Source URL: https://tech.slashdot.org/story/25/03/18/201213/nvidia-reveals-next-gen-ai-chips-roadmap-through-2028?utm_source=rss1.0mainlinkanon&utm_medium=feed
Source: Slashdot
Title: Nvidia Reveals Next-Gen AI Chips, Roadmap Through 2028
Feedly Summary:
AI Summary and Description: Yes
Summary: Nvidia’s announcement of its new AI processors, the Blackwell Ultra chips, showcases significant advancements in AI performance and memory capabilities. With faster processing speeds, these chips are positioned to enhance AI reasoning tasks, marking a strategic shift in Nvidia’s product release strategy amidst growing demand in the AI sector.
Detailed Description: The recent unveiling by Nvidia at GTC highlights the company’s continued innovation in AI processors, as it introduces several new chips aimed at enhancing computational capabilities for AI applications. The significant points from this announcement are:
– **Blackwell Ultra Chips**:
– Set to ship in late 2025, these chips retain the 20 petaflops of AI performance seen in previous Blackwell models but offer increased memory capacity.
– Boosting memory from 192GB to 288GB of HBM3e allows for improved data handling, crucial for memory-intensive AI applications.
– **Processing Speed**:
– The Blackwell Ultra chips are capable of processing 1,000 tokens per second, a tenfold improvement compared to 2022’s hardware.
– This increases the efficiency of AI reasoning tasks, significantly reducing response times from 1.5 minutes on H100 chips to just 10 seconds on the new models.
– **Future Innovations**:
– The upcoming Vera Rubin architecture, expected in 2026, aims to reach 50 petaflops and will include Nvidia’s first custom Arm-based CPU design, named Olympus.
– A novel approach to GPU design is noted with Rubin, which integrates two dies functioning as a single chip, suggesting a potential shift in how GPUs are architected and counted.
– **Strategic Shift**:
– Nvidia’s new annual release cadence contrasts with its previous model of biennial releases, indicating a responsive strategy to the burgeoning AI market and its demands.
This announcement represents a key development for professionals in AI and infrastructure security as these advancements can contribute to more sophisticated AI applications, necessitating a reassessment of security frameworks, performance expectations, and compliance measures in deploying AI solutions. This is particularly relevant in environments leveraging cloud infrastructure, where improved processing capabilities can enhance security analytics and AI-driven threat detection.