Intel is in trouble, and CEO Bob Swan knows it. If you want evidence, early last week Swan gave his leadership team a hard shake, and while many held on for the ride, Intel’s Chief Engineer Venkata Renduchintala did not.
The news came after revelations that Intel’s next-generation, 7-nanometer manufacturing process had been delayed. Honestly, I’m not sure why anyone who’s been following Intel’s 10-nanometer debacle, would be surprised, but this time, investors took notice.
And frankly, I’m glad. Nothing against Renduchintala, but over the past few years, I’ve watched in dismay as the once-proud chipmaker has lost momentum and its competitors have started nipping at its heels.
The story with Intel has become this never-ending cycle of too little too late. Intel’s mobile efforts with Atom, its cellular division that sold to Apple last year for $1 billion, and even its ill-fated Nervana AI collaboration with Facebook come to mind. So last week’s shakeup was at least a promising sign that maybe Intel has woken up to the fact it ain’t the king of the hill anymore.
Competition in this market breeds innovation, and sadly we haven’t been seeing much of that coming from Intel lately. There has been plenty of “us too” efforts, especially in AI, and Intel’s Optane persistent memory is certainly a bright spot in the company’s portfolio. Where it concerns desktop and data centers, the story is outright disappointing.
As of writing this, all of Intel’s high-end Xeon Scalable chips top out at 28 cores, 56 threads, and use the same 14-nanometer manufacturing process it’s been peddling since the end of 2014.
Yes, 10-nanometer data center chips are just around the corner, but rivals AMD, Marvell, and Ampere all have 7-nanometer chips available today and 5-nanometer chips are in the pipeline.
Critically, the cloud providers and hyperscalers, a key market for Intel’s lucrative data center group (DCG), have taken notice. All three of the top cloud providers have deployed AMD EPYC or EPYC 2 processors in their data centers, meanwhile many of those same cloud providers have also begun deploying high-core count Arm chips.
Investors have a real reason to be concerned. Continued losses in this segment are apt to hurt Intel’s bottom line. According to Swan, Intel’s data center business unit now accounts for more than half of the company’s quarterly revenues.
But it’s not just Intel’s data center efforts that are falling behind. Apple grabbed headlines earlier this summer when it announced it would begin phasing out Intel processors from its Mac line in favor of its own Arm-based Apple silicon.
The news certainly got people’s attention, but anyone who has seen benchmarks of Apple’s latest A-series chips knew this was only a matter of time. Even before Apple announced its pivot away from Intel, Microsoft has been preparing for a shift to Arm. The company’s latest Surface products show the Microsoft is as serious about Arm in their Windows business as they are in the cloud.
A Less-Than-Fab-ulous Predicament
Many of Intel’s shortfalls as of late can be sourced to poor yields in the company’s foundries.
Unlike most chipmakers — which contract out their designs to fabs like TSMC, Global Foundries, Samsung — Intel is one the few still designing and manufacturing its own chips. In theory, this kind of vertical integration should pay dividends, and in the past maybe it did, but today Intel is falling behind.
During Intel’s second-quarter earnings call, Swan acknowledged these shortcomings and laid out the company’s plan for addressing them. Swan assured investors that there was no “fundamental roadblock” to resolving the issues with the company’s 7-nanometer chips, claiming the first would ship sometime in late 2022 or early 2023.
But in addressing these concerns, Swan let slip the company’s contingency plans, which include the potential to leverage external chip fabs in order to satisfy performance demands.
Using external fabs won’t be a first for Intel either. The company is already using a combination of internal and external fabs for its upcoming data center GPU, which is slated for release next year.
Now, I’m not an engineer; I’m not going to say that Intel should swallow its pride and embrace external fabs the way AMD did. What I do know is by the time Intel can bring a 7-nanometer chip to market, TSMC is slated to start shipping 3-nanometer chips. Given Intel’s reputation for meeting deadlines, my money is on TSMC.
So Intel, maybe it’s time to take a cue from Apple and “Think Different” before the rest of your customers do the same.
"chips" - Google News
August 04, 2020 at 04:00AM
https://ift.tt/3i2dAiO
The Chips Are Down. Can Intel Catch Up? - SDxCentral
"chips" - Google News
https://ift.tt/2RGyUAH
https://ift.tt/3feFffJ
Bagikan Berita Ini
0 Response to "The Chips Are Down. Can Intel Catch Up? - SDxCentral"
Post a Comment