Bolt Graphics brings its RISC-V graphics cards to Ubuntu Summit — Zeus path tracing GPUs target film and animation industry

Bolt Graphics’ Zeus GPUs have slowly but surely been generating hype ever since the startup first announced its RISC-V GPUs in March. With its 2026 hardware launch steadily approaching, Bolt made a surprise appearance at the Ubuntu Summit 25.10 this week to talk about the software stack enabling its over-the-top hardware performance claims of being 13 times faster than Nvidia’s RTX 5090.

Antonio Salvemini, Director of Graphics Engineering at Bolt Graphics, presented a talk entitled “Beyond the Silicon: Redefining GPU Innovation Through Software and Methodology” at the Summit. Salvemini talked extensively about Bolt Graphics’ primary focus on path tracing, a graphics technology used primarily in the animation and film VFX worlds.

Path tracing is a step beyond ray tracing, but both rendering technologies have been an integral part of the Hollywood VFX pipeline for years and have only entered the consumer and gamer conversations in the current decade. The Zeus line of GPUs is being advertised, especially at the Ubuntu Summit, to be a way for visual effects artists to render path-tracing visuals in real-time.

You may like

• Legendary GPU architect Raja Koduri’s new startup leverages RISC-V and targets CUDA workloads

• China’s latest GPU arrives with claims of CUDA compatibility and RT support

• Nvidia Rubin CPX die shot allegedly reveals ROPs, raster units, and display pipes — forming the potential foundation for the RTX 6090

“The problem with ray tracing is that each light wave only bounces one way,” said Salvemini in his talk. “In path tracing, they can bounce anywhere, and you randomly select just some of these paths to display.” Path tracing traces light beams in reverse order, from the digital camera back along its exact path, bouncing from object to object, until it returns to the light source it would radiate from, creating a hyper-realistic light simulation only within the bounds of what the “camera” can see, reducing graphical overhead greatly.