Intel to Strike Thunderbolt in CPUs

With speeds of 40 Gbits/second, plug-and-play support and a single unified cable design, Thunderbolt should be the most adopted and accepted cabling interface in the market today. But sullied by a quagmire of complications around ports, licenses, hardware integration and more, its uptake has been somewhat stagnant.

When Thunderbolt was first released, the motherboards, systems and notebooks that integrated it had a significant price premium. With Thunderbolt 2 and now 3, costs have come down some, but accessories like expansion docks and external storage are still significantly more expensive than USB 3.0 or USB 3.1 counterparts.

Unfortunately there’s a large amount of confusion in the market around the Type-C connector USB 3.1 and Thunderbolt 3 share. Type-C is a form factor of cable and port that can be used for USB 3.1 Gen1 (essentially running at USB 3.0 speeds), Gen2 (at up to 10 Gbits/s) and Thunderbolt 3 (up to 40 Gbits/s with other features).

Many systems for sale today have USB Type-C ports that may support USB or Thunderbolt or both. To know the difference, the consumer is asked to read a tiny icon that differs from vendor to vendor on the side of their machine. In some instances, a machine might have multiple kinds of Type-C ports on it.

Another drawback for Thunderbolt 3 is that support for true plug-and-play compatibility varies depending on the chipset on either side of the connection. TB3 promised USB-like connectivity without rebooting systems, but in some cases that process only works one-way (plugging in or unplugging) or it might be slightly unstable.

The low quantity of Thunderbolt accessories on the market is another concern. External connectivity docks have been slow to come to market, external storage options are slim at full TB3 speed (and are expensive when found), and the number of Thunderbolt-ready displays is small.

Intel thinks it has solution. The company will begin to integrate Thunderbolt 3 directly onto the CPU although it hasn’t specified when. Because the underlying architecture of Thunderbolt is PCI Express, Intel could either add additional PCIe lanes to future architectures or it could redistribute the existing ones into the necessary Thunderbolt processing logic on-package or on-die.

This move will lower both cost and power draw. OEMs will not need a separate, external Thunderbolt controller and tracing on a motherboard, enabling thinner, lighter systems with longer battery life. I also assume integration will lower overall power consumption of the combined CPU+Thunderbolt product if Intel builds the controller on the same process technology node as the CPU.

We should see a near universal adoption of Thunderbolt 3 on platforms using these processors. Rising numbers of Thunderbolt systems will inspire more vendors and accessory companies to build more products for the interconnect.

Intel is also changing the licensing policy for Thunderbolt, moving to a non-exclusive, royalty-free license. This removes another cost restriction, allowing third parties to build Thunderbolt-capable controllers for storage, displays, docks and more without worrying about associated fees. Without the need for an Alpine Ridge chip from Intel, more accessories at lower prices will grow the market.

No one can predict when this will happen until Intel says when it will release the integrated CPUs. Nevertheless, I fully expect this move will drive Thunderbolt into mass adoption and even to the universality we see in USB today. External storage, networking connectivity, high resolution HDR displays, external graphics solutions, productivity docks and many more products should flourish.