic resource configuration primarily for environments. This model utilized the InterWave chip for hardware auto-detection of IRQs, DMAs, and I/O ports, thereby minimizing manual jumper configurations and resolving common resource conflicts in multi-device systems. It shipped with 256 KB or 512 KB of base sample RAM (512 KB for the Pro variant), expandable to 8 MB via onboard slots, and featured an integrated supporting 16-bit stereo recording and playback at up to 48 kHz. The card retained the core 32-voice capability while providing full emulation and complete DOS compatibility. Design changes in the Plug & Play included a shift to a green (PCB) and further reduction in jumper requirements compared to prior models, streamlining installation for end users. Although PnP compliance improved integration with operating systems like , the adoption of the InterWave chip introduced minor alterations to the instrument sound set relative to earlier GF1-based cards. Hardware specifications preserved 32 simultaneous voices for synthesis and added 32 channels for digital effects mixing, with firmware updates later enabling addressing up to 16 MB of RAM in compatible configurations. The UltraSound Extreme, launched in 1995 as an evolution of the Plug & Play line, featured 1 MB of sample RAM (or 512 KB in some variants) and enhanced emulation through integration with an ESS AudioDrive chip for improved legacy compatibility. This variant also underpinned the ViperMax OEM edition and incorporated early support for 3D positional audio effects via hardware mixing channels, targeting gamers amid rising demand for immersive sound in titles. However, sales volumes for these later models remained low, hampered by the industry's rapid transition from ISA to PCI bus architectures, which rendered ISA-based cards like the Extreme obsolete shortly after release despite its gamer-focused enhancements.

Accessories and Clones

The Gravis UltraSound ACE, released in as an external daughterboard, enhanced the original model by integrating a Crystal Semiconductor CS4231 and 512 KB of RAM, enabling improved 16-bit multichannel playback without requiring a full card replacement. Priced at $99 on the street, it provided 32 voices of and connected via a PC expansion slot or line-out jack to any 8- or 16-bit , though it lacked recording capabilities, support, or a port. This accessory targeted users seeking upgraded audio performance on a budget while maintaining compatibility with the GF1 chip's synthesis features. The CD3, introduced in 1994, served as a dedicated CD-ROM interface daughterboard for the UltraSound, supporting single- and dual-speed drives from manufacturers including Mitsumi, Panasonic, and Sony. It facilitated CD-quality playback and was often bundled with the UltraSound MAX variant to expand multimedia capabilities, integrating directly onto the base card for seamless operation with early optical drives common in mid-1990s PCs. Third-party clones and OEM implementations extended the UltraSound ecosystem by leveraging the GF1 chip or its successor, the AMD InterWave, often at reduced costs but with modified features for broader compatibility. Notable examples include the Synergy ViperMAX and Ultrasound Extreme (1995), which combined the GF1 with an ESS AudioDrive 1688 chip for Ad Lib and support, offering 512 KB or 1 MB RAM configurations at around $100. Other variants, such as the Core Dynamics DynaSonix series, provided GF1-based functionality with additions like FM radio tuners in select models, while InterWave-based cards like certain Ensoniq Vivo 90 OEM products incorporated onboard ROM samples for enhanced without user-loaded patches. These clones numbered around a in total, prioritizing affordability over full feature parity. Following Advanced Gravis's acquisition in , clones played a key role in sustaining the platform's availability, as original production ceased; estimates indicate 120,000 to 150,000 genuine Classics were sold during the company's active years. While clones boosted accessibility for enthusiasts, they frequently exhibited variable build quality and partial incompatibility with UltraSound-specific patches, limiting their appeal in demanding applications compared to official hardware.

