Tiger Lake is Intel’s upcoming line of processors, and at its 2020 Architecture Day, the company made some bold claims about the performance gains in this latest generation. Intel says Tiger Lake has a “greater than generational improvement” in performance over its predecessor, Ice Lake, resulting in the “largest single intranode enhancement in its history” and “performance improvement comparable to a full node transition.”
In other words, Tiger Lake won’t be your average generational update. You don’t need to know what an “intranode enhancement” is to hear Intel’s ambitions. For an industry as slow and iterative as processor design, a claim like that is enough to make you sit up and listen.
How’d Intel pull it off? Well, the company took aside an entire day to explain the engineering behind just how big of a leap forward this really is.
Refining transistors, not adding more
The most common way of increasing processor performance is a “node transition,” as Intel refers to in its audacious claims. Shrinking the die from 14nm to 10nm, for example, allows for more transistors — and more potential performance. Intel has infamously struggled with these transitions lately, falling behind the competition. Even having finally moved partially to 10nm with the current-generation Ice Lake, Intel hasn’t seen the greatest success in drawing the performance it needs out of that design.
But there’s another option in Intel’s toolkit: Redesign the transistors themselves. That’s the approach Intel is introducing with what it calls “10nm SuperFin.” The 10nm part we already know about. Intel’s upcoming Tiger Lake chips will be the second generation of its troubled 10nm process. In the past, Intel would have labeled generations of microarchitecture design as 10nm+, 10nm++, and so on. But now, it’ll have a more memorable name for this iteration on 10nm.
The new “SuperFin” terminology refers to a redesign of the billions of transistor used in Tiger Lake, which Intel calls a “high-performance” transistor. The goal of the refinements was to reduce the amount of current leaked through the gate of each transistor. Less wasted power means a lower operating voltage and potentially more headroom for performance, or diverting power to other components, such as the GPU.
Intel also redesigned the metal interconnect, a stack of intricately layered metals that acts as a capacitor for the current of the transistor. The new design supplies five times the capacitance by using ultra-thin layers of dielectric materials, which Intel claims is an “industry-first” accomplishment.
SuperFin is all of those changes wrapped up into a single catchy brand. The other half of the performance gains come from how those transistor efficiencies are used, which is where Intel’s new microarchitecture comes in, known as Willow Cove.
Higher frequencies with Willow Cove
Intel says Willow Cove’s primary advantage over its predecessor (Sunny Cove) is in frequency. Thanks to SuperFin, Will Cove cores feature a greater dynamic range and better power management. No specifics or numbers were provided, but Intel says Willow Cove performs far better than Sunny Cove in bursty workloads where heightened responsiveness is beneficial.
“We’re able to drive Willow Cove to much higher frequencies for much less voltage than what you’d expect,” an Intel spokesperson said. “But from a dynamic standpoint, if a CPU is augmenting a certain workflow, we’re able to do that at a much lower voltage and do something like give more power to graphics.”
Intel also claims that Willow Cove is not only faster when TDP is restricted (such as in small laptops), but also in unconstrained performance. That sounds promising for future products such as 15-inch laptops or even desktops, both of which are still stuck on older 14nm architectures.
Of course, there was no mention of core counts, specific clock speeds, or product details. Based on what Intel showed on Thursday, we wouldn’t expect Tiger Lake to make any large leaps in core count to match what AMD has done with its Ryzen 4000 chips. Graphics, though, is another story.
Xe graphics finally launch
The most exciting of the upgrades coming to Tiger Lake, by far, arrive in the graphics department. It’s the first to use Intel’s Xe GPUs, which promise a huge improvement to graphics, ranging from integrated graphics all the way up to the data center.
Tiger Lake graphics build on what was already a massive improvement in Ice Lake. Intel’s integrated Gen11 “Iris Plus” graphics were offered in 10th-generation Ice Lake laptops, which doubled the performance of the terrible Intel UHD integrated graphics of yesteryear.
Tiger Lake takes it a step further, bumping up the number of EUs (execution units) from 64 up to 96. Intel emphasized the ability of these graphics even in restricted form factors, as low as 15 watts, which is the standard size for many 13-inch laptops. In Battlefield 1, the company showed how a 15-watt Tiger Lake system had smoother gameplay than a 25-watt Ice Lake system.
This increase in performance won’t transform your laptop into a full-fledged gaming laptop — not by any means. But it looks to be a meaningful boost in frame rates, especially while playing games at low-quality settings (in 1080p, of course). Intel showed a number of games playing on Tiger Lake integrated graphics, including Doom Eternal, Battlefield V, and PlayerUnknown’s Battlegrounds. Intel says less demanding titles, like the racing game Grid, can even be played at higher quality settings. Again, no exact frame rates were provided.
Of course, increased graphics performance can also benefit content creation tasks, such as video editing or 3D rendering. And the chips will increase memory and fabric efficiency, feature an updated media engine, and driver improvements.
Intel’s Tiger Lake processors are still scheduled for launch before the end of 2020. Some laptops manufacturers, such as Acer, have already promised 11th-gen Tiger Lake laptops before 2021 as well.
Intel has an event scheduled on September 2, where the company is rumored to provide more details on specific Tiger Lake chips, including some concrete performance data and specific information on the processor lineup.
