Things are heating up in the X86 processor market. After taking a backseat to Intel for more than a decade, AMD (together with TSMC) has edged itself into technological leadership.
Intel is “investing to recapture process leadership going forward”, said CEO Bob Swan on an earnings call with investors. The truth had to come out, eventually: Intel is no longer king of the hill in integrated circuit technology. Even if the company starts delivering on its promises and 10nm volume shipments will get going for real, AMD’s equivalent 7nm chips have been in production at TSMC for some time now.
Intel’s 10nm products were originally slated to move into production late 2015. When the chips never appeared on the market in 2016, the processor maker explained it was shuffling some parameters of Moore’s Law around. The additional development time is accompanied by a proportional increase in density and performance, Intel assured. In other words: there will be no net deviation from the historical trend.
But by its own admission, the company’s technological goals had been too ambitious. Aggressive scaling goals combined with implementation of several new technologies, such as cobalt interconnects and self-aligned quad-patterning lithography, proved extremely problematic to realize. Manufacturing yields just wouldn’t reach acceptable levels.
And so 10nm chips didn’t appear at the rescheduled launch date, around the summer of 2017. The deadline then moved to 2018, and to an unspecified quarter in 2019 next. As we near the end of 2019, rumors are abound that the 10nm generation (or at least major product lines within it) will be a total failure.
Intel vehemently denies this. In fact, at the latest conference call, CEO Swan announced that the “10nm product era has begun”. A couple of products seem to be shipping indeed, but industry watchers now want to know if they’re actually worth their salt. While getting 10nm up and running, Intel put out several upgraded versions of 14nm chips (up to 14nm++++). These may now be so good, that there’s little incentive to buy 10nm parts on performance. In fact, even on early roadmaps Intel presented, it took one or two updates for 10nm (to 10nm+ or 10nm++) to start outperforming 14nm(++++).
Meanwhile, AMD has been earning praise for its 7nm consumer and server products, which seem to be on par with more expensive Intel processors, or even better at some tasks. And that’s before the 7nm design is moving to TSMC’s 7nm+ process, which employs EUV lithography and delivers some additional performance upgrades. This will presumably happen in 2020.
In the third quarter – the first ‘full 7nm quarter’ – AMD posted its highest revenue since 2005, despite game console chips weighing down sales. “I’m extremely pleased with our progress as we have the strongest product portfolio in our history, significant customer momentum and a leadership product roadmap for 2020 and beyond,” said AMD CEO Lisa Su. She’s not exaggerating: AMD may not be miles ahead, but Intel is playing catch-up right now.
It’s not like AMD is hurting Intel much financially, however: the latter posted record quarterly revenues as well in Q3. And Intel still controls roughly 75 percent of the market. But losing the edge technologically is nonetheless painful for a company that flaunted its leadership in the past and prided itself on flawless execution cycle after cycle. Even if Intel manages to retake pole position, it will take a long time before it can shake off the 10nm bust.