Thunderbolt Is Seeing A Lot Of Improvements For Linux 5.2

[nils_]

Even with PCIe 5.0 lanes ( 4Gbit/s bidirectional) you would still need like 10 lanes to provide a single 40Gbit/s link for USB4.
With PCIe 4.0 you would need 20, and so on, doubling for each PCIe revision you decrease.
This is just silly, even on a tower PC.

You're mixing up bits and bytes there, PCIe 5.0 can do (almost) 4 GiB per second, so almost 35Gbit/s. Though 40Gbit/s is less than what Intel provides in their connect to the PCH (I believe it's currently 4 Lanes of DMI which is similar to 4 PCIe 3 lanes).

[starshipeleven]

And then of course many CPUs don't have the PCIe/DMI lanes to support a larger number of ports (or even more than one).

Yes, that's one of the main reasons USB 3.0 didn't go much beyond the first gen in real life.

Even with PCIe 5.0 lanes ( 4Gbit/s bidirectional) you would still need like 10 lanes to provide a single 40Gbit/s link for USB4.
With PCIe 4.0 you would need 20, and so on, doubling for each PCIe revision you decrease.
This is just silly, even on a tower PC.

These high-speed interfaces will get any traction once they are integrated in the main CPU die, just like PCIe, RAM and whatever else controllers.
There you don't have bandwith issues as the controller will be on the on-die interconnect bus.

[nils_]

Yes but it won't matter just as it didn't with USB 3.0 revisions.

99.99% of current USB 3.0 devices aren't using controllers of a higher revision than the bare minimum needed to be USB 3.0 (aka 5Gbit/s), and even if they did it's irrelevant as most USB 3.0 host ports are still not really 10Gbit/s anyway.

And then of course many CPUs don't have the PCIe/DMI lanes to support a larger number of ports (or even more than one).

[starshipeleven]

Also a new Type-Ca cable is in the work that is non-reversible to restore the classic user experience of the 3xflip'n'plug™

[starshipeleven]

They'll get to that in due time my good man

Yes but it won't matter just as it didn't with USB 3.0 revisions.

99.99% of current USB 3.0 devices aren't using controllers of a higher revision than the bare minimum needed to be USB 3.0 (aka 5Gbit/s), and even if they did it's irrelevant as most USB 3.0 host ports are still not really 10Gbit/s anyway.

[nils_]

I'm fairly certain the next iteration will be named USB 3.3 x2 +3 TurboSpeed v2. Once it's released, they will retroactively rename the previous versions of course. But to avoid consumer confusion, they will use the marketing names SuperSpeed, UltraSpeed, MegaSpeed, and SpeedySpeed to differentiate them.

Also a new Type-Ca cable is in the work that is non-reversible to restore the classic user experience of the 3xflip'n'plug™

[nils_]

*USB4

And it's a good thing. I'm still having nightmares about the USB 3.2x2=6.4 Gen2 or some shit names they chose.

They'll get to that in due time my good man

[starshipeleven]

I don't think that "incompatible NIH" is applicable to a technology that's one of a kind. As of now, there is nothing for Thunderbolt to be potentially compatible to.

Wrong.
"Incompatible NIH" fits Intel's Thunderbolt because until they opened the spec the only available thunderbolt host controllers were designed by Intel and worked only on (some specific) Intel hardware. Sure you could connect any kind of client devices to your specific Intel hardware device.

Now that they opened the spec we might start seeing Thunderbolt host controllers that can run on AMD, or ARM or whatever else. Then and ONLY THEN we can talk of actual compatibility, standards and all that.

[torsionbar28]

Good thing Thunderbolt is going to be merged into USB 4 (or whatever the next version is called).

I'm fairly certain the next iteration will be named USB 3.3 x2 +3 TurboSpeed v2. Once it's released, they will retroactively rename the previous versions of course. But to avoid consumer confusion, they will use the marketing names SuperSpeed, UltraSpeed, MegaSpeed, and SpeedySpeed to differentiate them.

[shmerl]

However, I'm slightly worried, how this "merger" is ever going to work? Does it mean that any device that's going to formally support USB4 will have to implement PCIe? That's not gonna happen.

Not any different from how USB-C works now. Not every device that uses USB-C supports DisplayPort. So I don't see how PCIe is any worse. Device can support a subset or all features - the connection is still the same.