Key Takeaways
- Google Cloud has launched G4 virtual machines powered by Nvidia’s RTX Pro 6000 GPUs for advanced AI applications.
- The G4 VMs provide up to nine times the throughput compared to previous models, enhancing performance for AI tasks, graphics, and robotics simulations.
- Nvidia Omniverse is now available on Google Cloud Marketplace, offering tools for digital twins and real-time simulations.
New G4 Virtual Machines for Enhanced AI Capability
Google Cloud has strengthened its collaboration with Nvidia by introducing a new range of G4 virtual machines (VMs), designed specifically for sophisticated AI applications in the enterprise sector. These VMs are powered by Nvidia’s RTX Pro 6000 Blackwell Server Edition GPUs, marking a significant advancement in cloud computing resources.
According to a blog post by Google Cloud’s Roy Kim and Dai Vu, the G4 VMs deliver a remarkable leap in performance, capable of achieving up to nine times the throughput of the previous G2 instances. This enhancement supports a variety of applications, including multimodal AI inference, photorealistic design and visualization, and complex robotics simulations through tools developed on Nvidia Omniverse.
The G4 VM configurations are versatile, offering options with one to eight RTX Pro Blackwell GPUs, providing a total of 768 GB of GDDR7 memory. Future updates will include fractional GPU options to accommodate a wider range of workloads. The GPUs feature advanced fifth-generation tensor cores that improve AI performance significantly and fourth-generation ray tracing (RT) cores that facilitate high-quality graphics and realistic simulations.
With these improvements, Google Cloud anticipates several benefits for customers, particularly in generative AI. The compute capabilities of the G4 VMs enable the fine-tuning and inference of large language models in real time, effectively supporting text-to-image applications and other AI functions.
Moreover, the G4 architecture allows for a more efficient use of resources. Each GPU can be split into four isolated instances, each with dedicated high-bandwidth memory and compute capabilities. This flexibility allows for the simultaneous execution of various smaller workloads on a single VM, enhancing productivity.
The Google Cloud team highlighted the advantages of making Nvidia Omniverse available on the Google Cloud Marketplace. With the G4 VMs equipped to handle the demanding demands of real-time rendering and accurate simulations, the combined offerings pave the way for creating and deploying digital twins in industrial applications.
This partnership solidifies Nvidia’s position in the market, offering an end-to-end platform based on Blackwell architecture. Both companies aim to provide users with a seamless experience that accelerates their workloads within a unified cloud ecosystem, facilitating the advancement of AI and robotics technologies globally.
The content above is a summary. For more details, see the source article.
