Nvidia's Ian Buck marks two decades of CUDA, from one 2006 toolkit to a library for every workload
firstadopter · x · 2026-11-17
Nvidia executive Ian Buck pushed back on the current wave of Nvidia discourse, saying he's "so proud of the work that everyone at Nvidia does," and shared two retrospective slides:
- Two Decades of CUDA: from a single toolkit in 2006 to a library matrix covering every workload — linear algebra/FFT/Monte Carlo, deep learning and graph analytics, genomics/tensor networks/data science, up to quantum, lithography, digital twins, robotics, climate AI and drug discovery, spanning cuBLAS, cuDNN, TensorRT-LLM, Transformer Engine and hundreds more.
- A 2006 supercomputing flashback: CUDA's origin as a "simple, explicit" C-on-GPU programming solution, the GeForce 8800 GTX (128 processors, 768MB GDDR), and early CUDA speedup charts.
A condensed archive of how the CUDA ecosystem grew over twenty years.
More from Companies & People
- OpenAI Foundation hiring Director of Bio Partnerships, plans $1B+ first-year investment — MichaelRetchin · 2026-11-17
- OpenAI's jxnlco on building taste: 'If you want good taste, you have to eat' — jxnlco · 2026-11-17
- Yejin Choi raises $110M for new startup asking: monolithic or plural AI? — jeremyphoward · 2026-11-17
- Feds admit no spy evidence found against ex-OpenAI engineer Alfozan, lawyers say — ns123abc · 2026-11-17
- Developer nahtnam announces he's joining OpenAI — burny_tech · 2026-11-17
- XMoney shows up as the default payment method inside Tesla's Robotaxi app — zeroxjackson · 2026-11-17