NVIDIA RTX PRO 4500 Blackwell Workstation Edition datasheet - verbatim spec table (vendor re-fetch 2026-10-04)
Fields
sources/vb-nv-rtx-pro-4500-blackwell-datasheet-spec-tableNotes
VERIFY-ONLY re-fetch of the primary datasheet PDF (doc code 5108623, 'Apr 26'). Spec table transcribed character-for-character; nothing derived. SPEC TABLE VERBATIM: 'GPU architecture NVIDIA Blackwell' | 'NVIDIA CUDA Cores 10,496' | 'Tensor Cores 5th Generation' | 'Ray Tracing Cores 4th Generation' | 'AI TOPS (1,2) 1,617 AI TOPS' | 'Single-precision performance 51 TFLOPS' footnote 1 | 'RT Core performance 153 TFLOPS' footnote 1 | 'GPU memory 32 GB GDDR7 with ECC' | 'Memory interface 256-bit' | 'Memory bandwidth 896 GB/s' | 'System interface PCIe 5.0 x16' | 'Display connectors 4x DisplayPort 2.1b' | 'Video engines > 2x NVENC (9th Gen) > 2x NVDEC (6th Gen)' | 'Power consumption Total board power: 200 W' | 'Power connector 1x PCIe CEM5 16-pin' | 'Thermal solution Active' | 'Form factor 4.4in x 10.5in L, dual slot, full height' | 'Compute APIs CUDA 12.8'. FOOTNOTES VERBATIM: '1. Peak rates based on GPU Boost Clock.' '2. Effective FP4 TOPS with sparsity.' FLAG CHECK: identical convention to the RTX PRO 6000. 'AI TOPS 1,617' is the ONLY row carrying footnote 2, therefore 1,617 is the SPARSE FP4 figure and must be stored dense:false. 'Single-precision performance 51 TFLOPS' carries footnote 1 only and is dense. The datasheet prints no dense FP4 twin. TWO DIFFERENCES FROM THE RTX PRO 6000 WORTH RECORDING: (a) this datasheet's Compute APIs row reads 'CUDA 12.8, OpenCL 3.0, DirectCompute' and Graphics APIs reads 'DirectX 12, Shader Model 6.6, OpenGL 4.63, Vulkan 1.43', whereas the RTX PRO 6000 (Mar 25 doc) prints the older 'CUDA 11.6' and 'Vulkan 1.33' - so the 6000's API list is a stale row on that sheet, not a hardware fact; (b) this part has NO MIG row in the spec table at all, unlike the 6000's 'Up to 4x 24 GB / 2x 48 GB / 1x 96 GB'. NOT PUBLISHED: any FP16/BF16/TF32/FP8/INT8 figure, any boost clock, any NVLink figure, any GB/s host bandwidth. 'PCIe 5.0 x16' is the entire interconnect statement - scale-up on this part is PCIe only. | DIGIT CORRECTION 2026-10-04: this note previously read 'OpenGL 4.6, Vulkan 1.4' and 'Vulkan 1.3' - rounded. Re-extracted from the PDF text layer, the sheets print OpenGL 4.63, Vulkan 1.43 and Vulkan 1.33 exactly. Full side-by-side plus the Ada-template provenance for the stale row is in w3s-nv-rtx-pro-6000-and-4500-api-row-stale-in-6000-sheet.
Cross-references
This record references no other records.