Published in News

Intel reveals chip plans

by on26 April 2024


Gelsinger tries to put a silver lining into disappointing results

Despite Intel's disappointing results, the company used the announcement to reveal some of its chip plans.

During its Q1 2024 earnings call, Chipzilla confirmed that developing the 18A process node and its associated products was on track. The company will first increase production of the 20A process node, scheduled for the latter half of this year. This node will be used for two leading products launching later this year, Arrow Lake and Lunar Lake.

After introducing Arrow Lake and Lunar Lake, Intel plans to start production on its 18A process node. The company has confirmed that the 18A PDK (1.0) will be released this quarter, with the next-generation E-Core Clearwater Forest "Xeon 7" family for data centres and the Cougar Cove P-Core based Panther Lake family for consumer PCs (Desktops & Laptops) as the two leading products. Intel also noted that the 18A production ramp for both products will begin in the first half of 2025, with the product release scheduled for mid-2025.

Intel CEO Pat Gelsinger said, “We expect to release the 1.0 PDK for Intel 18A this quarter. Moreover, our lead products, Clearwater Forest and Panther Lake, are already in fabrication, and we anticipate starting the production ramp of Intel 18A in these products in the first half of '25 for product release in the middle of next year.”

He confirmed its Core Ultra ramp-up continues, with overall unit shipments expected to double this quarter. However, shipments are constrained by the supply of wafer-level assembly. Core Ultra "Meteor Lake" supply will likely improve in the second half of 2024, followed by adding Arrow Lake and Lunar Lake CPUs.

The three Core Ultra families, Core Ultra 100 (Meteor Lake), Core Ultra 200 (Arrow Lake), and Core Ultra 200V (Lunar Lake), are projected to push shipments above 40 million AI PC units by the end of 2024.

Gelsinger said that Intel’s Core Ultra ramp, spearheaded by Meteor Lake, is progressing faster than initially expected, with units forecasted to double sequentially in Q2, limited only by Intel’s supply of wafer-level assembly. The improved Meteor Lake supply in the second half of the year and the introduction of Lunar Lake and Arrow Lake will enable the company to exceed its target of 40 million AI PC CPUs in 2024.

On the Data Centre & Enterprise side, Intel is preparing for the launch of its first Intel 3-powered CPU family, codenamed Sierra Forest. Based on the optimised Sierra Glen E-Core (an enhanced version of Crestmont E-Core), the Xeon 6(E) family will offer up to 288 cores and is set to be released in the coming month. This will be followed by the Granite Rapids P-Core (Redwood Cove) family, offering up to 128 cores on the LGA 4710 and LGA 7592 sockets under the Birch Stream platforms.

Intel anticipates that the Granite Rapids-P Xeon 6(P) CPUs will signify the official production start of the Intel 3 process node. As mentioned, Clearwater Forest is one of the two lead products based on the 18A process node. Intel describes Clearwater Forest as a "Very Strong Product" using the Darkmont E-Core architecture.

“Within DC AI, as promised, we have achieved product release on our first Intel 3 server product, the first-generation E-core Xeon 6, codenamed Sierra Forest. The next-generation P-core Xeon 6 product, Granite Rapids, will be released in Q3,” Gelsinger commented.

Intel’s first 18A product, Clearwater Forest, is expected to launch next year, and Gelsinger believes it will enable the company to accelerate market share gains.

“Sierra Forest is our first Xeon 6 product on Intel 3. We now have leading process technology back on American soil for the first time in a decade. This is truly exciting. Sierra Forest, with a high core count of 144 to 288 cores, is very focused on power, performance, and efficiency, and we see a strong pipeline of customers and socket win-backs, as the area of power performance has been a deficit for us on an older node. We foresee market share gains now that we're on leading nodes,” Gelsinger said.

Granite Rapids, set to arrive in Q3, is a core part of the Xeon 6 family. Gelsinger views it as a more decisive element of the portfolio this year. Intel has yet to be as active in the power performance sockets, and Sierra Forest provides that capability.

“So, as we've described, it is a one-two punch, with Granite arriving in Q3 and a volume ramp on Intel 3. With that, we feel we have a very robust product line. Next year, Clearwater Forest, the second generation of the E-core part on the 18A in the server market, will be a solid product for us,” Gelsinger explained.

Moving on to the Foundry business, Intel highlighted another 18A customer, bringing the total to six. The customer is a leader in the aerospace and defence industry, hinted Gelsinger.

The US Department of Defense has also awarded Intel Foundry Phase III of the RAMP C program, which opens the door for more customers in the defence department. The company also mentions approximately 50 test chips based on the 18A node in the pipeline.

Intel expects the Foundry segment to grow significantly, with a Total Addressable Market (TAM) of $240 billion by 2030 (currently $110 billion). The main growth areas will include Extreme Ultraviolet (EUV) nodes and advanced packaging technologies.

“I am delighted to announce that this quarter, we signed another significant customer on Intel 18A, bringing our total to six, a leader in the aerospace and defence industry. This customer chose Intel foundry not only for the process technology benefits of Intel 18A but also for their desire to have a secure U.S.-only supply base. Just this week, we were very pleased to announce that the DoD awarded Intel foundry Phase III of the RAMP C program, which we are confident will lead to additional federal aerospace and defence customers,” Gelsinger said.

Intel mentioned its Falcon Shores AI GPU accelerator, which will combine the best of Gaudi 3 with a fully programmable architecture supported by a robust AI ecosystem (OneAPI). The Falcon Shores GPU is expected to launch in late 2025, marking a significant release for the company as it attempts to emulate NVIDIA's success in the AI segment.

Gelsinger stated: “We'll be introducing that late next year with Falcon Shores, combining the excellent systolic performance of Gaudi 3 with a fully programmable architecture. This all comes together with Falcon Shores. We then have a rich and aggressive cadence of Falcon Shores products following that.”

 

Last modified on 26 April 2024
Rate this item
(3 votes)

Read more about: