To ‘Meteor Lake’ and Beyond: How Intel Plans a New Era of ‘Chiplet’-Based CPUs

At Hot Chips 2022(Opens in a new window), Intel lifted the curtain to show off some tantalizing details around its upcoming 14th Generation “Meteor Lake” processors. Now, these cutting-edge CPUs aren’t likely to ship anytime soon. (We still have yet to see the release of Intel’s 13th Generation “Raptor Lake” processors, though Intel insists Meteor Lake is on track to launch in the second half of 2023.) But the details we’ve seen so far are deeply fascinating and show a growing change in the way that modern processors are designed.

Starting with Meteor Lake (and followed on by “Arrow Lake,” its presumptive 15th Generation Core family), Intel will transition to using a “chiplet” design for its consumer processors, with many tiny tiles that handle different functions fused together into a single chip. It’s a significant departure from the “monolithic” design of existing Intel processors, and it could lead to faster—and, maybe, more affordable—processors in the years to come.

The Birth of the Monolithic CPU Die

From the beginning, the computer industry has incessantly pushed for tighter integration. Way back in 1965, Intel co-founder Gordon Moore coined what became known as “Moore’s Law,” the much-flogged axiom that anticipated a doubling of the number of components in an integrated circuit every year. This hasn’t always held entirely true, but it nonetheless shows the critical importance of integration to the chip industry.

Over the last several decades, we have seen numerous components integrated into CPUs: floating point modules, cache, memory controllers, PCI Express controllers, video controllers, display controllers, graphics processors, and a host of other circuits. In general, this has had many positive payoffs, from lowering production costs and power consumption, to drastically increasing performance. But it has also led to challenges that companies have had to work to overcome.

How Chiplets Could Solve Monolithic Problems

Three key issues stand out with building large monolithic chips. First and foremost is the problem with chip yields.

No production process is perfect, and when it comes to silicon chips, even a seemingly small defect can cause a chip to not work correctly. This tendency makes the construction of large chips significantly more costly. That’s because when a defect occurs, more fabrication time and resources are wasted versus with a smaller defective chip.

(Credit: Intel)

Second, components like graphics processors and CPUs tend to function better when each is made under its own, optimal process technology. When integrating multiple types of components into a single chip, however, you are forced to use the same manufacturing process for all components. You ultimately must use a process that will negatively affect them both at least a little, or you will need to use a process that works great for one, but not as well for the other.

Last (but certainly not least), having all of these components tightly integrated can stymie development, to an extent. When everything’s baked together, it’s not as simple to make changes to just one component—say, the memory controller, or the video processor. You need to consider how everything is connected, and the whole chip then needs run through a lengthy verification process to ensure everything is working correctly after any change is made. After that, you still need to have the design changed at the factory, and work through existing chips, before starting production on the new design.

By using a chiplet design, however, these issues and others can be resolved or at least mitigated. A certain amount of inspection and verification needs to be done to make sure chips work correctly, to be sure, but you gain much greater flexibility in designing and updating chips. Also, you’ll see significantly less constraint on what production process you use for different parts of the chip. And a defect on any single chip will be less costly, as the chips are each smaller.

(Credit: Intel)

This change ultimately can help to hasten chip development, and potentially lower costs and improve performance, too. It’s also a proven method: AMD has been using chiplet designs for its Ryzen processors for some time now.

The Chiplet Push: Why Now, and What Could Go Wrong?

If the use of chiplets has clear and obvious benefits, so why hasn’t Intel employed them before? Well, the short answer is that the company has—many times. Intel used a chiplet-like design with its Pentium D processors way back in 2005 to combine two CPU cores into a single processor. It again used a chiplet-esque design with its 1st Generation Core “Arrandale” processors. And it has experimented with chiplets in other products since.

(Credit: Intel)

A key difference between these designs and the upcoming Meteor Lake and further-out Arrow Lake CPUs, however, is how tightly together the chips in question are coupled. One drawback to a chiplet design is that the chips simply cannot have the same level of interconnectivity that they enjoy in a single monolithic chip. Bandwidth gets hurt, as a result. This has hampered performance in the past on Intel’s products (and AMD’s, as well), but it’s something that has improved over time.

Another drawback: Older chiplets tend to use more power. But as power requirements are constantly changing between generations, it is difficult to determine how significant this problem could be in the future.

What the Meteor Lake Chiplet CPU Will Look Like

Intel’s enormous “Ponte Vecchio” chip combines 47 tiles (the company’s preferred term for these small chips) with more than 100 billion transistors. It’s with that same interconnect technology that Intel plans to connect the chips in its upcoming Meteor Lake and Arrow Lake processors. As you can see from the image below, the Meteor Lake design comprises six total pieces, including the package substrate, which will likely be little more than a PCB for connecting to the LGA socket on the motherboard.

(Credit: Intel)

Among the remaining five titles are the CPU tile, a GPU tile, and an IO Extender tile. (The last will likely contain just the PCI Express controller.) The CPU portion will be made with the upcoming “Intel 4” production process, whereas other tiles might be made by Taiwan Semiconductor Manufacturing Company (TSMC).

(Credit: Intel)

There’s also a somewhat confusingly named “SOC” tile, which contains everything that doesn’t fit into the three aforementioned tiles, including the memory controller. It does appear to be the largest tile, and it likely has a lot of functionality, but care should be taken not to confuse it with a System-on-a-Chip (SoC) that you might find in your phone or TV, as the latter would also contain a CPU and GPU.

The last tile, labeled as “Base Tile,” will likely function like a breadboard and serve to connect the other parts together. This is achieved using 36-micrometer Foveros Direct interconnects, which are also used on Ponte Vecchio.

The Graphics Tile image below helps to show the added flexibility that this design provides. Intel is able to design these tiles with varying amounts of resources to build processors with different amounts of CPU cores, graphics cores, or other resources.

(Credit: Intel)

The whole design marks a departure from tradition, but some of the technologies are actually backtracking a bit. Components like the memory controller that have long been integrated into the CPU are now breaking off once more into separate components—albeit closely tied to the CPU. Though it seems a bit contrary to prevailing wisdom, we can’t deny that chiplet designs have promise and that they appear to be the way things are going to go in the future, with both AMD and Intel adopting chiplet designs.

(Credit: Intel)

Unfortunately, it may be some time before we learn more details or see performance numbers from Meteor Lake processors. Intel is expected to launch its 13th Generation Raptor Lake processors sometime this fall, and it’s likely Meteor Lake won’t launch until sometime in the fall of 2023. That means it will be even longer before we hear anything about Arrow Lake. And in the meantime, we’ll see more evolutions and revolutions of AMD’s already-chiplet-based tech.