New Way Now: General Motors drives the future of autonomous vehicles with Google Cloud and NVIDIA
지금 보고 있는 영상
New Way Now: General Motors drives the future of autonomous vehicles with Google Cloud and NVIDIA
조회수 1,431
영상 설명 · 눌러서 펼치기
Featured in this video: Sony Mohapatra, Director of AV ML Infrastructure, General Motors
Executive summary: General Motors (GM) is partnering with Google Cloud and NVIDIA to build the next-generation infrastructure required to transition from "hands-free" to "eyes-off" autonomous driving by 2028. By migrating to NVIDIA G4 GPUs on Google Cloud and optimizing data architecture, GM achieved up to 8x faster runtimes and a 38% reduction in compute time for production workloads. This scalable infrastructure already supports Super Cruise customers, accelerating GM's roadmap to bring mainstream autonomy to millions.
Challenge: GM aims to accelerate autonomous vehicle (AV) development to achieve "eyes-off" driving capability by 2028. This ambitious goal raises the bar for GM’s infrastructure, demanding the ability to process massive amounts of data, train complex models, and run large-scale simulations. However, upgrading to new GPU technologies presented complex software dependencies, which threatened to slow down iteration speed.
Solution: To modernize its simulation infrastructure, GM transitioned from T4 and L4 GPUs to high-performance NVIDIA RTX PRO 6000 Blackwell GPUs on Google Cloud. GM collaborated closely with Google Cloud and NVIDIA early in the process to ensure a smooth transition to the new system. The team also implemented strategic architectural optimizations, such as preloading data rather than streaming it, to significantly lower GPU idle cycles and improve overall hardware utilization.
Result: Moving to NVIDIA RTX PRO 6000 Blackwell GPUs on Google Cloud allowed GM to achieve an 8x higher requests per second (RPS), an 8x faster runtime, and a 4x faster time to result for inference throughput. In large-scale simulations, GM saw a 38% reduction in compute time for production-like workloads and a 20% overall reduction in runtime.
Highlights and key takeaways from our interview with Sony Mohapatra, Director of AV ML Infrastructure at General Motors:
→ “Moving to G4s has really changed the way we scale our simulations. Just looking at throughput, we saw that there was a 20% reduction in runtime, and when we ran production-like workloads, we saw there was a 38% reduction in compute time.”
→ “The infrastructure that we are building with Google Cloud and NVIDIA takes us from hands free to eyes off, giving people back their time, and that's the real payoff.”
→ “With Google Cloud and NVIDIA, General Motors is building a new way to drive autonomy to millions.”
Google Cloud products used: G4 VMs with NVIDIA RTX PRO 6000 Blackwell
Learn more about how Google Cloud’s partner network can unlock value for your business → tinyurl.com/32bdu6w6
Executive summary: General Motors (GM) is partnering with Google Cloud and NVIDIA to build the next-generation infrastructure required to transition from "hands-free" to "eyes-off" autonomous driving by 2028. By migrating to NVIDIA G4 GPUs on Google Cloud and optimizing data architecture, GM achieved up to 8x faster runtimes and a 38% reduction in compute time for production workloads. This scalable infrastructure already supports Super Cruise customers, accelerating GM's roadmap to bring mainstream autonomy to millions.
Challenge: GM aims to accelerate autonomous vehicle (AV) development to achieve "eyes-off" driving capability by 2028. This ambitious goal raises the bar for GM’s infrastructure, demanding the ability to process massive amounts of data, train complex models, and run large-scale simulations. However, upgrading to new GPU technologies presented complex software dependencies, which threatened to slow down iteration speed.
Solution: To modernize its simulation infrastructure, GM transitioned from T4 and L4 GPUs to high-performance NVIDIA RTX PRO 6000 Blackwell GPUs on Google Cloud. GM collaborated closely with Google Cloud and NVIDIA early in the process to ensure a smooth transition to the new system. The team also implemented strategic architectural optimizations, such as preloading data rather than streaming it, to significantly lower GPU idle cycles and improve overall hardware utilization.
Result: Moving to NVIDIA RTX PRO 6000 Blackwell GPUs on Google Cloud allowed GM to achieve an 8x higher requests per second (RPS), an 8x faster runtime, and a 4x faster time to result for inference throughput. In large-scale simulations, GM saw a 38% reduction in compute time for production-like workloads and a 20% overall reduction in runtime.
Highlights and key takeaways from our interview with Sony Mohapatra, Director of AV ML Infrastructure at General Motors:
→ “Moving to G4s has really changed the way we scale our simulations. Just looking at throughput, we saw that there was a 20% reduction in runtime, and when we ran production-like workloads, we saw there was a 38% reduction in compute time.”
→ “The infrastructure that we are building with Google Cloud and NVIDIA takes us from hands free to eyes off, giving people back their time, and that's the real payoff.”
→ “With Google Cloud and NVIDIA, General Motors is building a new way to drive autonomy to millions.”
Google Cloud products used: G4 VMs with NVIDIA RTX PRO 6000 Blackwell
Learn more about how Google Cloud’s partner network can unlock value for your business → tinyurl.com/32bdu6w6
Google Cloud 다른 영상
It’s alive! Turn your chatbot into an Agent
조회수 342
Make sure your app data doesn't ghost you 💀
조회수 241
Take your local web app public instantly
조회수 191
Is your live app actually working? 😳
조회수 60
Code entirely in your browser window 💻
조회수 607
Build a Halloween app in minutes! 🎃
조회수 939
KeyBank improves infra provisioning by 90% with Google Cloud; shrinks manual coding work with Gemini
조회수 1,421
DOE’s National Laboratory of the Rockies accelerates materials discovery with Google AI
조회수 1,247