GPU Audio

GPU Audio, INC. operates at the intersection of high-performance computing and professional audio, developing technology to offload digital signal processing (DSP) from CPUs to the parallel processing architecture of Graphics Processing Units (GPUs). Originally founded in Switzerland as Braingines SA, the company is now based in the United States. The initial concept emerged around 2011 from a collaboration between Alexander Talashov, who had a background in programming GPUs for high-frequency stock trading, and Vasiliy Sumatokhin, a sound engineer. Joined by co-founder Jonathan Rowden, a saxophonist and composer with a Master's in Jazz Studies, they formally established the company to solve the long-standing challenge of utilizing GPUs for real-time audio. Early attempts faced significant hurdles, including near-bankruptcy, but research that began in earnest in 2014 led to key breakthroughs by 2018.

The company's core business model revolves around providing its patented technology as a platform for other developers through a Software Development Kit (SDK). Instead of focusing on being solely a plugin maker, GPU Audio enables third-party audio software companies, from pro audio to game audio and automotive sectors, to integrate GPU acceleration into their own products. This allows them to create more powerful applications, such as dynamic spatial reverbs, complex synthesizers, and real-time machine learning algorithms for audio, which would be too demanding for traditional CPU-based systems. The company has formed partnerships with entities like Vienna Symphonic Library, Audio Modeling, and Mntra Instruments to demonstrate the technology's capabilities.

GPU Audio's key technological achievement is a full-stack solution that includes a proprietary scheduler, overcoming the latency issues that previously made GPUs impractical for audio. This scheduler manages the complex, sequential tasks of audio processing on the GPU's parallel architecture, enabling real-time performance with latencies as low as one millisecond. Their platform is cross-platform, supporting a wide range of GPUs from NVIDIA and AMD, as well as Apple Silicon, without requiring developers to write device-specific code. By unlocking the immense, often underutilized, processing power of GPUs already present in most computers, the technology allows for a significant increase in processing capacity, enabling musicians and producers to run a virtually limitless number of plugin instances and handle demanding tasks like spatial audio mixing.

Keywords: GPU audio processing, digital signal processing, audio SDK, real-time audio, low-latency processing, VST plugins, audio software development, parallel computing, pro audio technology, accelerated audio computing, audio engineering, music production, game audio, spatial audio, neural amp modeler, convolution reverb, audio developer platform, Alexander Talashov, Jonathan Rowden