AMD CTO: 'We Went All In' On 7nm CPUs


AMD is betting big on its 7-nanometer processors, going so far as to move up the launch date of the technology, said CTO Mark Papermaster.

"We knew 7nm would be a big challenge, so we made the bet, we shifted our resources onto the new node," Papermaster said earlier this month in an exclusive interview with CRN. "We didn't just dip our toe in the water. We went all in."

[Related: Nvidia Q2 Data Center Sales Soar As AI Fuels Demand]

Papermaster said the company originally planned to release the 7nm Vega graphics processor in 2019 but was able to move the product up the calendar as a result of "immense focus." AMD's first 7nm product is now slated to arrive later this year with the company's next-generation Radeon Vega Instinct GPU, which was demonstrated at this year's Computex and comes with 32 GB of memory.

Sponsored post

The company's 7nm EPYC server CPU is expected to come out next year, and its 7nm Ryzen desktop processors will follow sometime after that.

As Santa Clara, Calif.-based AMD has accelerated its 7nm CPU and GPU plans, Papermaster sees an advantage for the company in the next-generation chip race because of delays from rival Intel in the launch of its 10-nanometer processors, which are expected to be comparable to AMD's 7nm offerings.

"I think we do have the opportunity to be positioned much stronger than we originally anticipated," Papermaster said, "but I have to say our original plan was to be positioned very strongly, so any delay from our competitor could simply strengthen the value that AMD brings to the market."

For AMD, going all in on the 7nm node meant aligning its design team, the foundries and the company's electronic design automation partners, which build all of AMD's computer-aided design tools.

"Turns out, they had to be the third leg of the stool because it was such a difficult [task]," Papermaster said. "It changed even the automation tools that are required to put together these types of complex leading-edge designs."

While a laptop and NUC mini PC with Intel 10nm CPUs were released this year, Intel has delayed mass production of its 10nm chips for multiple years and now plans for a wider launch with client CPUs in holiday 2019, with 10nm server CPUs to follow in 2020. Intel has maintained that the company will continue to deliver "performance leadership" with significant performance and process improvements, as well as architectural advancements.

At The Channel Company's XChange 2018 event this week, Intel U.S. channel chief Jason Kimrey said the breadth of the company's growing portfolio of products — which now includes Optane memory and reprogrammable FPGAs — represents the only way solution providers can address the "almost unlimited demand for compute" because CPUs can't handle it alone now.

"I believe Intel over the years has delivered very consistent performance, price-performance benefits and will continue to do so," said Kimrey, general manager, U.S. channel scale and partners at Intel, in an interview with CRN when asked how partners should view the 10nm delays. "We already talked about the roadmap we have laid [out] over the next couple of years, and I just think we'll continue to innovate and meet the requirements of the customers."

In a statement to CRN, Navin Shenoy, Intel’s lead data center executive, said the company has "actionable plans to win in the highest growth areas, and we have an unparalleled portfolio to fuel our growth – including performance-leading products and a broad ecosystem that spans the entire data-centric market."

As for AMD, Papermaster said the company's 7nm processors will see an improvement in performance and energy efficiency, the result of moving the chip's transistors closer to each other than AMD's current 14nm FinFET process node. The practical effect is that 7nm processors will provide "double the performance per watt of energy expended in a previous node."

"That's significant," he said.

There's a major difference that separates AMD's and Intel's manufacturing approaches: the former is fabless, meaning that AMD relies on external foundry companies like GlobalFoundries and TSMC for manufacturing while Intel runs its own operations.

"Before I got here, AMD had made a decision to go fabless, because the value we bring, of getting great, high-performance products to market, we need to be able to move very nimbly from a design standpoint," Papermaster said. "And to really leverage what it takes to be successful in a semiconductor foundry industry, it needs 100 percent focus."

An Intel spokesperson said that its "manufacturing capabilities are unmatched in the industry, providing us a unique advantage to deliver a broad range of industry-leading products to our customers."

Randy Copeland, CEO of Velocity Micro, a Richmond, Va.-based system builder, said AMD's advances with its 7nm processors, as well as with its progress with Ryzen and EPYC CPUs, have brought welcome competition to the CPU market, but success is not guaranteed.

"7nm is a very important next step, but that alone does not guarantee that the new product will be better than anything else," he said in an email. "Only the real-world testing will prove that, but it does pave the road for better processor and better margin flexibility for AMD."

For Intel, the 10nm delays mean "they will have to scramble with the current architecture to stay competitive and try to hold market share — which they could if they need to," Copeland said.

"In reality, these are two companies that both have undeniable strength, and it's just very refreshing that there is competition for the CPU again, and it's spurred an arms race that our customers are really benefiting from," he said.