AMD releases a AI-optimized Helion rack version that is purely CPU and can be scaled up to 32,768 cores.
Ai Focus
AMD announced today local time the launch of a Helion rack solution for proxy-based artificial intelligence computing loads. This rack is designed based on OCP design specifications and the open platform concept of AMD Helios rack solutions. Partners and customers can flexibly combine a variety of CPU and GPU computing resources according to their own needs. The pure CPU version, based on the sixth-generation EPYC processor, can be scaled up to 32,768 cores. AMD plans to contribute the AMD Helion specifications to OCP, and states that it is working with ecosystem partners to develop the Helion rack, which is scheduled to be officially launched in November 2026.
IT News on October 12th: AMD announced today, local time, the launch of a Helion rack solution for agent-based artificial intelligence computing loads.
This rack is designed based on the OCP design specifications and the open platform concept of the AMD Helios rack-based solutions. Partners and customers can flexibly combine a variety of CPU and GPU computing resources according to their own needs. The pure CPU version, which is based on the sixth-generation EPYC (Xiaolong) processor, can be scaled up to 32,768 cores.
AMD plans to contribute AMD Helion specifications to OCP, making it available for further development and adoption by the community. AMD is currently working with ecosystem partners to develop Helion racks, with a formal launch scheduled for November 2026.
Follow CoinMeta official accounts to stay updated
Views 18
CoinMeta reminds readers to view blockchain rationally, stay aware of risks, and beware of virtual token issuance and speculation. All content on this site represents market information or related viewpoints only and does not constitute any form of investment advice. If you find sensitive content, please click“Report”，and we will handle it promptly。
No comments yet. Be the first!
Lyon: China's AI model accounts for approximately 65% of the token usage shareLyon stated that data from OpenRouter shows a rapid growth in demand for the AI model. As of the end of September, the weekly consumption exceeded 146 trillion, and the model's annual recurring revenue totaled over 1.5 billion US dollars. In September, Chinese models accounted for approximately 65% of the total usage and 30% of the expenditures, benefiting from a price discount of about 80% compared to their American counterparts.
AASTOCKS
·2026-10-13 00:42:163 
NEXTPredict Spent $12,000 on hedging in the prediction market for its conferenceThe organizer of NEXTPredict purchased a contract through Kalshi: If the cancellation rate of flights to New York's Kennedy International Airport on October 21st exceeds 50%, the contract will pay out approximately $3 million. The organizer stated that this transaction, undertaken by Susquehanna as a market maker, had a premium of $12,000, which corresponds to their total investment in the event. This marks the first time in the convention and exhibition industry that market forecasting has been used to hedge against conference risks.
Decrypt
·2026-10-13 00:42:151 
AASTOCKS
·2026-10-13 00:31:259 
CoinDesk
·2026-10-13 00:31:2410 
Marvell Launches the Second Generation CXL Memory Expander, Supporting Up to 9TB of DDR4 MemoryMarvell releases the second-generation Structera X 3406 and 3506 CXL memory expanders, claiming them to be the industry's first six-channel CXL 3.2 memory expanders. A single controller supports up to 9TB DDR4 or 6TB DDR5 of memory, with samples expected to be provided to customers in the first quarter of 2027. The company stated that the new products are designed to alleviate memory bottlenecks in AI inference workloads and have gained support from companies such as AMD, Arm, Samsung, and SK Hynix.
Wallstreetcn
·2026-10-13 00:31:2110 View More
