Already last week we tried the designation trade for the skills of the "Thunderbolt / USB 4"-Sockets of the new arm Macs to explain with the then available information. Meanwhile, the first devices with M1 processor are in customer hands and some have disassembled their specimen immediately. The look inside a Mac mini, the EGPU.Io Forum was posted, brings some light into the dark, but completely clarity looks different.
For example, Apple uses two JHL8040R blocks on the Mac Mini MAIN block, the Intel’s specification database Ark as Thunderbolt-4-Retimer. However, according to ARK, these building blocks have been available since the third quarter of 2019, while Intel himself spoke for the first time in early 2020 Uber Thunderbolt 4 and many details were published in the middle of 2020. In the latter context, Intel presented the Thunderbolt 4-chips JHL8540, JHL8340 (both controller) and JHL8440 (client) – ie other chips than those in the M1-Mac.
In the Mac Mini with M1-SOC are two Thunderbolt 4-Retimer building blocks from Intel.
Of the Ratsel Solution: Retimer building blocks do not contain protocol logic, but only use electrical signals to produce electrical signals via the lines. Since the useful data rate between Thunderbolt 3 and Thunderbolt 4 has not changed, there are no differences in such "stupid" Building blocks or an old retail chip design also works with Thunderbolt 4.
But this does not apply to the logic of the actual controller. Unlike last suspected, Apple has actually integrated the Thunderbolt Controller into the in-house M1-SOC – Chapeau! However, as an open question, however, remains why Apple only applies this with Thunderbolt 3 ability.
Under the potential explanations, the logic does not support the slightly hottest frequencies in which Thunderbolt 3 and USB 4 / Thunderbolt 3 distinguishes is not supported. A retailer is the difference no matter, for a logic to be certified according to a standard, this looks very different. Topography extensions such as hubs instead of daisy chaining, a Thunderbolt 4 controller must also need to be supported.
Also, however, the Intel has launched its Thunderbolt 4 program for Spat for the first M1 products: mid-2020 was allowed to "regulatory" Tests for the now published products finally already run. Or it is attributable to subtleties such as the requirements for the screen or PCIe bandwidth support, which may not be fulfilled by a M1-SOC.
Like that too: that the final naming is confused by Krude, but is completely correct in itself, was already explained. And also that for users hardly reveal differences, because there are hardly any Thunderbolt-4 periphery and the maximum data rate remains the same.