NVIDIA announced on Wednesday that it will launch the RTX PRO 6000 Blackwell Server Edition, a Blackwell-based data center GPU, in May at the GTC in San Jose, California.
The RTX PRO 6000 Blackwell Server Edition is built exclusively for visual computing and enterprise artificial intelligence (AI). As part of the RTX PRO Blackwell series, it delivers high-performance AI and graphics application processing solutions.
A spokesperson for NVIDIA highlighted the impressive capabilities of the RTX PRO 6000. Compared to its predecessor, the Ada Lovelace architecture’s L40S GPU, the new model boasts up to 5 times the LLM inference throughput, a remarkable 7 times improvement in genomic sequencing analysis, 3.3 times better performance for text-to-video generation, and more than double the rendering speed.
The RTX PRO 6000 Blackwell Server Edition incorporates NVIDIA Confidential Computing to safeguard sensitive data. It has 96GB of GDDR7 memory and Multi-Instance GPU (MIG) functionality. With 24 GB of dedicated memory, each instance can simultaneously handle AI and graphics workloads.
According to NVIDIA, several companies have already leveraged the RTX PRO 6000 Blackwell Server Edition for various applications. Black Forest Labs will use it for their Flux AI image generator, while OTOY will implement it for real-time rendering with OctaneRender. KLA plans to utilize it for inference workloads in semiconductor wafer manufacturing processes, and Oxford Nanopore Technologies will employ it for genomic sequencing analysis.
The RTX PRO 6000 Blackwell Server Edition will be unveiled through global partners starting in May. Cloud service providers, including Amazon Web Services (AWS), Google Cloud, Microsoft Azure, and companies such as Cisco, Dell, HPE, and Lenovo, are expected to introduce related hardware solutions.