AMD InterWave Technology

Development and Features

The InterWave chip, designated as the AM78C201, was developed by between 1994 and 1995 under a licensing agreement with Advanced Gravis and Forte Technologies, the co-creators of the original GF1 chip used in earlier Gravis UltraSound cards. This collaboration allowed to produce an enhanced, backward-compatible successor to the GF1, enabling software designed for the original chip to operate seamlessly on InterWave-based hardware. The chip was formally announced and showcased at the Computer Game Developers Conference in 1995, with production ramping up for availability by late that year. Key enhancements over the GF1 included 24-bit memory addressing, supporting up to 16 MB of sample RAM via dual 30-pin slots, a significant increase from the GF1's 1 MB limit. Additional DMA channels were incorporated to handle 16-bit audio transforms, such as reverb and filtering effects, improving processing efficiency for complex sound effects. The core architecture provided 32 voices of 16-bit at up to 44.1 kHz sampling rates, with an integrated 16-bit DAC and ADC for high-fidelity stereo output and input. The chip itself was primarily designed for ISA architectures. Design goals emphasized enhanced compatibility while delivering superior wavetable audio quality and realism compared to FM synthesis, targeting consumer-priced cards under $100. To reduce CPU overhead, particularly in Windows environments, the chip offloaded more audio processing to hardware, including on-chip support for effects like , , chorus, , and phase shifting. Early hardware 3D audio positioning was introduced as a precursor to (HRTF) techniques, enabling spatial sound rendering with minimal software intervention. In production, the InterWave was deployed in the Gravis UltraSound Plug & Play series and various third-party cards from manufacturers such as STB Systems, , , ExpertColor, and Reveal, often with 1 MB or 2 MB onboard ROM for sample storage. was upgradable via , allowing configurations to emulate GUS Classic mode for legacy compatibility and custom patch loading in formats like .FFF or .DAT.

Integration in Gravis Products

The InterWave chip debuted in Gravis products with the Plug & Play in , replacing the original GF1 chip to provide full integration on a single chip. This design enabled native 16-bit file playback and recording at up to 48 kHz without requiring external hardware like the CS4231 used in earlier models. The InterWave's integrated architecture supported hardware mixing for 32 digital channels and full-duplex operation, allowing simultaneous audio playback and recording for applications such as voice chat. In the Plug & Play Pro variant, also released in 1995, the InterWave came with 512 KB of default sample RAM, expandable to 8 MB via 30-pin SIMMs, which facilitated seamless use of existing GUS patches for . Improved IRQ sharing through Plug & Play compliance reduced configuration conflicts, while the chip's backward compatibility allowed switching to GF1 emulation mode for legacy DOS software. Additional features included 16-bit DMA support for lower-latency audio processing and compatibility with onboard patches up to 5.6 MB when sufficient RAM was installed. Despite these advances, integration challenges arose due to differences in the InterWave's sound set compared to the GF1, often requiring updated patches for optimal playback in games and applications. The chip's higher power consumption in some configurations could strain older power supplies, though this was less pronounced in the ISA-based PnP models. Adoption was centered on Gravis' own lineup and OEM partners like , which produced compatible InterWave cards, thereby extending the viability of the ISA platform for high-fidelity audio into the mid-1990s before the shift to PCI and integrated solutions.

