inference/research
verifiedmedium · 0.80updated 2026-10-15

NVIDIA RTX PRO 5000 Blackwell Workstation Edition

nvidia blackwell 2025

Fields

id
accelerators/nvidia-rtx-pro-5000-blackwell
Vendor
nvidia
Architecture
blackwell
Release year
2025
Process (nm)
4
Form factors
pcieworkstation
VRAM (GB)
48
Memory type
gddr7 with ECC
Memory bandwidth (GB/s)
1,344
Bandwidth basis
NVIDIA product-page spec table prints 'Memory Bandwidth 1,344 GB/sec' on the shared row covering both the 48 GB and the 72 GB columns (source nv-rtx-pro-5000-blackwell-product-page, retrieved 2026-10-15). The linked datasheet PDF is image-compressed and was not text-extracted, so the bus width is NOT stored: Hardware Corner (Allan Witt, 2025-05-11) states a 384-bit interface, and 28 Gbps GDDR7 x 384 / 8 = 1,344 GB/s is arithmetically consistent, but that bus width is a secondary source and is left out of the field.
TDP (W)
300
Unified memory
no
Consumer part
no
notes
First-party spec table (nv-rtx-pro-5000-blackwell-product-page) has two memory columns: this record is the 48 GB GDDR7-with-ECC column. Shared printed figures: Blackwell, AI TOPS 2064 (sparsity footnote not printed — see flops note), 1,344 GB/s, 300 W, 4x DisplayPort 2.1, 4.4in H x 10.5in L dual-slot, active cooler. The 72 GB column is a separate accelerator id (nvidia-rtx-pro-5000-72gb-blackwell) because capacity changes which models fit; bandwidth and TDP are printed once and shared. CUDA core count and memory bus width are NOT stored: the datasheet PDF did not yield a text layer, and the 14,080 cores / 384-bit figures circulating in Hardware Corner's May 2025 analysis are secondary. process_nm 4 is the Blackwell node used by the sibling workstation records, not a number printed on this page. Measured LLM numbers live on the Hardware Corner March 2026 benchmark page, not here.

Notes

First-party spec table (nv-rtx-pro-5000-blackwell-product-page) has two memory columns: this record is the 48 GB GDDR7-with-ECC column. Shared printed figures: Blackwell, AI TOPS 2064 (sparsity footnote not printed — see flops note), 1,344 GB/s, 300 W, 4x DisplayPort 2.1, 4.4in H x 10.5in L dual-slot, active cooler. The 72 GB column is a separate accelerator id (nvidia-rtx-pro-5000-72gb-blackwell) because capacity changes which models fit; bandwidth and TDP are printed once and shared. CUDA core count and memory bus width are NOT stored: the datasheet PDF did not yield a text layer, and the 14,080 cores / 384-bit figures circulating in Hardware Corner's May 2025 analysis are secondary. process_nm 4 is the Blackwell node used by the sibling workstation records, not a number printed on this page. Measured LLM numbers live on the Hardware Corner March 2026 benchmark page, not here.

Cross-references

Interconnect (1)