What a USB-C Cable’s 80 Gbps Rating Actually Promises

USB4 and Thunderbolt 5 share a connector, a negotiation protocol and, on a marketing slide, the same impressive number. They do not share the same list of things a product is required to do. The gap between mandatory and optional is where the money goes, and it is why the cable in your drawer can be simultaneously correct, certified and the reason your drive is running at a third of its rated speed.

Two Specs, One Connector, Two Contracts

USB4 is written by the USB Implementers Forum and built on the USB-C connector with Power Delivery as the entry ticket: nothing happens at USB4 speeds until the two ends have negotiated a power contract and agreed to enter the mode at all. The minimum a USB4 host must deliver is 20 Gbps. Forty is optional. That single sentence explains most of the disappointment in laptops advertised as USB4 — the port is compliant, and the compliance bar was set where the cheapest silicon could clear it.

Thunderbolt 5 is Intel's specification, aligned with USB4 Version 2.0, and it tightens the contract in ways that cost money. PCIe tunneling is mandatory, so any Thunderbolt port can host an external GPU enclosure or a capture card without the vendor having to say which of its ports support it. DisplayPort tunneling, cable certification and a minimum guaranteed bandwidth are baked in rather than offered.

The practical difference is not a spec-sheet line, it is predictability. On a USB4 machine you read the fine print or you plug something in and find out. On a Thunderbolt machine the fine print is short, and you paid for that in licensing and in silicon.

Speed Is Negotiated, Not Printed on the Jacket

Every USB-C cable that carries more than 3 A or more than USB 2.0 data is supposed to hold an electronically marked chip at each end that declares its current rating, its supported data rate and whether it is active. The port reads that chip before it commits to anything. If the chip is absent or silent, the correct behaviour is to assume 3 A and USB 2.0 — a device that works fine on a plain charging cable, at four hundred megabits, is not broken. It is obeying the spec.

Above that floor everything is a link-training conversation. The two ends start at a low rate, probe the channel, and step up while the error rate stays acceptable. A cable with marginal signal integrity does not fail; it trains down and reports a slower link. This is why a 40 Gbps drive can negotiate 20, and why a dock that worked on one laptop negotiates a different set of tunnels on another.

The new generation also changes the signalling. Doubling the rate without doubling the connector's usable bandwidth means more bits per symbol, which is why the 80 Gbps generation moves to three-level PAM-3 — more throughput out of the same channel, at the cost of noise margin. Noise margin is exactly what a long, thin, cheap cable does not have. A passive cable that passes at 40 Gbps will not automatically pass at 80, and nothing about the jacket tells you which one you are holding.

Physics Decides What Length You Can Buy

Copper has a length limit and the standards admit it. A passive USB4 cable rated for the 40 Gbps modes tops out around 0.8 m in the specification; Intel rates passive Thunderbolt 5 cables to about a metre at 80 Gbps. Past those lengths the signal is gone and you need retimers inside the cable — an active cable, with silicon at both ends, a bill of materials several times higher, and a small power draw taken from the port.

That is the whole reason a two-metre 80 Gbps cable costs what it costs, and why the two-metre cable in the bargain bin is not the same product with a longer jacket on it. It is a passive cable that will train down, or an out-of-spec passive cable that works on a good day and produces intermittent retraining on a bad one. Intermittent is the worst failure mode available, because nothing in the operating system says "your cable is marginal."

There is a quieter trap in the connector itself. A full-speed USB4 link uses all four high-speed pairs, so a "USB-C 3.2" cable built for a single lane caps you regardless of what the ends can do — and charging cables routinely omit those pairs entirely.

How to Buy a Cable Without Getting Robbed

None of this requires a laboratory. It requires reading the listing the way the specification reads it:

• Match the cable to the mode you will actually run. Charging, USB 3.2 data, DisplayPort video, PCIe tunneling and 80 Gbps are five different products that look identical in a photograph.
• Treat "40 Gbps" and "80 Gbps" claims with length attached. If a passive cable is longer than about a metre and claims the top rate, it is either active or optimistic.
• Pay for certification. The USB-IF logo programme and Intel's Thunderbolt certification both issue testable identifiers; a listing that says "compatible" and shows no identifier is selling you a cable, not a guarantee.
• Buy active, deliberately, when you need length. Active cables are the correct answer above a metre at high rate — just know you are buying silicon, not copper.
• Keep one short certified cable as a reference. Swapping in a known-good cable is a two-second test that eliminates half the possible causes.

The branding does not help. The current USB-IF logos dropped the version numbers nobody understood for a data rate and a wattage — more honest, and still requiring you to know that the number is what the cable was tested for, not what your port will reach.

What I Think

The cable drawer is a compatibility problem we built ourselves, and the fix is not a better standard — it is buying fewer, better-specified cables and labelling them. Almost every "my dock is flaky" story I hear resolves to a cable that was never rated for what it was doing, and the reason nobody suspects the cable is that it charges the laptop perfectly. Power and data are different capabilities with different margins, and the connector is designed to hide the distinction.

If you are buying for a machine with Thunderbolt 5, pay for the certified short passive cable for the desk and one active cable for the long run, and stop pretending the cheap one will do. The spec sheet promised 80 Gbps. It promised it to the cable, and the cable is the part you chose badly.