Intel’s new Tiger Lake processors for ultra-thin laptops are packed with new silicon building blocks for AI, graphics and other technologies to serve a “spectrum of needs” in the mobile computing space, according to top Intel engineer Boyd Phelps.
In an interview with CRN, Phelps said the Santa Clara, Calif.-based company has taken a holistic and balanced approach to the engineering and design of the new processors, which are the first CPUs in the 11th-generation Intel Core family. This means that the company has devised ways, for instance, to offload certain AI workloads, like blurring the background in a Zoom video call, to new accelerators within the chip to make the workloads run faster while also saving on power.
“For us, we thought about it in the context of how the different workloads have evolved and emerged. They all have kind of a different sweet spot, so for us, we geared Tiger Lake to meet that spectrum of needs,” said Phelps, who is vice president of the Client Engineering Group and general manager of Client and Core Development Group at Intel.
Phelps said Intel baked in AI capabilities into the Tiger Lake processors in multiple ways. The processors’ new Intel Iris Xe graphics come with a new instruction set for neural networking inferencing. They come with the Intel Gaussian and Neural Accelerator 2.0, which can be used to blur background video backgrounds and suppress background noise. They also feature a faster and more efficient vector instruction set inside the processors’ Willow Cove cores.
“That‘s absolutely the direction that we’re going to with architecture. AI is not just one engine. It’s really about how do you actually meet the needs of the different workloads and how do you do it both power efficiently [while also enabling developers to take advantage of new features],” Phelps said.
When Intel started working on Tiger Lake, Phelps said, the company didn’t want to focus on a single benchmark to drive design and engineering decisions. Instead, the company wanted to make mobile PCs “capable of a much broader experience.” A big aspect of that was ensuring that laptops could provide the same level of experience, whether plugged or unplugged.
With the Intel Gaussian and Neural Network Accelerator 2.0, for example, Intel wanted to eliminate the need to use of the integrated graphics of the CPU’s cores — the Willow Cove cores in the case of Tiger Lake — to run certain AI workloads like background image blurring or background noise cancellation. The Gaussian and Neural Network Accelerator is designed to run those workloads more efficiently, meaning that the integrated graphics or CPU cores don’t have to spend more energy, which translates into more performance headroom for other workloads or more battery life.
“It‘s about making sure that we actually use the power in the right places at the right time and we’re not wasteful in that context,” Phelps said.
This factored into other design choices, like having multiple Thunderbolt 4.0 USB-C ports integrated into the processor. The Thunderbolt integration means OEMS don’t have to add their own discrete components, which lets them use smaller motherboards and larger batteries, according to Phelps.
“By having it also integrated, we can actually do better power management when we have devices that are connected to your laptop,” he said.
Outside of the specialized components, Tiger Lake’s Willow Cove cores can drive a high level of performance and even more than what the company has already promised, according to Phelps, which was made possible by Intel’s new 10-nanometer SuperFin technology that provides the “largest single intranode enhancement” in the company’s history.
“We talked about 4.8GHz when we launched Tiger Lake. We‘ll be able to drive that to even higher speeds, much higher speeds as we go throughout the year,” he said.
Taken together, all of Intel’s design and engineering decisions were made with the idea that mobile computing is becoming even more important — a perspective Intel held long before the pandemic.
“I think it‘s accelerated the curve of realizing the importance of mobility and connectedness and the experiences that people have,” Phelps said. “We care about [Intel Architecture] performance. We care about graphics performance. We think that shows in Tiger Lake and what we’ve delivered, but I think more than that, we care about the full experience people can get from that.”
