NVIDIA RTX PRO Blackwell Performance Delivering Excellent Linux Performance

Written by Michael Larabel in Graphics Cards on 21 May 2026 at 11:43 AM EDT. Page 3 of 9. 33 Comments.

👁 Llama.cpp benchmark with settings of Backend: GPU, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Text Generation 128. RTX PRO 6000 Blackwell was the fastest.

A variety of models were tested with Llama.cpp using each GPU vendor's native back-end of NVIDIA CUDA / Intel oneAPI / AMD ROCm HIP (Vulkan Llama.cpp benchmarks follow). Both large and smaller models were tested for being able to retain coverage going down to the smaller graphics cards tested with their limited memory capacities.

👁 Llama.cpp benchmark with settings of Backend: GPU, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Prompt Processing 512. RTX PRO 6000 Blackwell was the fastest.

The NVIDIA RTX PRO Blackwell graphics cards were delivering phenomenal performance both over the prior RTX PRO Ada Generation cards and against the Intel Arc Pro and AMD Radeon PRO offerings.

👁 Llama.cpp benchmark with settings of Backend: GPU, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Prompt Processing 2048. RTX PRO 6000 Blackwell was the fastest.

The NVIDIA RTX PRO 5000 Blackwell was outpacing the prior-gen flagship of the RTX 6000 Ada Generation. In fact, the RTX PRO 4500 Blackwell was even coming in close to the RTX 6000 Ada Generation performance in some of these AI benchmarks.

👁 Llama.cpp benchmark with settings of Backend: GPU, Model: Mistral-7B-Instruct-v0.3-Q8_0, Test: Text Generation 128. RTX PRO 6000 Blackwell was the fastest.

👁 Llama.cpp benchmark with settings of Backend: GPU, Model: Mistral-7B-Instruct-v0.3-Q8_0, Test: Prompt Processing 512. RTX PRO 6000 Blackwell was the fastest.

Meanwhile AMD's newest Radeon AI PRO R9700 performance was only competing with the RTX PRO 4000 Blackwell in these Llama.cpp benchmarks.

👁 Llama.cpp benchmark with settings of Backend: GPU, Model: Mistral-7B-Instruct-v0.3-Q8_0, Test: Prompt Processing 2048. RTX PRO 6000 Blackwell was the fastest.

Intel's new Arc Pro B70 graphics card was coming up short of the RTX PRO 2000 Blackwell in these benchmarks.

👁 Llama.cpp benchmark with settings of Backend: GPU, Model: gpt-oss-20b-Q8_0, Test: Text Generation 128. RTX PRO 6000 Blackwell was the fastest.

👁 Llama.cpp benchmark with settings of GPU Power Consumption Monitor. RTX 2000 Ada was the fastest.

Simply put, the NVIDIA RTX PRO Blackwell graphics cards were delivering incredible performance across the board with very nice generational gains each step of the way and coming well past where the AMD Radeon and Intel Arc Pro graphics cards can currently compete.

👁 Llama.cpp benchmark with settings of Backend: GPU, Model: gpt-oss-20b-Q8_0, Test: Prompt Processing 512. RTX PRO 6000 Blackwell was the fastest.

👁 Llama.cpp benchmark with settings of Backend: GPU, Model: gpt-oss-20b-Q8_0, Test: Prompt Processing 2048. RTX PRO 6000 Blackwell was the fastest.

👁 Llama.cpp benchmark with settings of GPU Power Consumption Monitor. RTX 2000 Ada was the fastest.

With every model tested under Llama.cpp, the NVIDIA RTX PRO Blackwell line-up was delivering incredible results.

33 Comments - Next Page

Page: 1 2 3 4 5 6 7 8 9 Next Page

URL: https://www.phoronix.com/review/nvidia-rtx-pro-blackwell/3

⇱ NVIDIA RTX PRO Blackwell Performance Delivering Excellent Linux Performance Review - Phoronix

NVIDIA RTX PRO Blackwell Performance Delivering Excellent Linux Performance