GPU Audio has partnered with AMD to open up a plethora of music production and innovation opportunities through the revolutionary parallel processing of graphics cards. After months of the two companies collaborating together to make this happen, AMD Radeon™ GPU users around the world can now experience ultra-responsive audio production technology first-hand.

"It was April when I was pulled into the project — I was astounded at how much [GPU Audio] had done on their own to enable pro audio on our GPUs," Carl Wakeland, Fellow at Advanced Micro Devices, begins. "They had to pretty much rip up the existing infrastructure and replace it in order to bring real-time audio to the point where they're scheduling a lot of complex kernels and getting them to flow in a very precise way." AMD provided the hardware, tools, and consulted on the development of AMD GPU support, but, in Wakeland's words, GPU Audio did "nearly all of the heavy lifting on its own."

The AMD ROCm™ platform, an open-source development platform for GPU computing, was initially designed in C++ for a Linux environment. This was an issue for the pro audio industry as most users work with Windows or Mac OS X ecosystems. The landscape changed when AMD launched HIP, the Heterogeneous Interface for Portability, which provides efficient methods for converting CUDA code to run on AMD Radeon GPUs.

Since the pro audio environment largely exists on the Linux compute platform and there wasn't anything like it on Windows, the team was surprised how GPU Audio developers were able to leverage HIP by converting the CUDA code using the tools on Linux and then bring the port to Windows. "It was quite amazing to us that they took the initiative to do that work," Wakeland shares, saying that while AMD has done some internal work to show that pro audio suits AMD Radeon GPUs really well, they weren't able to find a company that was willing to invest their time and energy to bring this idea to life. "We talked to quite a few but we didn't really run into any that were willing to put in the amount of effort it was going to take to leverage audio on GPUs to its fullest."

Along the way, GPU Audio contributed to HIP implementation and improvement through their own development, thus building on top of HIP, contributing to even better performance for pro audio users.

Looking into the future, Wakeland foresees AMD helping to enable further capabilities on the company's hardware that isn't currently available on the market. Talking about the capabilities of its products, he is certain that AMD holds some particular — and unexpected — advantages in regards to pro audio processing support. "We support asynchronous queues and reserved real-time queues that we developed quite a long time ago," he continues. "They weren't all initially developed with audio in mind but it turned out that they're a really good way to help facilitate running audio scenarios because you can set up multiple queues that run in parallel and that are scheduled independently."

Pro audio is due for a low-level technological innovation. "Audio has always been in the second chair [and] the only way to move it forward is to give it a really strong push to enable something that's never been done," Wakeland says. "And with the latest GPU hardware, you don't have to worry about conserving computer resources anymore."

When it comes to gaming audio on PCs, for instance, people want the highest frame rate and the highest level of resolution. "It's been ten years since we started looking at [that]," Wakeland conveys. "Even if you come up with a way, as we did, to make the whole experience of audio and graphics on GPU completely reliable, the reality is we are still using part of your GPU for audio that you could be using to generate a higher frame rate."

A lot of the time, you might have to set up one scenario for the world that has audio-capable GPUs and another one for the rest of the world that doesn't. As such, these logistics made it very difficult for GPUs to penetrate deeply into the gaming audio world, even though the company could create demos that were "just impossible to do on the CPU with respect to acoustic modeling scenarios."

Now, with growing interest in pro audio and other key markets — from virtual rehearsal to AI applied everywhere such as plugins and audio speech — the collaboration between AMD and GPU Audio is key for audio innovation. "We have a mutual interest and we saw immediately that if [GPU Audio] wants to do this in the pro audio world, it's going to be great and we will help them where we can," Wakeland tells me. He hopes that the engagement is just getting started. Given how robust graphics cards have become and how AMD continues its endeavor to enhance the power of performance-per-watt and performance-per-dollar, it's only a matter of time.

"As time goes on and generations progress, you see that the capability of what was once limited to these big brick discrete GPUs is becoming available in your little gaming laptop," he says. "Where it doesn't work, we evolve it — the curve doesn't stop."

Both GPU Audio and AMD agree that it is a big achievement not only for the two teams, but for the future of audio too. After spending more than half a year developing the technology to add support for AMD Radeon GPUs, GPU Audio is excited to be putting it out and letting the community try it. The company has now revealed public Early Access testing for AMD Radeon GPUs1 using the AMD Software: PRO Edition drivers and with the community's help, the company will expand the support as quickly as possible to many AMD Radeon GPUs.

1* Testing conducted with AMD Radeon RX 5600M, AMD Radeon RX 6700M, AMD Radeon RX 6800M and AMD Radeon PRO W6400. Testing of additional AMD Radeon and AMD Radeon PRO GPUs is ongoing.