In announcing Intel's second quarter earnings, CEO Bob Swan piled on the praise (PDF), saying the results came in "well above expectations," and chest-thumped the company's "continued focus on innovation and execution." And indeed, Intel raked in $19.7 billion in revenue, a 20 percent year-over-year jump. But then came the news of a defect in Intel's 7-nanometer technology, which means another delay. Meanwhile, Intel's first 10nm desktop CPUs won't be here until the second half of next year. Oh, boy.
What the hell is going on, Intel? I pose this question only partially in a rhetorical manner, because this is not the same Intel that dominated the semiconductor space for so long, both financially and in technological design and manufacturing.
Granted, Intel is still making money hand over fist. But it's also given up ground to rival AMD, which is firing on all cylinders these days. AMD is already on 7nm with its latest generation Ryzen processors based on Zen 2, and later this year, it will launch its first Zen 3 chips built on an enhanced 7nm node, before transitioning to 5nm next year.
This requires a bit of a disclaimer. There does not exist a universal standard to labeling nodes, and the measurement only tells part of the story. Generally speaking, Intel's 10nm node is comparable to AMD's 7nm tech. So in that regard, Intel isn't as far behind as it may seem.
Nevertheless, the point is, AMD is now the one that is pushing the envelope and leading on process technology, rather than the other way around. Had things gone to plan for Intel, it would have been at 10nm on the desktop five years ago, with a dual-core version of Cannon Lake. Yes, in 2015! A series of delays ultimately derailed Cannon Lake altogether, and the only 10nm parts now and on the immediate horizon are laptop CPUs (Ice Lake currently, and Tiger Lake soon). Getting to 10nm on the desktop is still a year away.
"In the second half of 2021, Intel expects to deliver a new line of client CPU’s (code-named 'Alder Lake'), which will include its first 10nm-based desktop CPU, and a new 10nm-based server CPU (code-named 'Sapphire Rapids)," Intel states.
This opens up a whole other can of worms. Early indications (leaks and rumors) peg Alder Lake as adopting a design similar to ARM's big.LITTLE approach, which Intel is calling Hybrid Technology. This entails pairing high performance cores with lower performing, power efficient cores. It makes sense in mobile devices like smartphones and tablets where battery life is key, and maybe even in x86 laptops. But I'm not sold on the benefits of a hybrid x86 configuration on the desktop. We'll see.
Looking beyond 10nm, things with 7nm have not gone as smoothly as Intel had hoped.
"We are seeing an approximate six-month shift in our 7-nanometer based CPU product timing relative to prior expectations. The primary driver is the yield of our 7-nanometer process, which based on recent data, is now trending approximately 12 months behind our internal target. We have identified a defect mode in our 7-nanometer process that resulted in yield degradation," Swan said during an earnings call with investors (see the full transcript at Seeking Alpha).
So after delaying volume shipments of 10nm silicon for five years, Intel is now looking at a minimum one-year delay for 7nm, and that's assuming it actually has things sorted out. To that end, Swan added that Intel has identified the root cause of the defect and, as things stand, believes there are "no fundamental roadblocks" in the way.
"But we have also invested in contingency plans to hedge against further schedule uncertainty," Swan said.
This came with a eyebrow-raising caveat, in response to a question asking at what point Intel would consider outsourcing production to other foundries. After all, Taiwain Semiconductor Manufacturing Company (TSMC) will be on 3nm by the time Intel gets to 7nm.
"The extent that we need to use somebody else’s process technology and we call those contingency plans, we will be prepared to do that. And if we do, there’s lots of moving parts," Swan said.
Let me parse this for you. Plagued with one delay after another, things have gotten so out of control that Intel is actually considering tapping a third-party foundry to manufacturer its chips. So like AMD, Intel would design the silicon, and TSMC or Globalfoundries or Samsung would manufacturer them.
Even that wouldn't be easy, though. Sanford C. Bernstein analyst Stacy Rasgon told Yahoo Finance that TSMC is not an option "because it doesn't have the capacity." Rasgon also believes the situation is a hit on Intel's reputation as a chip leader.
"You didn't need to read any more. Whatever little credibility they had is out the window," Rasgon added.
Certainly none of this is ideal. Intel's saving grace is that it has been able to squeeze quite a bit out of its 14nm technology on the desktop. Despite falling behind in process nodes, Intel's 14nm Core i7 10900K is the best CPU for gaming, because clock for clock, it's still generally ahead of AMD.
However, in multi-threaded workloads, AMD offers the better value, with more cores and threads for the money. AMD is also the only game in town for PCI Express 4.0 support, which paves the way for using ultra-fast PCIe 4.0 SSDs sporting 5,000MB/s read and write speeds.
So again, this is all to say, what the hell is going on, Intel?