It was barely a year ago when Intel finally slayed the demons that had been holding back its 10nm process and keeping it from most buyers for years. But we now know the first 7nm processor will only hit the streets in early 2023. Much like the 10nm process, it will begin in data centres.
Intel CEO Bob Swan said last week that 7nm was 12 months behind schedule and that the company might look at using external foundries to stay on track. He also said Intel was planning more products based on 10nm.
It really does sound like a replay of the procession of 14nm products that Intel would put out in years past.
In those days, Intel was far enough ahead of its rivals that it had some time to burn. This time, it doesn't have that luxury, and the computing industry is shifting beneath its feet. The chip giant knows this and recently announced its Lakefield hybrid processors that mix low-power and high-power cores in ways Arm has done for a long time.
The other difference is that in 2020, AMD is far from being an also-ran. Recent Ryzenannouncements from AMD have gleefully boasted of its performance per watt numbers and is finally beating Intel across a wide range of benchmarks.
"What you would have seen if you want to look at [these benchmarks] a few years ago, is that AMD across the board here would have been a substantial deficit," AMD director for its commercial client business Matthew Unangst said in May.
"When we look at Cinebench 1T, what used to be a large double debit deficit in performance, we've pretty much eliminated that deficit and we're very close to parity."
While AMD is parading around its 7nm process, it's worth noting that not all fabrication measurements are created equal and your 7nm is not necessarily my 7nm, nor Intel's 7nm. It's more a marketing term now than any reflection of absolute physical sizing.
In 2017, Intel was claiming its 10nm process would have double the transistor density of its competitors, and last year, Swan said Intel 7nm would be the equivalent to the 5nm process from Taiwan Semiconductor Manufacturing Company (TSMC) used for iPhone chips. The numbers comparing Intel 10nm to TSMC 7nm certainly lend credence to this view.
This might be a workable situation for Intel if the world stayed static, but Samsung has been talking about its 3nm process for a year, and even though volume production is reportedly set to arrive in 2022 following a COVID-induced delay, the Korean giant will have its foundries running on a new Gate-All-Around process before a 7nm Intel processor becomes available for sale.
A potential wildcard in the chip world arrived last week when Bloomberg reported Nvidia was having a good think about buying Arm from Softbank. Should the deal happen, Intel would eventually have to face yet another competitor with superior graphics processing credentials in the CPU space.
Clearly the sharks are circling and Intel is facing an existential crisis, right? Not quite. The company also released its second-quarter earnings last week and posted revenue growth of 20% to $19.7 billion for the quarter. Of that number, $9.5 billion was from the PC business, representing an increase of 7%.
If that is an existential threat in the midst of a global pandemic, many other companies would love to get such a deal.
The other factor in Intel's favour is, outside of performance per watt, raw performance numbers are not as crucial as they used to be. Apple absolutely spanks Qualcomm when it comes to mobile chip performance, and yet the regular user is blissfully unaware since the silicon in most Android devices is good enough.
The threat to Intel isn't so much going out of business, it's more about losing its prized leadership position, whether real or perceived, which has allowed it to have healthy margins and a reliable ecosystem of support. As a sometimes user of AMD hardware, issues do tend to pop up from time to time, showing that the support isn't quite there compared to the full Wintel experience, but the gap is closing.
This is before Cupertino kicks off an Apple Silicon advertising blitz that sways the public into thinking chips other than Intel ones in Macs will make them the fastest ever, greatest ever, or whatever barrel-full of adjectives Apple decides to throw at it.
To stay on top, Intel will need to pull a 10nm rabbit out of its hat to stay in the game in order to compete with the 7nm and lower processes from other foundries.
Intel is in real danger of losing its place on the pedestal, a place it has grown used to over many decades.
ZDNET'S MONDAY MORNING OPENER
The Monday Morning Opener is our opening salvo for the week in tech. Since we run a global site, this editorial publishes on Monday at 8:00am AEST in Sydney, Australia, which is 6:00pm Eastern Time on Sunday in the US. It is written by a member of ZDNet's global editorial board, which is comprised of our lead editors across Asia, Australia, Europe, and North America.