Legacy NVIDIA Pascal GPUs repurposed for real-time 8K60 HEVC V2X encoding

A new paper from Kasidis Arunruangsirilert and Jiro Katto (Waseda University) presents a surprisingly effective method for achieving real-time 8K60 HEVC encoding using repurposed legacy NVIDIA Pascal GPUs. By leveraging 2-Way Split Frame Encoding across dual NVENC units found on GP104 and GP102 dies—think GTX 1080 and GTX 1080 Ti—the team achieved full 8K60 throughput with a rate-distortion penalty of under 1%. This is a breakthrough for Vehicle-to-Everything (V2X) and Tele-Operated Driving applications, which demand ultra-low-latency video telemetry but face supply chain and power budget constraints. The key insight is that smaller Pascal dies are actually more efficient than flagship GPUs because fixed-function encoding forces general-purpose Streaming Multiprocessor cores to idle at high frequencies. Fewer CUDA cores mean less wasted power, making mid-range cards like the GTX 1080 ideal for edge deployments.

While the paper notes a generational compression efficiency gap versus the state-of-the-art RTX PRO 6000 Blackwell, Pascal's native lack of B-frames aligns perfectly with the low-latency requirements of V2X pipelines. The researchers argue that repurposing these old GPUs—which would otherwise become e-waste—offers a highly cost-effective and sustainable solution for modern Intelligent Transportation Systems. The work will be presented at the 2026 IEEE 104th Vehicular Technology Conference (VTC2026-Fall) in Boston. It challenges the assumption that only the latest hardware can meet 8K edge encoding demands, opening the door for circular economy practices in autonomous driving infrastructure.