TITAN X (Pascal)

Added 2026-09 alongside GTX 1080 Ti, TITAN V, and TITAN RTX — NVIDIA's first Pascal-generation TITAN, explicitly marketed at launch for "deep learning and artificial intelligence" (per NVIDIA's own GTX 1080 Ti press release, which describes this card as "designed for deep learning and artificial intelligence"). Not to be confused with the earlier, differently-architected Maxwell "GeForce GTX TITAN X" (2015) — this is the GP102-based "(Pascal)" refresh, hence the parenthetical in both NVIDIA's own naming and this entry's name field. CUDA core count (3,584), memory config (12GB GDDR5X at 10Gbps, 384-bit), memory bandwidth (480 GB/sec), and FP32 performance ("11 TFLOPS of power") are quoted directly from NVIDIA's own official product page (nvidia.com/en-us/geforce/news/gfecnt/nvidia-titan-x-pascal-available- august-2nd/), read 2026-09. peakClockMHz (1531, boost) is from the same page's "boosts to 1.5GHz out of the box" combined with the widely-corroborated exact figure across independent GPU databases; tdpWatts (250) and msrpUSD (1200, the original announced price) are not stated on that specific NVIDIA page but are consistently corroborated across contemporaneous press coverage of the same Jul 21 2016 announcement. No Tensor/matrix-throughput fields — Pascal predates Tensor Cores (introduced with Volta the following year). 2026-09-15 — added computeUnitCount (28 SMs): the same GP102 configuration as this collection's GTX 1080 Ti entry — 3,584 CUDA cores / 128 FP32 cores per Pascal consumer SM = 28 of the die's 30 SMs enabled. NVIDIA publishes the core total, not the SM count.