NVIDIA GeForce RTX 5090

GeForce RTX 5090 is the fastest consumer GPU of the Blackwell generation, with 21,760 CUDA cores, fifth generation Tensor cores rated at 3,352 AI TOPS, fourth generation RT cores at 318 TFLOPS, 32 GB of GDDR7 on a 512 bit interface, and a 575 W power rating that calls for a 1,000 W class system power supply. It drives three DisplayPort and one HDMI output and supports DLSS 4 and hardware AV1 encode and decode.

We supply it in a dual GPU content creation workstation and in 4 GPU and 8 GPU 4U servers, for buyers who want the lowest cost per GPU for inference, fine tuning of small and mid size models, image and video generation and rendering, and who accept consumer licensing terms. It has no ECC memory and no professional driver certification, and NVIDIA does not license the GeForce driver for data center deployment, so we confirm the intended use with every quotation. For AI development with large models, ECC and MIG, the RTX PRO 6000 Blackwell editions with 96 GB are the better fit.