Musk: Tesla AI5 chip brings roughly 12X more compute than current in-car hardware
elonmusk · x · 2026-11-22
Adding to his reality visualizer breakdown, Musk explained the compute constraints:
- After moving the tiny decoder into SRAM, only half of the three TRIP engines on SoC B remain available for the visualizer
- TRIP essentially only handles 8-bit integer math, so softmax must run in emulation mode across 28 steps
- He teased the next-gen AI5 chip as "a gigantic improvement" with roughly 12X more compute
Related event: Tesla Rolls Out FSD V15 with New Architecture and 25x Fewer Interventions(49 posts)→
More from Embodied
- First Drive: Tesla FSD V15 Shines With Reality View Birds-Eye Map and Garage Auto-Parking — Scobleizer · 2026-11-22
- AI drones hunt moths in greenhouses with precision strikes, replacing pesticides — TinfoilTricorn · 2026-11-22
- Robot manipulation training data: teleop, sim, or egocentric video? Practitioners weigh in — TendiesTown3 · 2026-11-22
- One photo, 8 VLAs, only 3 correct: ERQA embodied-reasoning benchmark questioned — giffmana · 2026-11-22
- China Installed 59% of New Industrial Robots in 2026 Report, ~354,000 Units — ingliguori · 2026-11-22
- Apple Watch bricks after one shower while a $15 Keep band survives many — oran_ge · 2026-11-22