Microsoft Expands Partnership with Intel; Azure’s In-House AI Chips to Use EMIB Packaging
Microsoft and Intel have entered into a long-term partnership under which future generations of Azure’s in-house AI chips will use Intel’s EMIB advanced packaging technology, while their Xeon processor supply partnership will continue through Diamond Rapids. The two companies did not disclose the contract value, purchase volumes, or specific chip models involved, nor did they say whether existing packaging orders would be transferred. The partnership also covers CXL memory efficiency, accelerated SQL database backups, and security for cloud workload migration.
Why read this
AI score 57
The partnership covers advanced packaging, server processors, and cloud infrastructure optimization, offering insight into how Microsoft is expanding the supply chain for its in-house chips.
Topics
Body · AI translation
Article Body EM_StockImg_Start
EM_StockImg_End
On Wednesday local time, cloud computing giant Microsoft and U.S. chipmaker Intel jointly announced a long-term partnership agreement to further expand their cooperation on cloud computing infrastructure.
(Source: Microsoft’s official website)
Overall, the agreement has two main highlights: Intel’s advanced packaging business will receive Microsoft’s backing, while a long-term supply arrangement will extend the two companies’ server processor partnership to next-generation products. However, the announcement did not disclose the contract value, procurement volume, or specific revenue contribution.
The most notable aspect of the partnership is Microsoft’s adoption of Intel’s EMIB advanced packaging technology for its in-house Azure chips.
Put simply, advanced packaging connects multiple chip units responsible for different tasks so that they can work together efficiently. As artificial intelligence systems become increasingly complex, chip performance depends not only on the power of individual computing units, but also on whether connections between different chips are fast enough.
Intel’s EMIB advanced packaging has previously been used in its own server processors and GPUs, as well as Altera’s programmable chips. Microsoft’s explicit announcement that it will use the technology for its in-house Azure chips marks new progress in expanding its adoption among major cloud computing customers.
Publicly available information shows that Microsoft’s earlier in-house Maia 100 and Maia 200 AI chips both used TSMC manufacturing processes and CoWoS-S advanced packaging technology. The Maia 300, which Microsoft released last month, uses a multi-chiplet design, with its compute chip manufactured using TSMC’s N2P process.
The announced adoption of Intel’s EMIB technology means Microsoft is further broadening its advanced packaging options for in-house chips, while also giving Intel the backing of a major customer as it competes in this market. It should be noted that the two companies did not disclose which specific chip models are covered by the EMIB partnership, nor did they say whether it involves transferring existing packaging orders.
In addition to advanced packaging, the two companies also confirmed multiyear Xeon processor supply capacity and will extend their partnership to the next-generation Xeon server CPU code-named Diamond Rapids.
Naga Chandrasekaran, Intel executive vice president, chief technology and operations officer, and general manager of Intel Foundry, said in the announcement:
“Microsoft and Intel share a clear vision for the direction of AI infrastructure: systems are becoming increasingly complex and must be designed through coordinated optimization across silicon, packaging, compute, memory, and software. By using our EMIB advanced packaging capabilities to support the scaling of future generations of Azure infrastructure, Intel and Microsoft can combine the flexibility to integrate different types of chips with the high-volume manufacturing capabilities required in the AI era, delivering greater performance and cost efficiency to Azure customers.”
The two companies will also deepen cooperation on memory efficiency, database backups, and data security. These projects share the goal of enabling cloud servers to process workloads more efficiently while strengthening protection for sensitive data.
On memory, the companies will use a high-speed interconnect technology called CXL in conjunction with Intel processors to improve server memory expansion and utilization efficiency. According to the announcement, Azure M-series v4 virtual machines using the relevant technology delivered performance of up to 2.3 times that of the comparison configuration in certain tests.
On databases, Microsoft will use Intel’s hardware acceleration capabilities to increase backup speeds. According to the announcement, SQL database backup speeds in relevant tests were up to 2.5 times those of the comparison platform.
The security partnership will focus on migrating cloud workloads. When a virtual machine carrying sensitive data is transferred from one server to another, the relevant technology can encrypt its memory and processor state, protecting the migration process.
(Article source: CLS) Middle of Article Embedded Advertisement
AI Trends