AMD tips manufacturing in data center announcements



AMD this week announced a flurry of new products in what is best described as silicon beasts for the data center. With the announcement of new data center processors and a GPU for accelerated computing, each takes single chip / socket performance to new levels and demonstrates the impact of product design for computing applications. high performance computing (HPC). While the performance expectations for new products are impressive, what struck me was how AMD is taking advantage of the latest semiconductor packaging and manufacturing technologies 12 years after its split. own manufacturing group to form GlobalFoundries (GF).

The first announcement was for a new Epyc server processor called Milan-X. Milan-X shares the same Zen 3 architecture as existing Epyc processors with up to 64 processor cores, but Milan-X adds a stacked memory array on top of the chip, which AMD calls V-cache. The V-cache processor and arrays will be fabricated using TSMC’s 7nm process node. The result is a 300% increase in processor cache for a total of 804MB cache per socket. The V-cache uses direct copper connections without microbosses for increased interconnect density and energy efficiency. The new parts are compatible with existing Epyc Gen 3 processors with just a minor BIOS upgrade. According to AMD, the result can be a significant increase in performance for technical workloads, such as 66% faster semiconductor RTL verification. Microsoft’s Azure uses Milan-X in its HBv3 platform and has reported increases in workload performance ranging from 50% to 80%. The Milan-X will generally be available in the first quarter of 2022.

AMD also introduced the first product to use its next-generation GPU architecture for computing – CDNA 2. The new product is the Instinct MI200 OAM, the second generation of the GPU family of products. The MI200 combines two of the new CDNA 2 GPU cores into a single package, resulting in 58 billion transistors in a single chip. It includes 220 compute units, 880 matrix cores and 128 GB of HBM2E (high bandwidth stacked) memory. The MI200 GPU will also take advantage of a 3rd generation of AMD Infinity fabric capable of supporting up to 3.2TB of total memory bandwidth. The chip is made using TSMC’s 6nm process node and will consume up to 550W but, incredibly, it’s packaged in an air-cooled module, although liquid-cooled options are available. The module is available today and a PCIe card version will be available in the future. The MI200 was designed in cooperation with the Oak Ridge National Laboratory for the Frontier supercomputer, and Frontier assembly has already started. If completed on time, Frontier should be America’s first Exaflop supercomputer.

The company also provided an overview of the 4e generation of Epyc processors with two new processors – Genoa and Bergamo. Genoa will be the standard processor which will include up to 96 of the next generation Zen 4 processor cores, as well as support for DDR5 memory and CXL and PCI Gen 5 interconnect technologies. Bergamo will offer up to 128 cores Zen 4c processors for cloud applications, including a new cache hierarchy and greater power efficiency. Both products will be manufactured on TSMC’s 5nm process node and will be socket and software compatible. Genoa is sampling its customers today and will be in production in 2022. Bergamo production is scheduled for 2023.

What’s even more impressive to me than the products is how AMD has come to be at the forefront of semiconductor technology. While the company has always invested in new semiconductor technology, AMD has often fallen behind its competitors in manufacturing processes and even spun its manufacturing group in 2009. However, since the development of the architecture Zen, AMD has continued to take advantage of advanced semiconductor technology, processes, and packaging. With the announcement of the Instinct MI200, the company highlighted the use of a new Elevated Deployment Bridge (EFB) to reduce the cost and complexity of building chips with multiple chips, called multi-chip modules ( MCM), chiplets or 2.5D packaging. In addition to being on the cutting edge processing technology of TSMC, AMD has become a leader in semiconductor technology. As we have seen in the past, having a leading product architecture and a manufacturing advantage is a combination of being successful in the technology segment.

Source link


Comments are closed.