Samsung Announces New 11nm Process, Targets EUV Ramp in 2018

This site may earn affiliate commissions from the links on this page. Terms of use.

Samsung made a pair of announcements today about the future of its foundry manufacturing, and they imply things are going fairly well for the company. First, the company is announcing a new “11nm” process, which it claims will deliver “up to 15 percent higher performance and up to 10 percent chip area reduction with the same power consumption.” It’s also the first prime number we’ve used for a process node in a very long time; you have to go back to the 2um node for the last one. (This last bit isn’t exactly called out in Samsung’s marketing materials).

Practically, this appears to be a reaction to TSMC’s plans for a 12nm node (Samsung’s is better, see, because it’s a smaller number). TSMC’s “12nm” node is a respin of its 16nm, but with better performance and power consumption characteristics. Samsung’s 11nm node appears to be something similar. Adjustments like this, and the entire concept of slow, step-wise incremental improvements, are a further polishing of a technique TSMC began using nearly 20 years ago. Back in the early 2000s, the company began offering half-nodes that essentially split the typical “node” improvement in half. Instead of shifting from 180nm to 130nm, TSMC’s offered a 150nm half-node. Between 130nm and 90nm there was a 110nm node; between 90nm and 65nm, an 80nm node, between 65nm and 45nm, a 55nm, etc.

Samsung Fab Line

The term half-node has fallen out of favor as the foundries and fabs have begun using other ways to designate these improvements. Intel uses 14nm, 14nm+, and 14nm++, while Samsung has been using LPE (Low Power Early) and LPP (Low Power Plus). In both cases, the intent is to signal that improvements have been made to the process, and hopefully entice manufacturers that might not have been persuaded by the early improvements to 14nm compared with 28nm to step up to the later manufacturing process as the benefits become more substantial. 11nm will be ready in the first half of 2018 and is expected to complement Samsung’s 10nm process node. Samsung’s 11nm node also sounds quite similar to its 14nm LPU node, albeit with improved area.

EUV Supposedly Ramping Up

Initial EUV production is supposedly on-track for the second half of 2018 and the 7nm node, with initial production targeted for the back half of 2018. That’s the same thing Samsung has said previously and it comes with the same qualifiers — initial production doesn’t mean high volumes, and it doesn’t mean shipping product.

Samsung’s own communication on EUV is still shot through with qualifiers. The company states that: “Since 2014, Samsung has processed close to 200,000 wafers with EUV lithography technology and, building on its experience, has recently seen visible results in process development such as achieving 80 percent yield for 256 megabit (Mb) SRAM (static random-access memory).”

If you think about it, “processed,” is an odd verb to use compared with “manufactured.” To hit that impressive-looking 200,000 figure, Samsung appears to be counting any wafer it ever ran through a test machine, regardless of whether the final product was usable or what, exactly, was being tested. In a context this broad, “processed” is a meaningless metric. Hitting an 80-percent yield on 256Mb SRAM (32MB) is interesting, but we don’t know how long the company has been reaching that target, how regularly or easily it hits it, or how many wafers it can expose per hour at the 80-percent yield. Ramping up exposure power and reducing wafer exposure time have been twin anchors around EUV’s neck for the past few years.

We’ve discussed the manufacturing problems facing EUV and the slow ramp of progress at several points in the past, so check our previous coverage for more details. GlobalFoundries has said it plans to introduce EUV in high volume manufacturing at some point in 2019, and Samsung’s own timeline is loosely compatible with that as well. Don’t be surprised if this tech still slips a year or two; if companies are still predicting EUV will arrive by 2022 come 2020 it’ll just be a continuation of the trend these past 17 years.

Let’s block ads! (Why?)

ExtremeTechExtremeTech

Leave a Reply