Re: thunderbolt: AMD Phoenix USB4 host advertises Gen3 Support=0 in SB LINK_CONF with TB5 device, link stuck at Gen2

[James R Taylor <jrtaylor11@gmail.com>]

(Resending to the list in plain text: my first copy was rejected for
containing HTML. Sorry for the noise.)

Hi Mario,

On Sun, Nov 15, 2026 at 02:13:37PM +0000, Mario Limonciello wrote:
> I'd lean on more likely a PD controller issue. Did you already look for
> a BIOS update? Most vendors distribute PD updates with BIOS/EC.

Thank you for looking at this. I'm on BIOS 1.06. Minisforum have a 1.07
image for the UM890 Pro with no release notes; its AGESA (PhoenixPI
1.2.0.0a) is the same as 1.06, and I couldn't tell from the image
whether it carries a PD or EC update, so I haven't flashed it yet.

> If there is no update available; can you start with reporting to your
> vendor and let them reproduce and get some PD traces to confirm?

I reported it to Minisforum on 8 November and it is with their firmware
team now. I'll pass on your PD controller suggestion and ask whether
1.07 or a separate EC/PD update changes anything.

Andrei, thank you for testing your SER8 and for laying the values out
side by side. The host-side values are identical on both machines,
including the retimer ID, PORT_CS_18 and the SB LINK_CONFIGURATION
03 d3 03, and your MacBook result rules out the cable and the DEG2. One
difference from Framework #56: the StarTech TBLT5MM50CM240W I have been
using is an active cable and it still trains TBT3, 2 x 10 Gb/s; the
DEG2's bundled cable gives the same result. Like your SER8, the UM890
Pro exposes no UCSI device, so I can't see the PD side from Linux
either.

Thanks,
James