The publication
About Root Complex
Root Complex reads public specifications and vendor documentation about PCI Express and writes down what they say, in the order a working engineer needs them.
What this site is
Root Complex is an independent publication about PCI Express and the interconnects that grew around it. It exists for one kind of reader: the person holding a datasheet, a board or an error log who needs to know what a figure means before deciding anything. Every page is written to be read once and used, not to be skimmed.
The site covers six subjects, and it covers them in the same order every time: what a lane and a link are, what each generation changed, which connector holds the link, how the hierarchy is shaped, how a part is tested, and which product categories a solutions catalog is built from. Two standing references sit underneath all of it, the link speed table and the glossary of the link, and every article points back to them.
What this site is not
This publication is not the PCI Special Interest Group. It is not a working group, a standards body, a trade association, a member organization or a chip vendor. It does not publish, sell, license or distribute any specification, and it cannot grant, withhold or interpret compliance for anyone. Where a specification is the authority, the page says so and sends the reader to the body that publishes it.
The domain carried a developer network for PCI Express between 2002 and roughly 2017. That organization is gone. Nothing of its membership, its roster, its mailing list or its people is reproduced here, and no continuity with it is claimed. What is reused is the subject, which nobody owns, and the addresses that still receive links, which are answered honestly rather than left to rot.
Who writes it
Pages are signed by the desk rather than by an individual name, because they are edited collectively and revised without a byline change. The desk is reachable at one address, which is on the contact page, and corrections sent there are treated as the highest priority work in the queue.
How a page gets published
A page starts as a question a reader could plausibly type. The desk then collects the primary documents that answer it, records each figure with the document it came from and the date it was read, and only then writes. A claim that cannot be tied to a document either gets a plainer formulation or does not appear. That procedure, including what happens when two sources disagree, is written out on the sources and method page.
Nothing on this site is generated on a schedule. A page is revised when a specification revision, a vendor document or a reader correction makes it wrong, and the revision is announced with its date on the bench notes page. Silent edits to a factual claim are not made.
Why this address
This domain carried a developer network for PCI Express from 2002 until the middle of the 2010s. Documents printed in that period still point at it. A chipset datasheet from 2004 sends its reader to one address on this domain for guidance on receiver compliance measurement, and a vendor product page still online today links to another for a catalog roster. Those two addresses answer again, at their exact historical form, with material written now from sources that can be checked now.
Every other address the domain once served is handled from a published table rather than by guesswork. Where a page here genuinely covers the same subject, the old address redirects to it. Where the content belonged to the previous organization and is deliberately not taken over, the address says the content is gone. Everything else answers not found, because sending an unknown address to the home page is a way of pretending to have content that does not exist.
The alternative was to leave a purchased domain parked while documents in circulation kept pointing readers at a dead server. Answering those readers with something accurate seemed better than answering them with nothing, and it is the reason this publication exists at this address rather than at any other.