started · updated
AMD and Intel unveil new AI infrastructure and rack designs
At the Open Compute Project Global Summit, major semiconductor companies unveiled new hardware architectures specifically designed to support agentic AI and large-scale inference workloads.
AMD introduced Helion, an open rack-scale blueprint. The system offers two primary configurations: a hyperscale CPU-only version supporting up to 32,768 Zen 6 EPYC cores, and an enterprise version that integrates EPYC nodes with up to 16 AMD Instinct MI350P GPUs. Helion includes an integrated token router to manage agent model calls and supports both air and liquid cooling.
Intel announced its GPU D-series, code-named Crescent Island, aimed at memory-intensive AI inference. The flagship D270 model features up to 480 GB of LPDDR5X memory and provides 1.3 petaflops of FP8 compute. Intel also presented new Xeon-based rack designs that can scale to over 46,000 CPU cores per rack, utilizing the Open Rack v3 standard to facilitate easier data center integration.
Additionally, AMD launched the Ryzen AI Halo developer platform, which utilizes the Ryzen AI Max+ PRO 495 processor. This platform provides up to 192 GB of unified memory, enabling the local execution of large language models with up to 300 billion parameters.
NVIDIA also contributed to the shift toward open standards by releasing its Vera Rubin rack and compute tray designs through the NVIDIA MGX architecture, aiming to standardize AI factory infrastructure.
Entities
AMD · Arm · Dan McNamara · Intel · MSI · Nvidia · Open Compute Project · Ryzen AI Halo · San Jose · TSMC
Claims
What the coverage asserts, and how many sources carry each claim.
- [● 3 SOURCES] The Ryzen AI Halo developer platform features the Ryzen AI Max+ PRO 495 processor and up to 192GB of unified memory. www.tokenpost.com · www.guru3d.com · time.news
- [● 5 SOURCES] The Helion CPU-only configuration can support up to 32,768 Zen 6 EPYC cores per rack. world-today-journal.com · enterados.pe · futurumgroup.com · www.nerdsheaven.de
- [● 5 SOURCES] Intel unveiled its GPU D-series, code-named Crescent Island, designed for AI inference and agentic AI workloads. technews.tw · www.igorslab.de · pulse2.com · www.pcgameshardware.de · www.tweakpc.de
- [○ 1 SOURCE] NVIDIA announced the opening of its Vera Rubin rack and compute tray designs at the OCP Global Summit. www.ithome.com
- [● 5 SOURCES] AMD introduced Helion, an open rack-scale blueprint for agentic AI infrastructure based on Open Compute Project specifications. world-today-journal.com · enterados.pe · futurumgroup.com · www.nerdsheaven.de
- [● 3 SOURCES] The Intel GPU D-series can feature up to 480 GB of LPDDR5X memory and provides 1.3 petaflops of FP8 compute. technews.tw · pulse2.com · www.tweakpc.de
- [● 2 SOURCES] The Enterprise configuration of Helion can integrate up to 16 AMD Instinct MI350P GPUs, delivering 74 petaflops of FP4 compute. futurumgroup.com · www.nerdsheaven.de
- [● 2 SOURCES] Intel's new Xeon-based rack designs can scale to over 46,000 CPU cores per rack. technews.tw · pulse2.com