Intel Talks Sapphire Rapids Respin, Granite Rapids Power Up, and AWS Deal

Intel revealed during its earnings call on Thursday that it needs to relaunch its 4th generation Xeon Scalable ‘Sapphire Rapids’ processor before launching high-volume production. This overhaul and removal took time, which is why Intel is delaying the ramp-up of SPR processors to late 2022. Meanwhile, Intel’s “Granite Rapids” Xeon processor compute tile, slated for 2024, is on track to be fueled later in the third quarter. . Intel has also signed an agreement with AWS to develop custom data center solutions.

Not so brilliant for Sapphire Rapids

It turns out that Intel’s previous iteration of its next-generation Xeon Scalable “Sapphire Rapids” processor had a security vulnerability that required hardware-level mitigation.

(Image credit: Los Alamos National Laboratory)

As a result, Intel had to reboot the processor, fix the problem, save the new stepper, and then go through regular testing and validation procedures. This naturally postponed the volume launch of the product. Some Intel customers (likely among hyperscale data center operators) are already using initial Sapphire Rapids processors unaffected by the security issue (perhaps because they ship with some features disabled).

“We are not shipping at the quality levels, at the security levels that we needed to […] it’s clear we shouldn’t have had this bug in the product in the first place,” Intel chief executive Pat Gelsinger said during the company’s earnings call. […] So we made another [stepping]what I will say for higher volume SKUs, and those will ship in volume in the second half of the year.”

Intel has hundreds of customers with different requirements for its Xeon processors. For example, hyperscale cloud data center operators usually need some specific features as well as high performance, but sometimes they don’t need all the features that Xeons have to offer. As a result, Intel can ship new processors to them before general availability without any negative consequences. But for general customers like traditional server makers, Intel needs to deliver processor models that will satisfy many different types of customers with different workloads and software. Therefore, a bug in a chip may not affect giants like AWS, Meta, or Microsoft Azure. But it could affect businesses with different needs and software stacks. So it’s better to solve the hardware problem before shipping.

But a new step (and resulting new testing/revalidation) automatically delays the high-volume launch, which is why Intel typically won’t release its 4th Gen Xeon Scalable “Sapphire Rapids” processors until later this year. So while the production ramp for these processors will begin in 2022, it will progress through 2023.

“Sapphire Rapids [is] ramp up later,” Gelsinger said. “We’ve released a few SKUs, which is good, but the major SKUs haven’t released, and they’re coming later in the year. Of course, they will contribute a lot more next year than they will this year.”

Intel admits its Sapphire Rapids processors are late to the party and won’t be widely available until 2023. It remains to be seen how this affects the availability of Intel’s 5th Gen ‘Emerald Rapids’ processor which is compatible with Sapphire Rapids. (and is manufactured using the same Intel 7 aka 10nm Enhanced SuperFin process technology as SPR) and is slated for a 2023 debut. However, at this time Intel has no plans to delay Emerald Rapids and claims that this processor indeed seems “sound”.

“Emerald [Rapids] go in the sapphire [Rapids] platform, so we are working closely with our customers and the time is right, the product looks very healthy, so we are on the right track,” said Intel CEO. “So it will be a 2023 product, then Granite [Rapids] and Sierra Forest are the 2024 products.”

Granite Rapids will be powered up this quarter

But while there are issues with Intel’s 4th Gen “Sapphire Rapids” Xeon Scalable processor, development of one of its successors – the Granite Rapids processor, codenamed for 2024 – seems to be proceeding correctly.

“We have now registered the first stage of the Granite Rapids processor and expect a power-up this quarter,” the Intel official said. “During the second half of this year, we expect to register many internal customer and foundry test chips on various process nodes, including Intel 3 and Intel 18A.”

(Image credit: Intel)

Intel’s Granite Rapids codename is an important product for Intel. The processor is expected to arrive in 2024. It will use an all-new platform that will succeed the upcoming LGA4677 socket-based platform that will support 4th Gen Intel Xeon Scalable “Sapphire Rapids” and Xeon Scalable “Emerald Rapids” processors. 5th generation. It could go by the name of 6th Generation Xeon Scalable Processor while increasing performance and bringing additional features, which will improve Intel’s competitive position in the server market.

The Granite Rapids Compute Tile will be fabricated using Intel 3 fabrication technology which promises to deliver denser high-performance libraries, increased intrinsic drive current, and reduced resistance – three characteristics particularly beneficial for processors. data center processors. Meanwhile, Intel claims its I3 node will deliver an 18% performance/watt gain over its I4 node. Intel’s I3 process node has never been listed for any upcoming client PC products from Intel (Meteor Lake with an I4 compute tile will be replaced by Arrow Lake with an I20A compute tile), so we can assume that this technology is indeed very well suited for heavy-duty data center applications rather than typical customer products.

Custom Data Center Solutions for AWS

Additionally, Intel announced on the call that it will be developing custom data center solutions for Amazon Web Services.

“In the second quarter, we agreed to expand our partnership with AWS to include the co-development of multi-generational data center solutions optimized for AWS and Intel infrastructure as a strategic customer for internal workloads, including EDA,” Gelsinger said.

(Image credit: Intel)

Intel’s wording about its expanded partnership with AWS is vague at best. The company did not specify what exactly it plans to co-develop with the cloud giant. However, since Intel’s core prowess is chip development, we might assume Intel is talking about data center-grade x86 SoCs or multi-chip/multi-chip Systems-in-Packets (SiP). -roof tiles.

Intel’s intention to create highly customized datacenter-grade x86 processors for large customers like AWS or Meta has always been a core part of its IDM 2.0 strategy. AWS may be a perfect customer for Intel and its foundry division. AWS consumes a lot of chips, can take advantage of almost any IP address Intel has to offer, can bring a properly developed and explicitly designed internal IP address for AWS-enabled services and platforms like Netflix, and is ready to pay large sums for the material he needs.

Strategically, AWS’s sticking to x86 means the company will continue to use this architecture for the foreseeable future. Of course, AWS is a company that uses all available platforms including AMD, Nvidia as well as Arm (eg Ampere Altra/Altra Max based platforms). Still, the custom Xeons, of course, indicate that AWS is very serious about custom x86-based solutions.

Comments are closed.