Software and Game Support

GUS-Specific Patches and Tools

The Gravis UltraSound utilized .PAT files as its primary patch format for , enabling high-fidelity playback through multi-sampled instruments stored in the card's onboard RAM. These files begin with the "GF1PATCH110" identifier and include a header specifying manufacturer details, instrument count, voices, output channels, , master volume, and size, followed by instrument headers and patch records containing , loop points, sample rates, frequency ranges, envelopes for and filter modulation, and effects like and . Each .PAT file could reach up to 1 MB in size, accommodating looped 16-bit samples to optimize memory usage, though the full General MIDI-compatible set provided by Gravis in 1993 totaled approximately 5.7 MB across 128 instruments and effects, requiring at least 512 KB of RAM for practical loading. Official tools facilitated patch creation and management, including UltraSound Studio, a DOS-based digital recording and editing suite that allowed users to audition, load, and modify patches via its Patch Manager, with features for waveform editing, mixing, and GM mapping. The suite supported 8-bit recording at rates up to 44.1 kHz, with upgrades enabling 16-bit stereo capabilities, and included basic scripting-like controls for custom voice modulation through envelope adjustments. Additionally, the driver suite version 4.11, released in 1995, provided comprehensive setup utilities for DOS and /3.11 environments, incorporating Ultrinit for hardware initialization and drivers like Ultramid for seamless integration. Community efforts expanded the patch ecosystem, particularly within the , where users created optimized collections to fit the card's limited RAM, such as the EAWPATS pack compiling high-quality samples for enhanced module playback on 512 KB configurations. These packs focused on efficient looping and compression to maximize without exceeding memory constraints. Patch management often involved batch scripts like those invoking LOADPATS to preload game-specific instruments from the ULTRADIR directory, ensuring compatibility with DOS applications. The format also supported tracker modules, including XM files from FastTracker II, through libraries like libxmp, which rendered them directly to GUS hardware for authentic wavetable output. Later InterWave-based models, integrating AMD's for up to 8 MB RAM, extended patch handling via converted formats; tools like GIPC transformed standard .PAT files into InterWave-specific .FFF (patch headers) and .DAT (data) files, allowing larger, uncompressed sample sets for improved fidelity. Today, archived GUS patches from official releases, community packs, and OEM variants collectively exceed 100 MB, preserved in repositories that include expansions like Pro Patches Lite 1.61 (14.9 MB) for broader instrument variety.

Notable Applications and Games

The Gravis UltraSound (GUS) found notable adoption in several early 1990s , particularly those leveraging its for superior playback compared to FM synthesis alternatives. id Software's Doom (1993) supported GUS for music through patches that loaded instrument sets into the card's sample RAM, enabling richer soundtracks during gameplay. Similarly, Doom II: Hell on Earth (1994) included native GUS drivers for enhanced audio immersion. Several of id Software's early titles, such as Doom and Doom II, featured optimizations tailored for GUS hardware mixing. (1996) provided full native support, utilizing the card's 32 channels for dynamic sound effects and music, which contributed to its acclaim in the FPS genre. In the demoscene, the GUS excelled with tracker-based music formats, showcasing its multi-channel capabilities. Future Crew's Second Reality (1993), a landmark PC demo that won first place at Assembly '93, incorporated MOD-style music composed by Purple Motion and Skaven, rendered through GUS for high-fidelity playback that highlighted the card's sample-based synthesis. Trackers like Scream Tracker 3, developed by Future Crew, natively exported modules to GUS-compatible formats, supporting up to 32 channels of 8-bit samples and integrating FM synthesis for AdLib-compatible output. This made S3M files particularly well-suited for GUS hardware, allowing demoscene artists to create complex, polyphonic compositions without software mixing overhead. However, many games required manual loading of .PAT files using utilities like LOADPATS for full GUS support, adding to setup complexity. Applications beyond gaming also benefited from GUS integration, especially in and music production. The Miles Sound System (MSS), a popular audio library for DOS games, included drivers for GUS in versions from 1994 onward, enabling seamless wavetable and in adventure titles like (1993), where it powered atmospheric soundtracks via UltraMID patches. sequencers such as Express supported GUS as a wavetable , allowing users to sequence multi-track compositions directly to the card's onboard RAM for professional-grade playback. Support for GUS expanded significantly from 1994 to 1996, with over 50 titles incorporating native drivers by mid-decade, as documented in contemporary compatibility lists totaling around 156 games. This growth reflected developer interest in the card's audio quality for niche PC titles. Today, GUS emulation in via its dedicated core preserves these experiences, routing synthesized audio through modern systems while maintaining hardware-accurate mixing. Despite this, GUS remained rare in mainstream releases due to the dominant preference for compatibility, which offered broader emulation and simpler integration. It thrived instead in tracker music communities, where its sample RAM and channel count enabled intricate MOD and S3M productions that defined the era's evolution.

