NVIDIA Quadro RTX 8000 48GB

The Quadro RTX 8000 is the 48 GB flagship of NVIDIA's first ray tracing workstation pair, built on the same 754 mm2 TU102 die as the Quadro RTX 6000 with 4,608 CUDA cores, 576 second-generation Tensor cores and 72 RT cores. Compute is identical between the two cards, clock for clock and core for core; the RTX 8000 simply doubles the frame buffer to 48 GB, and a pair bridged over NVLink presents 96 GB. That capacity is why the card still turns up on rental marketplaces years after launch, since 48 GB at a low hourly rate remains useful for inference and for models that will not fit a 24 GB card. Turing tensor cores cover FP16, INT8 and INT4 and stop there, so there is no BF16, TF32 or FP8 on this part, and no structured sparsity, which arrived a generation later with Ampere. The datasheet's 130.5 TFLOPS of tensor performance is therefore a dense figure, directly comparable with every other accelerator on this site.