NVIDIA has expanded its Blackwell series of professional graphics cards with the new RTX PRO 5500 Workstation Edition. This model is designed for workstation and server systems that require large amounts of video memory to perform artificial intelligence, physical simulation and graphics tasks.
The accelerator is based on the GB202 GPU, similar to the flagship RTX PRO 6000 and GeForce RTX 5090. The card has 21,760 CUDA cores and a TDP of up to 600 W. Three ninth-generation NVENC and six sixth-generation NVDEC engines are responsible for video processing. The main feature is 84GB of GDDR7 memory – 36GB more than the RTX PRO 5000 and more than double the RTX 5090.
The claimed memory bandwidth reaches 1398GB/s. However, PNY specifies an unusual 416-bit bus, which is why it is expected to use different density DRAM modules. Up to four DisplayPort 2.1b are provided for connecting monitors, and the card itself uses PCIe 5.0 x16 and a 16-pin power connector. Available in air and liquid cooling, the MIG feature lets you split the accelerator into two virtual instances of 42 GB of memory. Pricing hasn’t been announced yet, but it’s certainly not worth looking forward to for its budget-friendliness: the 48GB RTX PRO 5000 is estimated to cost around $9,200.
