Caveat Emptor: USB4/TB4 cables

I've been running into all manner of problems connecting a USB4/TB4 docking station (Satechi) to my ThinkPad P16V gen 1 laptop. I would get intermittent connections, oscillating connections (going up/down).

Initially I blamed this on the USB4 ports on the laptop. Makes perfect sense to me, the laptop is almost 3 years old, and wear and tear on the port could make for intermittent connections.

For the below, I'll call out what I did, what mistake it was, and then explain the mistakes later on.

I searched Amazon (my first mistake) for USB4/TB4 cables, looking specifically for the right angle connectors, reasoning that the right angle design would place less stress on the physical port. This has worked well for my work machine (a gen 2 version of my personal laptop).

I had noticed (my second mistake) that wiggling the connector in the port would cause drops and reconnections. Again, this aligned with my priors (physical connection problem) rather than looking for other confounding factors (my third mistake).

I had Lenovo send a technician, twice, to fix the ports. The first time, he did a board swap, and things seemed to improve (my fourth mistake). The second time was to fix the problems this tech created on the laptop (it didn't go back together well, had bumps and valleys, etc.). I could have fixed that, but hey, it was still under warranty.

Ok, so here we are. I've got a laptop with USB4/TB4 ports and cables connecting to a Satechi USB4/TB4 dock, so I can put my extra networking doo-dads off of it, and save wear and tear on the ports on the laptop. The issues are still there. Unstable connections. Sometimes working, sometimes not. Usually with a 4 second duty cycle.

Dmesg logs (what, you thought I would run Windows on such a wonderful machine?) showed the connection/reconnections. The extra networking devices being added/removed. I've got 10GbE throughout 1st floor and basement office, with Wifi 6 as backup, and those are in a mesh format with MOCHA 2.5Gb backhaul over unused Coax.

Around mid summer this year (this was well after the tech left for the second time), I spent $30 on Amazon for a cable tester, as I had lots of USB C cables, and the whole 3.0 vs 3.1 vs 3.2 bit was really annoying. Same cable appearance, no way to easily discern what it was. The tester worked great, and I set up a shelving unit with boxes for USB A, USB C in its flavors, and of course the real USB4/TB4. You know those cables that have that lovely "4" printed on them.

I mean, they wouldn't have that 4 if they were not USB4, right? That would be ... misleading ... at best.

I think you might see where I am headed with this.

Brief segue, when you are trained to be a scientist, one of the core aspects of your training is a systematic determination of what you really know versus what you think. Feynman once quipped about biases, indicating that you are the easiest person for yourself to fool with misbegotten assumptions. A systematic test, which is repeatable, is always superior to a guess. Or an assumption. Or reading something printed on the damn cable.

About 1/2 hour ago, I ran down to my office and got my tester. I tested some of the USB4/TB4 cables I just bought from Amazon. I'd had a nagging feeling that I was missing something obvious.

And sure enough, they were USB3.2 or below!!!?! cables. In a product box that read USB4/TB4.

Hmmm. Ok, Let me look at some of the other cables I bought. Tester goes to USB4.2/TB5. I found one that tested positively for these specs. I brought it upstairs, connected to my docking station, and then to the laptop ...

And it freaking worked. No oscillations. No disconnections. Correct behavior.

FML.

Discussions on the interwebs talked about certified vs uncertified cables, and how sensitive the Lenovo laptops are to them.

The mistakes

1) Amazon: Purveyor of occasionally good products, but more often than not, complete garbage. The USB4 cables I bought ... no ... let me correct that ... "USB4" cables, scare quotes mandatory, were not-so-cheap knockoffs of real certified cables. Some of these are nearly as expensive as the certified.

I fell for this. Amazon is simply the high end of Temu, for an American/Western market.

Very much related to this is that the "USB C" USB4 standards are a complete mess.

I'll be leaving appropriate feedback (which I always look at, specifically the 1 ratings, as they are most informative) on their site for the vendors pushing this garbage. But it won't matter, as they come and go, as you can see from their effectively random company/brand names.

We need a label like "certified crap/slop product" to apply to help with Caveats.

2 and 3) Wiggling the connector causing connection oscillations. I did not test the null hypothesis, that not touching the connector would yield the same connection oscillations. Or at least, I didn't do that until today. That is, whether or not wiggling the connector was in any way correlated other than accidentally with the connection oscillations, was a poorly formed hypothesis.

Think of it this way. Your neice shoots a basket in basketball every time you turn your car on. The two aren't causally connected. They may be time correlated but that would be an accidental correlation. That is, unless the universe works in a completely different way than we think.

4) Before and after the board swap, I did not keep careful records of the oscillations. Would be easy to do with a script to pull logs, but hey, I had my priors, and so .... (yadda yadda yadda).

You are the easiest person for you to fool. Just remember that priors are your guesses as to the probabilities. Never completely discount alternative theories, and maybe test one or two. Like leave it alone and see if anything changes. I fooled myself with my own prior biases.

So here we are. View USB cables on Amazon with great skepticism. Spend the $30 USD on the USB tester. I no longer trust. So I don't trust, and I verify.

Annoying that we have to do this, but it appears Amazon lets people sell anything labeled as anything else.

Thus Caveat Emptor, let the buyer beware.

Joe Landman

Joe Landman

OG High Performance Computing (HPC) veteran. Non-practicing physicist, Ph.D in condensed matter theory. Closet mathematician, aspiring astrophotographer. Hubby, Father, Son, Uncle. Human.