Google has launched a private preview of its next-generation A3 GPU supercomputer. Google Compute Engine A3 supercomputers uses Nvidia’s H100 Tensor Core GPUs, and are purpose-built to train and serve AI models for generative AI and large language model innovation, according to the company.
A3 uses Google’s custom-designed 200 Gbps IPUs, with GPU-to-GPU data transfers bypassing the CPU host and flowing over separate interfaces from other VM networks and data traffic. Its intelligent Jupiter data center networking fabric scales to tens of thousands of interconnected GPUs and allows for full-bandwidth reconfigurable optical links that can adjust the topology on demand, says Google.
The A3 supercomputer’s scale provides up to 26 exaFlops of AI performance.
Key features of Google’s A3 supercomputers:
- 8 H100 GPUs utilizing Nvidia’s Hopper architecture
- 3.6 TB/s bisectional bandwidth between A3’s 8 GPUs via Nvidia NVSwitch and NVLink 4.0
- 4th Gen Intel Xeon Scalable processors
- 2TB of host memory via 4800 MHz DDR5 DIMMs
- 10x greater networking bandwidth powered by hardware-enabled IPUs, specialized inter-server GPU communication stack and NCCL optimizations
You can sign up for early access to the A3 here.
[Image courtesy: Nvidia]