Nvidia and TSMC working on multi-GPU solutions based on silicon photonics
A few years ago, Nvidia’s long-lasting partnership with the TSMC fabs came into question as the GPU giant decided to use Samsung’s 8 nm production nodes for the RTX 3000 models. Things appear to have returned to normal with the upcoming RTX 4000 series, and, according to a new report from industry sources close to DigiTimes, the two companies are looking to further strengthen the collaboration with a project involving silicon photonics (SiPh) tech, which should allow very fast connections between multiple GPUs.
DigiTimes specifies that the SiPh project is called COUPE (compact universal photonic engine) and uses CMOS circuitry similar to those found in digital cameras. SiPh chips can be combined with CMOS processes and integrated together with co-packaged optics (CPO). Multiple AI GPUs can thus be interconnected via a chip-on-wafer-on-substrate (CoWoS) 2.5D package. This improved method benefits from the low-latency advantage of optical data transmissions, which also ensures a significantly reduced signal loss between a larger group of interconnected GPUs.
The first GPUs built with TSMC’s SiPh tech are expected to launch in at least a few years. Apparently, Intel was also planning to develop a similar technology with Taiwanese epiwafer maker LandMark Optoelectronics. TSMC avoided any patent infringements by placing the laser light transceivers externally rather than internally, which would reduce transfer speeds, but also ensure a higher yield rate.
Nvidia is not the only company that stands to benefit from TSMC’s SiPh tech. AMD and Xilinx, as well as Broadcom, Cisco and Marvell Technology have also signed contracts with the Taiwanese foundries.
The SiPh module market is estimated to grow to US$4 billion in 2024, almost a 10x increase over the total market valuation recorded in 2018.