Decline and Legacy

Factors Leading to Demise

The Gravis UltraSound line suffered from several technical limitations that hindered its adoption among mainstream users. Notably, the cards lacked a dedicated OPL FM synthesis chip, relying instead on imperfect software emulation of AdLib and OPL modes through the GF1 chip, which often resulted in suboptimal sound quality for FM-based games and applications. Additionally, Sound Blaster 16 compatibility was incomplete, requiring third-party TSR drivers like SBOS for partial emulation, which introduced configuration complexities and potential conflicts in multi-card setups. The necessity for manual patch loading via utilities such as Patch Manager further deterred casual users, as it demanded explicit management of instrument samples in the card's onboard RAM, unlike the more seamless wavetable solutions from competitors. Market challenges exacerbated these issues, with the UltraSound priced at a premium—typically $129 for the model by mid-1994 and up to $249 list price for the MAX variant—positioning it as a high-end option in a market dominated by more affordable, versatile alternatives. Total unit sales remained low, estimated at 120,000 to 150,000 for the alone based on analysis, far below the volumes achieved by Creative Labs' series. Creative's integrated cards, such as the AWE32, offered superior and broader game support without additional setup, capturing the majority of the market and leaving Gravis struggling to gain traction beyond niche tracker music and demo scene enthusiasts. Company mismanagement contributed significantly to the line's struggles, including overreliance on the as a core product despite its compatibility shortcomings and misleading claims about software support that failed to materialize widely. Internal decisions, such as heavy investment in proprietary development without prioritizing industry-standard emulation, alienated developers and users, leading to perceptions of incompetence in product strategy. The timing of Gravis's later innovations proved detrimental, as the 1995 shift to AMD InterWave-based models like the Plug & Play came too late to counter the emerging PCI and USB audio standards by 1997, rendering ISA-based cards increasingly obsolete. These InterWave variants improved emulation but arrived after Creative had solidified its dominance with PCI-ready solutions, limiting their market impact. Financial pressures culminated in the company's decline, with revenues peaking in 1995 before dropping in 1996 amid mounting losses from low sales and development costs. Unable to sustain operations, Gravis ceased hardware production and was acquired by Technology Group in late 1997, effectively ending the UltraSound line.

Cultural Impact and Modern Relevance

The Gravis UltraSound pioneered affordable for PC-compatible systems upon its 1992 launch, integrating sample-based synthesis directly into the card without the need for separate daughterboards, which set a for subsequent hardware designs. Its capabilities made it a staple in the , where the card's hardware mixing of up to 32 channels enabled high-fidelity playback of MOD files, fostering a strong following among PC creative communities in the 1990s. The retains a in retro computing circles, with enthusiasts preserving its software ecosystem through dedicated online archives that host patches and tools for its Ensoniq GF1 synthesizer. Instrument patches compatible with the card, originally bundled as sets like DGGUSPAT for compliance, continue to be archived and shared on sites such as ModArchive.org, supporting ongoing music production and preservation efforts. In modern contexts, the Gravis UltraSound is emulated in software like DOSBox-X and to deliver authentic audio for DOS-era games and applications without requiring original hardware. Hardware reproductions have emerged in the 2020s, including ISA clones like the PicoGUS—based on microcontroller emulation—and the ARGUS board, which replicate the original's functionality for authentic retro PC builds, with ongoing availability as of 2025. Original UltraSound cards command collectible value in the vintage hardware market, with functional units typically selling for $200 to $400 on as of November 2025, reflecting demand from preservationists and gamers seeking period-accurate setups. The card's enduring appeal is highlighted in retrospectives like ' 2018 video analysis, which explores its role in 1990s PC audio innovation. The UltraSound's emphasis on programmable, influenced audio quality in PC gaming, contributing to higher standards for tracker music and productions that echo in contemporary revivals and software instruments emulating its wavetable approach.

References

Add your contribution
Related Hubs
User Avatar
No comments yet.