Last week after receiving the Intel Arc Pro B70 review hardware I began with some benchmarks looking at how the Arc Pro B70 compared to existing Intel GPUs on Linux with their fully open-source driver stack. Today’s article features the latest Arc Pro B70 benchmarks under Linux in looking at how the performance and value compares to other NVIDIA RTX and AMD Radeon (AI) PRO workstation graphics cards in the lab.
We know the gains in performance on the Intel Arc graphics side with the Arc Pro B70 from last week’s comparison while this article is seeing how the Arc Pro B70 stacks up against the competition. All of the graphics cards for this article were tested on Ubuntu 26.04 with the Linux 7.0 kernel. For the graphics tests this meant Mesa 26.0 for the Intel Arc Pro and AMD Radeon (AI) PRO while on the NVIDIA side was their 595.58.03 driver stack. The Intel Compute Runtime 26.09.37435.1 was used for the Intel compute stack as the latest release there while for the AMD compute tests was ROCm 7.2.1 and with NVIDIA was CUDA 13.2.
For comparing against the Intel Arc Pro B70 on the AMD side was the Radeon AI PRO R9700 that features 32GB of vRAM like the B70 while retailing for $1379 USD, compared to the Arc Pro B70 being available at online retailers for as little as $949 USD. The prior-generation Radeon PRO W7900 was also included with these benchmarks for that workstation graphics card featuring 48GB of RAM. Radeon PRO W7900 48GB retail listings still put that card for around $3999 USD.
On the NVIDIA side, unfortunately, I have yet to receive any review samples of the RTX PRO Blackwell graphics cards. Thus for this comparison I am unfortunately only able to test against the RTX Ada Generation graphics cards. On that side there is the RTX 4000 Ada Generation that retails for around $1899 USD and has 20GB of video memory. Additionally there was the RTX 6000 Ada Generation with its 48GB of video memory but retailing for $7499 USD. The comparison is unfortunately limited based on the review samples I have available.
With those details out of the way, let’s get straight to the benchmarking.
