During the prestigious IEDM 2024 conference, NVIDIA presented its vision for the future AI accelerator design, which the company plans to chase after in future accelerator iterations. Currently, the limits of chip packaging and silicon innovation are being stretched. However, future AI accelerators might need some additional verticals to gain the required performance improvement. The proposed design at IEDM 24 introduces silicon photonics (SiPh) at the center stage. NVIDIA's architecture calls for 12 SiPh connections for intrachip and interchip connections, with three connections per GPU tile across four GPU tiles per tier. This marks a significant departure from traditional interconnect technologies, which in the past have been limited by the natural properties of copper.

Perhaps the most striking aspect of NVIDIA's vision is the introduction of so-called "GPU tiers"—a novel approach that appears to stack GPU components vertically. This is complemented by an advanced 3D stacked DRAM configuration featuring six memory units per tile, enabling fine-grained memory access and substantially improved bandwidth. This stacked DRAM would have a direct electrical connection to the GPU tiles, mimicking the AMD 3D V-Cache on a larger scale. However, the timeline for implementation reflects the significant technological hurdles that must be overcome. The scale-up of silicon photonics manufacturing presents a particular challenge, with NVIDIA requiring the capacity to produce over one million SiPh connections monthly to make the design commercially viable. NVIDIA has invested in Lightmatter, which builds photonic packages for scaling the compute, so some form of its technology could end up in future NVIDIA accelerators

Lightmatter, a leader in silicon photonics processors, today announces its artificial intelligence (AI) photonic processor, a general-purpose AI inference accelerator that uses light to compute and transport data. Using light to calculate and communicate within the chip reduces heat—leading to orders of magnitude reduction in energy consumption per chip and dramatic improvements in processor speed. Since 2010, the amount of compute power needed to train a state-of-the-art AI algorithm has grown at five times the rate of Moore's Law scaling—doubling approximately every three and a half months. Lightmatter's processor solves the growing need for computation to support next-generation AI algorithms.

"The Department of Energy estimates that by 2030, computing and communications technology will consume more than 8 percent of the world's power. Transistors, the workhorse of traditional processors, aren't improving; they're simply too hot. Building larger and larger datacenters is a dead end path along the road of computational progress," said Nicholas Harris, PhD, founder and CEO at Lightmatter. "We need a new computing paradigm. Lightmatter's optical processors are dramatically faster and more energy efficient than traditional processors. We're simultaneously enabling the growth of computing and reducing its impact on our planet."