Source URL: https://tech.slashdot.org/story/24/11/29/1128242/ai-ambition-is-pushing-copper-to-its-breaking-point?utm_source=rss1.0mainlinkanon&utm_medium=feed
Source: Slashdot
Title: ‘AI Ambition is Pushing Copper To Its Breaking Point’
Feedly Summary:
AI Summary and Description: Yes
Summary: The text discusses the trend of increasing power demands in datacenters, driven mainly by the growing complexity of AI models. It highlights the shift towards direct liquid cooling and advanced interconnects like optics for data transfer, emphasizing the challenges faced by chip designers and the implications for infrastructure security and power efficiency.
Detailed Description:
The narrative describes several critical issues currently influencing data center infrastructure, particularly as it relates to the burgeoning demands of AI technologies. This is particularly relevant for professionals working in data center management, infrastructure security, and cloud computing.
– **Rising Power Demands**:
– Datacenters are shifting toward denser configurations.
– High-density racks now exceed power demands of 120 kilowatts.
– Direct liquid cooling is becoming a standard to manage heat generated by these systems.
– **Growth of AI Models**:
– AI models are getting significantly more complex, with an increase of parameters roughly 32-fold every three years according to Fujitsu research.
– This complexity drives advancements in computer hardware to keep up with operation needs.
– **Advanced Interconnects**:
– Chip designers, notably Nvidia, employ extremely high-speed interconnects of about 1.8 terabytes per second.
– This technology allows multiple GPUs (graphic processing units) to function cohesively as single entities, which is critical for AI processing tasks.
– **Signal Maintenance Issues**:
– As data transfer speeds increase, signal integrity over copper cables diminishes, limiting operational distances to one or two meters.
– The text indicates that optics could potentially enhance signal maintenance over longer distances, essential for robust data center operations.
– **Efficiency Challenges**:
– Current pluggable optics are inefficient and not fast enough to meet demand.
– Nvidia’s CEO, Jensen Huang, noted that an all-optical setup for their NVL72 system would lead to even higher power consumption requirements (an additional 20 kilowatts).
These points outline significant trends and challenges within data center infrastructure security and cloud computing, necessitating a deeper focus on efficient cooling systems, advanced material usage, and the intricacies of data transport technologies—all critical for maintaining operational integrity and efficiency in the age of AI. Additionally, the implications for power management, infrastructure investment, and long-term strategic planning are essential considerations for professionals in the security and compliance landscape.