- Nvidia reveals Blackwell Ultra at GTC 2025
- The new flagship GPU offers more power and efficiency than previous generations
- There is also a new NVIDIA RTX Pro 6000 Blackwell Server edition designed for corporate workloads
Nvidia has removed the wraps of its latest flagship GPU material from Blackwell Ultra while trying to assert its place as the world leader in AI computer science.
Revalled during its NVIDIA GTC event in San Jose, the new ultra blackwell equipment offers more power and efficiency in the data center than previous generations while the company seeks to establish itself as the backbone of development and future deployment of AI.
Declaring that it is “built for the age of AI reasoning”, Nvidia says that Blackwell Ultra can help democratize the adoption of AI around the world, which allows more organizations to take advantage of the advantages that this calculation can bring.
NVIDIA Blackwell Ultra
Nvidia noted that the rapid growth in use of AI in the world in recent years has resulted in huge demand for calculations, because companies and consumers slam as more.
“The AI has made a giant jump – the reasoning and the AI application agency of the request for IT performance,” said Jensen Huang, founder and CEO of Nvidia.
“We have designed Blackwell Ultra for this moment-it is a single versatile platform that can easily and effectively make the inference of pre-training, post-training and reasoning.”
The increase in reasoning models in particular has led to a boom requirements for a complete offer which is profitable but which also does the work.
Built on the first Blackwell architecture unveiled at GTC 2024, Blackwell ultra will offer 1.5x more FP4 inference and should be available India built by Nvidia Partners in the second half of 2025.
It will be present in a multitude of new NVIDIA offers, including the improved GB300 NVL72 Rack, which, according to him, offers an improved energy efficiency and ease of service, with bandwidth speeds up to 130 TB / s.
NVIDIA RTX PRO 6000 Blackwell Server Edition
However, it was not everything about New Blackwell Hardware, because the company also revealed the Nvidia RTX Pro 6000 Blackwell Server Edition, designed for corporate workloads such as IA multimodal inference, the creation of immersive content and scientific IT.
NVIDIA says the new version offers huge advances on the ADA Lovelace Architecture L40S L40S GPU of previous generation, offering up to 5x higher LLM inference flow for AI original applications, almost 7x faster genomic sequencing, 3.3x exhibits for the generation of empty text, almost 2x faster for recommendation systems and more 2X networks Rendering.
Each RTX PRO 6000 can also be partitioned in up to four fully isolated instances with 24 GB each to execute simultaneous workloads of AI and graphics, with 96 GB of ultra-fast GDDR7 memory and care for the multi-instance GPU.
They can also be configured in high -density accelerated IT platforms for distributed inference workloads – or used to provide virtual workstations using NVIDIA VGPU software, for high -intensity applications or to supply the development of AI.
“With the RTX Pro 6000 Blackwell Server edition, companies in various sectors – in particular architecture, automotive, cloud services, financial services, game development, health care, manufacturing, media and retail and retail – can allow revolutionary performance for workloads such as AI Multimodal generators, data analysis, Sandeep and Sandeep NVIDIA GUPTE.
