Intel to Produce Custom AI Chips and Xeon 6 Processors for AWS

[Nomad76]

Intel Corp. and Amazon Web Services. Inc., an Amazon.com company, today announced a co-investment in custom chip designs under a multi-year, multi-billion-dollar framework covering product and wafers from Intel. This is a significant expansion of the two companies' longstanding strategic collaboration to help customers power virtually any workload and accelerate the performance of artificial intelligence (AI) applications.

As part of the expanded collaboration, Intel will produce an AI fabric chip for AWS on Intel 18A, the company's most advanced process node. Intel will also produce a custom Xeon 6 chip on Intel 3, building on the existing partnership under which Intel produces Xeon Scalable processors for AWS.

"This expansion of our longtime relationship with AWS reflects the strength of our process technology and delivers differentiated solutions for customer workloads," said Pat Gelsinger, Intel CEO. "Intel's chip design and manufacturing capabilities, combined with the comprehensive and broadly adopted cloud, AI and machine learning services of AWS, will unleash innovation across our shared ecosystem and support the growth of both businesses, as well as a sustainable domestic AI supply chain."

"At AWS, we're committed to delivering the most powerful and innovative cloud infrastructure to our customers," said Matt Garman, CEO at AWS. "By co-developing next-generation AI fabric chips on Intel 18A, we continue our long-standing collaboration, dating back to 2006 when we launched the first Amazon EC2 instance featuring their chips. Our continued collaboration allows us to empower our joint customers with the ability to run any workload and unlock new AI capabilities."

With this expanded collaboration, Intel and AWS underscore their commitments to accelerating U.S.-based semiconductor manufacturing and creating a vibrant AI ecosystem in Ohio. Intel continues to be committed to the New Albany area and its plans to build leading edge semiconductor manufacturing. AWS is planning to invest $7.8 billion to expand its data center operations in Central Ohio, in addition to the $10.3 billion it has invested in the state of Ohio since 2015.

"This collaboration between Intel and AWS is a great development for U.S.-based manufacturing and solidifying Ohio as a leader in AI," said Ohio Governor Mike DeWine. "Today's announcement furthers Intel's commitment to U.S. manufacturing sites, like Ohio's, as well as AWS's commitment to its nearly decade-long investment in our state."

AWS and Intel have a more than 18-year relationship dedicated to helping organizations of all sizes to develop, build, and deploy their mission critical workloads in the cloud, while also supporting them to reduce cost and complexity, increase security, accelerate business outcomes, and scale to meet their current and future computing requirements. Going forward, Intel and AWS also intend to explore the potential for further designs to be produced by Intel based on Intel 18A and future process nodes including Intel 18AP and Intel 14A, which is expected to be produced in Intel's Ohio facilities, as well as the migration of existing Intel designs to these platforms.

View at TechPowerUp Main Site | Source

 

[Ravenas]

Glad to hear they are drumming up some major clients.

 

[Minus Infinity]

Irrespective of how you feel about Intel, it will be good for the consumer and industry if IFS succeeds and we have a real alternative to TSMC, because Scamsung have been a miserable failure so far.

 

[Daven]

Hopefully Intel will lean more into manufacturing and less into chip design.

 

[john_]

One of the rare occasions where we have good news for AMD (PS6), good news for Intel and no news for Nvidia.

In any case Intel seems to have something in mind with AI capabilities in it's CPUs, ALL CPUs, while AMD is having a difficulty to stop being short sighted keeping their CPUs, with the exception of the new mobile series, without AI capabilities(meaning NPU). They where short sighted with their 7000 series, not seeing the importance of RT performance in the current generation, they still don't integrate strong NPUs in their desktop Ryzen and EPYC CPUs. They are confused as the main competitor of Nvidia, thinking that GPUs is the only way for AI when the market will buy ANYTHING with AI on it at this moment. AMD will start losing market share in servers if they keep sleeping and unfortunately they have a habit of sleeping while in frond of the competition.

 

[kondamin]

Hopefully Intel will lean more into manufacturing and less into chip design.

They were keeping up with amd that had a 3 node advantage, intel chip designs are great.
previous intel management not so much.

 

[Scrizz]

They were keeping up with amd that had a 3 node advantage, intel chip designs are great.
previous intel management not so much.

 

[john_]

They were keeping up with amd that had a 3 node advantage, intel chip designs are great.
previous intel management not so much.

Intel engineers have more resources and experience over the years. Intel building CPUs in it's own fabs probably also offers some advantages. Intel's biggest idea was the hybrid designs. It gave them the opportunity to start selling more cores compared to AMD, while also claiming better efficiency when we all know that Intel CPUs hit easily over 200 watts when having to work hard. But Intel's experience with little cores for over 15 years really helped them. So, while AMD had 2-3 nodes advantage, Intel had others advantages. Fabs, experience with little cores, probably more engineers working on CPU designs, stronger ties with big OEMs, better name in the public and better treatment from the press. Where many of those advantages don't really count, is the server market. There little cores don't help and power consumption is critical. That's why AMD is more successful in servers.

 

[kondamin]

Intel engineers have more resources and experience over the years. Intel building CPUs in it's own fabs probably also offers some advantages. Intel's biggest idea was the hybrid designs. It gave them the opportunity to start selling more cores compared to AMD, while also claiming better efficiency when we all know that Intel CPUs hit easily over 200 watts when having to work hard. But Intel's experience with little cores for over 15 years really helped them. So, while AMD had 2-3 nodes advantage, Intel had others advantages. Fabs, experience with little cores, probably more engineers working on CPU designs, stronger ties with big OEMs, better name in the public and better treatment from the press. Where many of those advantages don't really count, is the server market. There little cores don't help and power consumption is critical. That's why AMD is more successful in servers.

Amd made a brilliant move by going for chiplets when they did, they Can offer 128 core chips that cost them A fraction of what it takes intel to make a 56 core chip as those are big slabs of silicon.

in hindsight intel should just have followed them on that before they were finished with foveros and taken the slight interconnect penalty.
4 x 16 p core on something similar like an epyc package, it would have needed 3000W but it would wipe the floor with anything on the market.