Signal

A deep dive into AMD's Zen 6 EPYC Venice: it uses TSMC's N2P, the 9686F often beats Nvidia's Vera, a dual 9996 is up to 93% faster than a dual Zen 5 9965, more

First reported by Tomshardware ·

The signal ●●●● Compiled by AI from Tomshardware, Techmeme, Hacker News, The Register, Chips and Cheese and 13 more
Why you might care

If you run AI or HPC workloads, expect substantially faster and potentially more power-efficient compute options to become available.

What happened

AMD's upcoming Zen 6 EPYC processors, codenamed 'Venice,' are set to leverage TSMC's N2P manufacturing process. Benchmarks suggest the EPYC 9686F will often outperform Nvidia's Vera AI accelerator. Furthermore, a dual-configuration of the EPYC 9996 processor is reported to deliver up to a 93% performance increase compared to a dual-socket Zen 5 9965 setup. The 'Venice' architecture also introduces preliminary support for MEXT, with claims of achieving 95% of peak performance even with 50% less DRAM capacity. These advancements indicate a significant generational leap in processing power and efficiency for AMD's server offerings.

What it means

The adoption of TSMC's N2P process for AMD's Zen 6 EPYC 'Venice' CPUs signifies a move towards more advanced manufacturing nodes, enabling higher transistor densities and improved power efficiency. This technological leap is crucial for maintaining a competitive edge in the high-performance computing and AI markets, where power consumption and thermal management are critical factors for large-scale deployments. The competitive performance against Nvidia's Vera also signals an intensifying battle in the AI accelerator space, suggesting AMD is positioning its CPUs as viable alternatives for certain AI tasks.

The claimed 93% performance gain in dual-socket configurations and the preliminary MEXT support with reduced DRAM dependency point to architectural innovations aimed at boosting throughput and memory efficiency. This could have broad implications for data centers, potentially reducing infrastructure costs and allowing for denser compute solutions. As these processors approach release, the industry will be watching how AMD's pricing and availability strategies align with these significant performance improvements, especially in the context of an increasingly AI-centric market.

AI-written summary. May contain errors.

Chips