Earlier last week, a report over Intel’s 18A process node drew attention to its allegedly low 10% yields, which sounds bad, especially when viewed without the right context. Well, Intel’s ex-CEO Pat Gelsinger has taken to Twitter to clear the air hovering over the issue and set the record straight, stating that the report was “fake news”.
Gelsinger took to X to explain that looking at the 18A process’ in terms of yields wasn’t appropriate. Not unless you were talking about a specific die. “Anyone using % yield as a metric for semiconductor health without defining die size, doesn’t understand semiconductor yield,” Gelsinger said. “Yields are represented as defect densities.”
speaking about yield as a % isn't appropriate. large die will have lower yield, smaller die – high yield percentage. Anyone using % yield as a metric for semiconductor health without defining die size, doesn't understand semiconductor yield. yields are represented as defect…
— Pat Gelsinger (@PGelsinger) December 7, 2024
Defect density is a metric used to measure the quality of software or products by quantifying the number of defects relative to the size of the product. Basically, the fewer defects within the area, the better. As for the ones still insisting on using chip yield over defect density, the good Dr. Ian Cutress showed that the chip yield of 18A was currently at 99%.
To put that in terms of defect density, 18A was achieving better than 0.4 defects per centimetre square (cm2). Again, according to Cutress, that’s only an 8% yield if the die is reticle-sized. For something smaller, such as chips within a laptop or a smartphone, the yields are at a greater 65%.
Intel test chip yield on 18A is 99%. 🤷♂️ https://t.co/uWAhWJq7OF pic.twitter.com/7Y8zCMkIV1
— 𝐷𝑟. 𝐼𝑎𝑛 𝐶𝑢𝑡𝑟𝑒𝑠𝑠 (@IanCutress) December 6, 2024
Since his “resignation”, Gelsinger has clearly become more vocal and it’s easy to forget that the man was essentially a chip engineer before he was donned the mantle of CEO at Intel. Now that he’s no longer on the blue throne, it wouldn’t be wrong to reasonably expect the man to start dropping fact bombs on some industrial myths.
(Source: X [1] [2], Hot Hardware, Yahoo!)
