Intel's Sapphire Rapids server chips face yet more delays, with AMD set to benefit

INFRA

Intel Corp. today confirmed that an initial launch event for its long awaited Xeon Scalable central processing units will take place in January, but it’s still not clear when the chips will become commercially available.

Intel’s announcement came in the wake of a report by industry watcher TrendForce, which claimed that mass production of the chips, code-named Sapphire Rapids, has been delayed until 2023. However, Intel insisted it is now upping production of the chips and they meet product release qualifications.

Although the launch event is scheduled for January, there was no word on general availability of the chips. In recent years, there have been long gaps between Intel’s launch events and the new products actually going on sale.

TrendForce’s report suggests there could be yet another gap. The Taiwan-based analyst firm claims that Intel is struggling with yield issues on its Intel 7 node, with a current production yield rate of just 50% to 60% affecting higher-core-count chips.

As a result, mass production likely won’t kick off until the first half of next year, TrendForce said. What that means is that customers may have to wait several months after the official launch event before they can actually get their hands on Sapphire Rapids.

Intel didn’t confirm the delay, instead simply reiterating what was said in its original announcement. “Sapphire Rapids volume SKUs are product release qualified with a very strong volume ramp expected across all SKUs,” the company said in a statement to The Register.

The seemingly inevitable delay with Sapphire Rapids means the chips will arrive more than a year behind schedule. Intel had originally planned to ship the new Xeon Scalable CPUs to cloud customers and enterprises in the fourth quarter of 2021, with wider availability planned for the first half of 2022. However, Intel said in June 2021 that it was shifting those windows to the first and second quarters of 2022.

Fast forward another year, to June 2022, and Sandra Rivera, Intel’s executive vice president and general manager of the Datacenter and AI Group, told investors that production of Sapphire Rapids was being pushed back again to “later in the year.” According to Rivera, the delay was related to the need for additional product validation time, rather than any manufacturing issues.

Regardless of what the reason is, TrendForce said, the delay will be a big “boon” for Intel’s rival Advanced Micro Devices Inc., which is said to be on track to shift its own next-generation data center chip, code-named Genoa, later this month. AMD’s new EPYC server chips will be the first in the industry to support DDR5 memory, PCEe 5.0 connectivity and Compute Express Link 1.1 standards, TrendForce said. That should result in a strong performance boost, making AMD’s server chips a more tempting option than Intel’s current generation server silicon.

With no concrete launch date for Sapphire Rapids, it’s likely that AMD will be able to sell a lot of extra server chips in the coming months. That seems to be the expectation anyway, for AMD said in an earnings call just today that it’s expecting to see strong sales growth in its data center business over the coming quarters.

TrendForce is forecasting that AMD’s share of the server CPU market will increase from 15% to 22% in 2023, with most of that growth coming at Intel’s expense.

Image: Intel