Discrete Thunderbolt 5? Intel new Core Ultra Chips are confusing. Am I alone?

General Topics
1

Intel launched new Intel Core Ultra H and HX processors at CES 2025. Those new laptop Arrow Lake CPUs are labeled to deliver “discrete” Thunderbolt 5 support. What that means?

Are we going to be able to have Thunderbolt 5 expansion cards/ports for the upcoming 2025 Framework devices?

Is this only an internal component that can’t be “transformed” into a connector?

This may be a question that the Framework can address if there are talks to include Thunderbolt 5 ports in the next lineup products. Not sure where I should post it.

2

This means that they only support Thunderbolt 4, but will (unsurprisingly) not prevent manufacturers from connecting a separate Thunderbolt 5 controller via PCIe.

1 Like
3

The Arrow Lake-H chips have native support for four Thunderbolt 4 ports and the Arrow Lake-HX chips have built-in support for two Thunderbolt 4 ports.

It is however possible to add a separate controller chip to add additional capabilities. That statement by Intel is referring to the fact that it is possible to add Thunderbolt 5 controllers to gain Thunderbolt 5 ports.

1 Like
4

Thank you for the clarification.

It is unbelievable that Intel is “transferring” the responsibility to add a TB5 controller to the laptop manufacturers. This will just create confusion and quality control issues.

I just don’t get why laptop manufacturers don’t simply go the Oculink route. I understand that it isn’t the same thing, but they would save on license costs by using it instead of trying to integrate Thunderbolt into their devices. Many users would like to have a PCIe connection for an eGPU dock that already exists. Although the hotplug, single purpose, and limited insertion capabilities of Oculink are inferior to Thunderbolt, everything else about it is better to handle.

5

I suspect this is a temporary thing. Intel’s current CPUs are actually divided into 4 separate chips (which Intel calls tiles) and afaik they reused the same IO Expander tile from last gen (the chip that contains the Thunderbolt controllers).

I expect that Intel’s next gen CPUs (in a year or so) will have built-in Thunderbolt 5, however their current generation doesn’t yet.

6

Yes. I heard that TB5 will be part of the next Intel CPU release. I was not expecting any TB5 possibility with Arrow Lake in the H series. This was actually a surprise. I did not understand this nuance that they created in this interaction with the discrete hardware.

I hope manufactures find good supliers to work with. Not sure how the third party firmware will be updated on those devices. Those components should be validated by Intel, like they used to do with their NUCs certifying parts that are integrating with their CPUs. I’m already foreseeing a lot of consumers reporting problems and frustration with TB5 to learn later that the problem is the controler that isn’t working properly.

Nevermind, this doesn’t makes any sense. The controller is from Intel.

7

This actually could be the issue that those people are having.

8

The Thunderbolt 5 discrete controllers don’t need CPU compatibility so much as motherboard firmware compatibility.

I haven’t watched the video you linked however the article mentions 2 issues.

Charging and Thunderbolt 5 function mostly independently using separate wires in the connector. The first part sounds more like an issue with the laptops power delivery firmware rather than a Thunderbolt issue.

Partially a laptop issue, partially an issue with Intel’s requirements for Thunderbolt 5.

Intel requires that Thunderbolt 5 devices support carrying at least two distinct DisplayPort signals over Thunderbolt. However Intel’s Thunderbolt controllers are physically capable of carrying three.

So Maingear only wired it for two distinct DisplayPort signals, however the Kensington dock mentioned is trying to access all three. This results in it being limited to two.

It is worth noting that it is also possible to split a single DisplayPort signal (using a feature called MST) and drive multiple displays from a single signal, however the Kensington dock doesn’t implement that.

The Maingear laptop isn’t the only laptop to forgo including the third signal. For example Thunderbolt 5-equipped Macs with the M4 Pro chip also only support two signals (although I think the M4 Max may support three).

I suspect that when Intel integrates the controllers they will be hooked up for three distinct DisplayPort signals.

9

Hey thanks for adding some more context to the discussion.

I have a follow up question based on your response:

Are you saying that the TB5 discrete controller could eventually work paired with an AMD CPU? I thought there should have some kind of “head control” in the CPU to pair/match the controller with the CPU. In the previous interactions I think AMD would only work with USB3/4v1.

10

Theoretically it should be possible. In the past companies have paired Thunderbolt 4 and Thunderbolt 3 discrete controller chips with AMD CPUs, however it was much less common than with Intel CPUs. I suspect Intel may have more resources and support for helping manufacturers make motherboard firmware for Intel CPUs be compatible with Intel Thunderbolt chips, whereas pairing with AMD leaves it up to motherboard/laptop manufacturers to figure out the firmware.

Although in the context of AMD’s laptop CPUs in high-end laptops like the FWL16 there’s an additional hurdle: PCIe lanes

PCIe lanes are the extremely fast connections between the CPU and other system components (such as discrete Thunderbolt controllers). Each discrete Thunderbolt controller needs 4 PCIe lanes. The current Framework Laptop 16 uses 17 PCIe lanes (8 for the Expansion Bay, 4 for the Primary SSD, 4 for the Secondary SSD, and 1 for the Wi-Fi card).

