Can AI Microprocessors be Sustainable? New NVIDIA Blackwell Chip Points to a Breakthrough for High Tech Industry
Author: Stephen Russell
Nvidia’s announcement of the anticipated Blackwell Hopper series of GPUs last month has tech leaders salivating to get their hands on these new chips that promise to deliver the processing power needed by AI, but now at a lower cost to manufacturer and lower electricity requirements to run. The claim that they will run with 25x less energy consumption sounds like a giant step forward for semiconductors becoming more efficient and sustainable.
Read Full Article Start My Free Trial
If only that was the only lever needing be pulled to reduce the carbon footprint of semiconductors. Other elements that need to be considered include manufacturing for these advanced nodes, the e-waste they will drive as older tech is retired, and even the energy costs of data centers that support them.
From a sustainability perspective, the improved performance metrics are impressive with several indicators towards a more sustainable technology, such as the use of a more advanced manufacturing node. However, with increased power consumption there are implications on electricity demand in data centers – an area where carbon impacts come to the fore.
Leveraging TechInsights’ sustainability tools with cradle-to-gate carbon footprint analysis along with unique insight in terms of the power consumption of a final product in operation, we get a more complete view on how much more sustainable the NVIDIA Blackwell platform might be compared to predecessors.
To see more on that breakdown, as well as additional information on how manufacturing impacts might counterbalance gains made in efficiency, read our full report in the Platform.
