The next-generation GB200 and GX200 Data Center & AI GPU accelerators from NVIDIA have been formally confirmed for release in 2024 and 2025, respectively. NVIDIA announced its accelerated computing vision at a recent investor presentation earlier this month. The vision largely discusses how AI has altered the current data center and how NVIDIA is providing the hardware and software capabilities to power these new workloads.
With the advent of Blackwell in 2024, everything that NVIDIA has been relying on over the past two years to meet the needs of AI & HPC customers throughout the world will change. Previously, NVIDIA relied on its Hopper H100 and Ampere A100 GPUs. The popularity of AI helped NVIDIA significantly increase its data center and overall business revenue, and it appears that trend is continuing apace as the green team plans to introduce two completely new GPU families by 2025.
The Blackwell GPU family, named for David Harold Blackwell (1919–2010), will be the first of these new NVIDIA AI/HPC GPU families. The B100 chip will be used in the GPU, which will be the GH200 Hopper series’ replacement. The business intends to sell a range of goods, such as the GB200NVL (NVLINK), the regular GB200, and the B40 for visual compute acceleration. The next GTC (2024) is projected to see the unveiling of the next-gen range, followed by a launch later that year.
According to current speculations, NVIDIA will produce its Blackwell GPUs using the TSMC 3nm manufacturing node, and the first customers will receive the chips by the end of 2024 (Q4).
The GPU will compete with AMD’s Instinct MI300 accelerator, which the red team has predicted will also be a major player in the AI field, being the first HPC/AI accelerator to use a chiplet design.
The GX200, the successor to Blackwell with a launch date of 2025, is the other chip that has been made public. Since NVIDIA has traditionally released its AI and HPC products every two years, it’s possible that the chip won’t even be announced until 2025, with sales of actual goods starting in 2026.
The range will be built around the X100 GPU and feature the GX200 line of goods as well as an exclusive X40 line of products for enterprise clients. We can anticipate a different scientific name for the X100 series since NVIDIA is known for naming its GPUs after illustrious scientists and it already uses the Xavier codename for its Jetson series.
Also Read:
