Intel pointed to 'capacity' as the main metric that determines internal or external production of Graphics Processing Unit (GPU). Production capacity means how high yields ensure excellent productivity.

Photo Image
<Intel's Senior Vice President Raja Koduri recently held a press conference about Intel's GPU and computing architecture for Asia Pacific media.>

Intel Senior Vice President Koduri said, "In the current situation, (GPU's) production capacity is very important. Unlike the central processing unit (CPU), GPU design was easy to easily take advantage of external technology and basically considered cost and production capacity together.” This is the first time Intel has explained the background since it recently announced that it would consign all production of its next-generation GPUs to TSMC in Taiwan. In August, Intel announced that it would consign the production of both the PC external GPU 'Ark' and the supercomputing GPU 'Ponte Vecchio' to TSMC, which was raised as putting orders to Samsung Electronics. This is interpreted as Intel highly regarded 5-7nm production capacity of TSMC, which Intel decided to consign next-generation GPU products.
 
Intel Arc is produced using TSMC's 6-nano process, and Ponte Vecchio is produced using 7- and 5-nano processes. It is proof that Intel recognized TSMC's 5-7nm process production capacity more than Samsung Electronics' foundry. Senior VP Koduri said, “We review internal and external production according to the design focus of each product. External GPUs are new to manufacturing line, so internal and external production capabilities need to be optimized.” Intel needs to use an external foundry such as TSMC in the short term before securing its own production capacity.

Photo Image
<Intel Senior Vice President Raja Koduri>

Intel is also considering the production of its own flagship products in the future. Citing an example of an external GPU for notebook computers that Intel first launched last year, Senior Vice President Koduri said, “The first external GPU, Xe Max, was produced internally based on Intel SuperFIN technology. We will carefully look at the Intel roadmap leading to Angstrom and review how we can use it.”
 
According to the process technology roadmap announced by Intel in July, Intel declared that it would open the era of 'Angstrom (0.1nm)' beyond the nano process, provide extreme ultraviolet (EUV) and ribbon FET (Gate All-Around: Intel’s name of GAA technology) by 2025, and improve the process using cutting-edge technologies. Once Intel is equipped with internal production capacity, it means that it will sufficiently be able to produce its own GPU using its advanced technology. Currently, it is known that Intel produced 80% of its total product quantity in-house.
 
Intel also announced its ambition to gain a significant share in the external GPU market within a few years, especially emphasizing that the platform that encompasses CPU and GPU is the core competitiveness of Intel to lead the market. “Intel has the ability to leverage both CPUs and GPUs to optimize overall platform performance,” said Roger Chandler, Intel Vice President and General Manager of Client Graphics Products and Solutions. “First, we will focus on product positioning and build a brand and ecosystem step by step.” The high-performance computing architecture that Intel announced is also believed to have contributed to this confidence. Intel has unveiled various technologies for high-performance computing, such as Efficient and Performance Core that realize low power and high performance, and a Thread Director that allows each core to work efficiently within the Intel architecture.
 
“Intel’s core is product and technology leadership,” said Senior Vice President Koduri. "All the core technologies will come together to make Intel a great company advancing the world's technology," he said.
 
By Staff Reporter Dong-jun Kwon / djkwon@etnews.com