Intel shows its first wafer made in 10 nm, will arrive first at fpga

Without a doubt Intel is at the forefront of silicon processing technology and has always been one of the strongest advocates of an industry-wide categorization both to avoid confusion and to put its process leadership in perspective. This is because not all silicon-based chipmakers use the same standards to measure the size of their transistors and play "dirty" to appear to be more advanced than they actually are.

Intel begins manufacturing with its 10nm Tri-Gate process

Intel's leadership extends to 10mm where they intend to improve transistor density 2.7-fold. The production of Intel chips at 10 nm is going to start in the sector of FPGAs that are the most suitable candidate due to their extremely redundant nature, in that a defect would not bring catastrophic problems with the affected chips, Intel can simply disable individual door arrays with defects to take advantage of. All manufacturing processes are immature in their beginnings, so they are not initially suitable for the manufacture of very complex monolithic chips in which the success rate would be too low.

That's the main reason Intel is putting its "Falcon Mesa" FPGA architecture to the test with the 10nm process. This enables the company to further fine-tune the 10nm production process with a relatively low-risk product, which is less sensitive to performance issues and defects, while optimizing for the manufacture of its most critical products., mainly CPUs. Mesa Falcon's “FPGA” design will also take advantage of Intel's EMIB packaging solution, where chip packaging is done with additional silicon substrates that allow for faster connection and data transfer between separate silicon blocks. This avoids the need for a full silicon interposer as AMD uses its Vega graphics cards, a more efficient, but much more expensive way to do this.

This means that Intel does not need to manufacture all the components of a chip in the same low-risk, high-performance 10nm process as they can use other process nodes at 14nm or even 22nm for parts that are not critical in terms energy-consuming or do not require state-of-the-art manufacturing.

Source: techpowerup