- Intel arc pro b50 doubles the XE nuclei over A50, offering a stronger performance
- Intel’s workstation GPU brings a 16 GB memory against the 8GB of NVIDIA
- DESIGN DECLARATIONS OF Cards 168 mm Low Profiles Compact Professional Work Station
The latest Intel work station graphics card, ARC Pro B50, has impressed how well compared to the NVIDIA RTX A1000.
Marketing as a professional input level GPU, the B50 is compactly but of ambitious scope, with characteristics that lead to a severe work station territory.
The AR B50 ARC is based on Intel’s XE2 architecture, the same base as BattleMage Consumer Line, which also feeds the GPUs on the laptops of the games.
A work station card with modest ambitions
This chip doubles the XE nuclei tracking units and the anterior ARC Pro A50 rays and is equipped with new XMX matrix engines.
ARC Pro B50 is executed in the BMG-G21 GPU, a smaller chip that intellly limited for costs.
The card offers 16 XE2 and 128 XMX engines, backed by 16 GB of GDDR6 memory in a 128 -bit interface, more than the 8 GB in the A1000.
This additional capacity provides a larger buffer for professional work loads such as CAD models, video editing or execution of small models of Locally.
With a bandwidth of 224 GB/s and 170 FP8 calculation tops, it is directed to efficiency instead of unprocessed power.
In this sense, the B50 leads the A1000, which runs at a lower bandwidth of 192GB/Sy offers only 107.8 tops in Int8 workloads.
Unlike consumption designs, the B50 comes in a low profile form that measures only 168 mm, cooled by a single radio fan.
Its 70W TDP means that it does not require auxiliary energy, and there are four mini displayport 2.1 outputs, marking an update on the oldest 1.4a connectors of the A1000.
These additions point to workloads such as representation, automatic learning inference and content creation.
Intel has also secured more than 50 certifications with the main software suppliers, a necessary step to compete in the work station market.
In reference points, Intel’s B50 surprised many observers. Geekbench AI showed it far ahead of NVIDIA through precision and half -precision tests.
Procyon AI’s reference points also highlighted faster results in computer vision and text generation.
The AR B50 also achieved a modest 7% advantage over the RTX A1000 in Photoshop, with even greater profits by isolating GPU tasks.
Premiere Pro showed similar results, and the B50 beat its rival in almost 20% in general.
The results after the effects were closer, although the B50 demonstrated clear forces in 2D workloads.
Perhaps the most surprising result came from the blender, since the representation has traditionally been an area of the Nvidia domain, but the B50 recorded a 20% advantage over the A1000.
This generational jump, more than 130% on the previous A50, shows that Intel has approached the previous deficiencies.
The card also stood out at the MLPERF reference points, generating tokens almost twice as quickly as the Nvidia entry level professional option.
However, in areas such as Revit and Inventor, the B50 often followed the RTX A1000, sometimes by broad margins.
In the inventor graphics test, the NVIDIA card was executed more than four times faster. However, in SolidWorks, the B50 reversed the result, publishing a 33%advantage.
The Intel Arc Pro B50 is not the best GPU for each task, nor will it move to NVIDIA in all markets, but for a card sold at $ 350, it offers great value for professional users.
Through Puget Systems / Hardwareluxx
