After spate of delays, Intel promises Sapphire Rapids Xeons for early 2023
x86 giant offers January launch, AMD eating its chips
Intel said its next-generation Xeon Scalable CPUs will launch in January of next year, shortly after a report out of Taiwan stated that mass production of the server chips was delayed to 2023.
The semiconductor giant announced a January 10 launch date for the oft-delayed processors, code-named Sapphire Rapids, on Tuesday. The company said mass-production Sapphire Rapids units now meet product release qualifications, and it has begun ramping up production. This echoes statements made by Intel CEO Pat Gelsinger during the company's earnings call last week.
"We are ramping the product as we speak... We expect this will be our fastest-ever Xeon to a million units. And we're going to push that quite aggressively, and the factories are ramping up as we speak," he said.
While Intel has announced a launch event for Sapphire Rapids, it is still not known when next year the processors will be commercially available for the broader IT market. Increasingly in the past few years, there have been gaps between launch events for products and general availability.
Intel's launch announcement comes after Taiwanese research firm TrendForce said this morning that mass production of the Silicon Valley stalwart's server chips had moved from the fourth quarter of this year to the first half of next year due to manufacturing issues.
TrendForce said Intel was experiencing yield issues with the Intel 7 node, saying the current production yield rate for Sapphire Rapids is estimated to be in the range of 50 percent to 60 percent. The firm said this was impacting higher core count models.
What a coincidence!
When The Register asked about these claims, an Intel spokesperson did not directly address them and instead pointed to Tuesday's launch announcement, which arrived hours after the TrendForce report.
"As we shared at earnings, Sapphire Rapids volume SKUs are product release qualified with a very strong volume ramp expected across all SKUs. In fact, we just announced our January 10 launch timing today," the spokesperson said.
At the very least, Intel's January 2023 launch event for Sapphire Rapids indicates that the server processors will arrive sometime early next year, after experiencing multiple delays over the past year and a half.
- AMD's Epyc 4 will likely beat Intel Sapphire Rapids to market
- Nvidia not cutting it? Google and Amazon's latest AI chips have arrived
- Intel accidentally leaked its 34-core Raptor Lake chip. What do the dies tell us?
- Intel has a secret club in the cloud for devs to try out new chips – and you ain't in it
AMD confirmed to hit market first with new server chips
Intel's plan to launch Sapphire Rapids early next year means that AMD will beat the company to the general market with the rival's forthcoming Epyc processors, code-named Genoa, given what we just learned from one major server vendor.
Hewlett Packard Enterprise execs confirmed to The Register that 4th-gen Xeon Scalable-equipped systems won't ship before servers with AMD's Genoa CPU. HPE said the new Epyc-powered servers will launch later this month, and it doesn't yet have a date for when the new Xeon machines will arrive.
TrendForce called Sapphire Rapid's mass production delay a "boon" for AMD, which will now have the first datacenter-class processors available to support DDR5 memory, PCIe 5.0 connectivity, and Compute Express Link (CXL) 1.1 standards.
The research firm also says it now expects AMD's x86 server CPU share against Intel to reach 15 percent in 2022 and grow to 22 percent in 2023.
A series of delays, new questions about health of Intel 7 node
The latest delay means Sapphire Rapids will arrive roughly a year after the company originally said the server chips would be available for the wider market.
Back in the day, Intel said it would ship the new Xeon CPUs to cloud and other big customers in the fourth quarter of 2021, then to everyone else in the first half of 2022. Those targets then moved to the first and second quarters of 2022, respectively.
Intel then signaled another delay when the company's datacenter boss, Sandra Rivera, told investors in June that the company had pushed Sapphire Rapids production "later in the year than what we had originally forecasted." They still believed then that it could beat AMD to market with Sapphire Rapids, though Rivera warned that the delay would narrow Intel's "leadership window."
At the time, Rivera said the delay was needed for additional "platform and product validation time," noting it was not the result of manufacturing issues with the Intel 7 node, which was previously known as the company's advanced 10nm process.
However, TrendForce's new report raises questions about the health of the Intel 7 node for Sapphire Rapids chips. The semiconductor giant began using Intel 7 for mass production with its Alder Lake chips for PCs and laptops last year, and it's now using the node for the newly released Raptor Lake processors, so it seems possible that manufacturing issues are specific to Sapphire Rapids and not all Intel 7 chips.
What separates Sapphire Rapids from Alder Lake and Raptor Lake is that the server chips will be the company's first CPU to embrace a chiplet architecture. Alder Lake and Raptor Lake, on the other hand, have monolithic dies, and the company won't move its PC processors to a chiplet-based design until Meteor Lake, which is slated to launch sometime next year.
Ronak Singhal, senior fellow at Intel, told The Register last month that the semiconductor giant is prioritizing quality with this generation of Xeon processors.
"We won't compromise that quality, even if it means delaying the product," he said. "We're going through a rigorous validation cycle, and when we find issues, sometimes it will push out the schedule, but we think that's the right trade-off for our customers, and what they're expecting from us." ®