Leaked Slides Offer Details on Intel’s Meteor Lake, Arrow Lake CPUs

The slides are labeled with both Intel and Hot Chips badges, indicating they were presented at the recent silicon symposium. They were unearthed by Wccftech, and look like the types of slides that would be shown off at a conference. We already knew some of this information, including Meteor Lake’s tile assignments. It’ll be built on a new 7nm process dubbed Intel 4. It’s supposed to be in testing now, with high-volume production starting in early 2021. It was originally supposed to come after Alder Lake, but it was delayed by roughly a year. This caused Intel to come up with a workaround, which it named Raptor Lake. Pat Gelsinger has already said on the record that Intel 4 testing is going well.

What’s notable is that Intel has been increasing core counts recently, but it’s seemingly reversing that with Meteor Lake. It went from 16 to 24 cores moving from the 12900K to the 13900K. However, Meteor Lake’s theoretical 14900K will offer just six P-cores, for 22 cores total (thanks to 16 E-cores). That’s two fewer performance cores than it used in both of its hybrid architectures. This would indicate Intel is favoring efficiency with Meteor Lake over raw horsepower. Another departure for Meteor Lake is only the top SKU will offer a 125W TDP. The rest of the lineup is 65W and lower, which bolsters the theory it’s a mobile-first design. Or it’s possible the slides are only referring to its mobile CPUs, not its desktop parts. A previous leak of a Meteor Lake mobile block diagram indicated it would max out at 14 cores.

Another big change for Meteor Lake will be the Xe-based integrated GPU. It’s currently using an older UHD 770 design for Alder/Raptor Lake, and this will be jettisoned in future architectures. Meteor Lake will offer four XE cores, with a total of 64 execution units. That’s double what it’s currently offering. Meteor Lake could use an integrated version of its current Alchemist architecture, with Arrow Lake using the next-gen Battlemage design via TSMC’s 3nm process.

For Arrow Lake, Intel will return to the 24-core design used by Raptor Lake. All of its top SKUs will offer the same design with eight P-cores and 16 E-cores, with 125W, 65W, and 35W variants. This will be Intel’s second tile-based design, on its 20A process. This will be the first non-FinFET CPU from Intel since Ivy Bridge in 2011. It’ll use RibbonFET transistors along with PowerVia backside power delivery.

Intel’s chart also notably doesn’t include any mention of its Intel 3 process, which will be an evolution of its 5nm Intel 4 node. As we wrote previously, Intel 4 will be designed for high-performance computing, with Intel 3 optimized for GPUs and ASICs. It’ll be the first node for Intel that adopts EUV, making it the last foundry to do so.

It’ll be very interesting to see how Intel proceeds with these future architectures. Alder and Raptor Lake were its first E/P core hybrid designs. It’s largely succeeded in those efforts, as they are very competitive with AMD’s Zen offerings. With Meteor Lake and beyond it’s changing to tile-based architectures similar to AMD’s current CPUs, so we’ll get to see a true head-to-head battle play out. Though TSMC still has a node advantage over Intel, the latter has always claimed its nodes are competitive despite what numbers the marketing teams use.

Meteor Lake has also been the subject of a lot of rumors concerning delays and parts being swapped out. It supposedly wanted TSMC’s 3nm node for the GPU tile. However, that was reportedly jettisoned in favor of its 5nm technology. A lot is on the line for Intel as it attempts to jump from node to node with its upcoming chips. We all know what happened at 14nm so no need to repeat that here. It’ll be interesting to see if it can make a smooth transition to Intel 4, then eventually, to 20A as well.

Continue reading

Intel Launches AMD Radeon-Powered CPUs

Intel's new Radeon+Kaby Lake hybrid CPUs are headed for store shelves. Here's how the SKUs break down and what you need to know.

RISC-V Tiptoes Towards Mainstream With SiFive Dev Board, High-Performance CPU

RISC V continues to make inroads across the market, this time with a cheaper and more fully-featured test motherboard.

VIA Technologies, Zhaoxin Strengthen x86 CPU Development Ties

VIA and Zhaoxin are deepening their strategic partnership with additional IP transfers, intended to accelerate long-term product development.

What Does It Mean for the PC Market If Apple Makes the Fastest CPU?

Apple's M1 SoC could have a profound impact on the PC market. After 25 years, x86 may no longer be the highest-performing CPU architecture you can practically buy.