However AMD’s new Ryzen AI 300 series CPUs have only 16 PCIe lanes (4 fewer than last gen), so the PCIe lane count is already going to be a problem for Framework to deal with even without a discrete Thunderbolt 5 controller.

Intel’s CPUs support 28 PCIe lanes, so they could more easily have a Thunderbolt 5 controller added.

Regardless I expect that the next generation of CPUs from Intel and AMD (which are still a ways away) will have integrated USB4 80 Gbps controllers (Thunderbolt 5 is an Intel certification and brand name for USB4 80 Gbps ports that meet certain requirements). So eventually that will be the solution.

1 Like
11

This is very interesting didn’t know that Thunderbolt could work with other CPUs that weren’t Intel. Thanks for educating me on that.

About the PCIe lanes I was aware about the typology differences between the new CPUs. It was the same with the last generation. Meteor Lake had more lanes and therefore Lenovo adventured releasing a China only Intel Thinkbook model with Oculink and could not do the same for AMD because they didn’t have PCIe lanes enough.

Sad to see that AMD is missing this opportunity again. I hope we see some ultrabooks with TB5 using the new Arrow Lake CPUs.

12

Even with current chips that have the controllers on board it is up to the manufacturers to use them which a lot of times does not or very jankily happen.

In this case tb5 is probably still a expensive enough not to just put on the chips by default and just let manufacturers not use it if they don’t want to.

TB5 and 80gbit usb4 will likely eventually migrate on chip again as the tech matures.

Occulink only does 1 thing thunderbolt does and does so with major drawbacks (hotplug) but cheaper and higher bandwidth.

Egpu use is extremely niche, the main reason people use thunderbolt is for docks.

Not impossible, discrete intel tb3/4 controllers kinda do work on amd platforms even if the firmware support is very janky.

Likely amd will use someone elses 80gbit usb4 or tb5 chips when(if those show up.

Intel certainly put a lot of effort into making that a pretty big pain in the ass XD

13

All good with the inputs but I might say something about the line below:

Yeah, I’m not sure that I’m cool with this way of putting the message out. I keep hearing this being propagated over and over: “eGPU use is niche.” Why?

So far, I believe that the interest in eGPUs hasn’t really skyrocketed because of technological limitations. If we overcome those, I think things might change.

People who game on powerful stations build desktops with high-performance GPUs, or they get powerful gaming laptops with dedicated GPUs. Creators typically tend to stay within the reliable realm of Apple. Those who don’t care about GPU performance—which is a massive number—aren’t interested in eGPUs at all. As you said, they only look for a dock that works. However, the previous groups, and even some others that I couldn’t list, may have some interest in it. Let’s face it: if we can make a Thunderbolt 5 dock that isn’t an eGPU, how much more difficult is it to create a Thunderbolt 5 dock that includes an eGPU?

If the technology reaches a point where we can easily extract very good GPU performance equivalent to desktop use, we would be accessing a new realm of possibilities.

The market of people who may have a desktop for gaming and an ultrabook for portability might see a chance to experience something much less cumbersome. One could have the possibility of a single laptop that is light, small, battery-efficient, thermally efficient, and quiet. We don’t need to carry a powerful GPU everywhere; when we game, we are usually at a desk in front of a monitor.

The only ones potentially not interested in unifying consumer devices may be the PC industry.

The solution could be modular: an upgradable laptop like Framework paired with an eGPU that is far better than a dedicated GPU for both the laptop vendor and the consumer could represent a new world. Laptop iGPUs are becoming efficient enough to accommodate the potential need for some GPU power on the go.

Users would have the flexibility to upgrade their graphics performance without needing to replace their entire laptop. This modular approach allows for customization based on individual needs, whether for gaming, content creation, or professional applications. As more manufacturers adopt these technologies, the market for eGPUs will likely grow, leading to better products and more competitive pricing.

Again, this is my vision of the future—something aligned with what we are seeing happen with handhelds and mini PCs. People are attracted to mobility and simplicity. New laptops with efficient iGPUs and Thunderbolt 5 plus eGPUs could be the next leap in home computing.

In my view, one area for improvement is how we sometimes, albeit unintentionally, overlook the benefits of eGPUs and fail to foster communities that genuinely excite and engage newcomers. The existing online communities focused on this topic often lack the resources and knowledge to effectively welcome and nurture the interests of new members.

14

You did ask why manufacturers don’t “simply go the occulink route”.

I was not downplaying how cool egpus are but they are just a side use of the thunderbolt ports so you can’t just replace them with occulink. It’s not the egpu port that can also do other stuff, it’s the other stuff port that happens to be usable for egpu.

That is certainly a major factor.

Significantly more difficult and especially expensive.

With current trends compute power seems to be improving faster than thunderbolt bandwidth so I hope that changes.