We have the first glimpse of Nvidia’s upcoming Blackwell GPUs designed for enterprises. While not optimized for gaming like Cyberpunk 2077, this preview gives us a peek into Nvidia’s future plans for its RTX 50-series GPUs, set to be launched in the upcoming year.
The showcased GPU is the B200, or Blackwell 200, boasting a hefty 208 billion transistors. It is based on TSMC’s 4NP node, an advanced version of the 5nm node. Nvidia’s use of this custom process sets it apart from the Ada Lovelace GPUs, even though they are both related to TSMC technology.
Nvidia highlights six key features that define Blackwell. Among them is a custom 4NP process for connecting two GPU dies with a high-speed interconnect, an updated Transformer Engine, and fifth-generation NVLink for seamless GPU scaling in enterprise environments.
Enterprise customers benefit from AI security features and a dedicated decompression engine in the B200. The latter, especially, has potential implications for gaming performance if integrated into future gaming GPUs.
Get your weekly breakdown of PC gaming tech developments
Nvidia’s Blackwell gaming GPUs are anticipated by the end of this year or early next year. Currently, enterprise customers can access the B200 through the HGX B200 server board and the GB200, which combines the B200 GPU with Nvidia’s Grace CPU in the GB200 NVL72 configuration.
Editors’ Recommendations
