Inspired by strong customer desire. The original G.fast plan for 8-16 ports isn't nearly enough now. BT and others have decided to use far fewer terminals, each with more lines. Vectoring 48 or 96 lines is not a trivial problem. Imagine the calculations for so many lines running at 500 megabits over 106 MHz. The heat alone is a problem for small field terminals. It will require at minimum semi-custom chips optimized for the application.
Huawei is confident that "long reach" G.fast will soon see improved performance. Confirming the early field trials at BT, they expect it will be practical to increase the power used by a few dBm. My cable friends are working on 2 gigabit systems, possibly using full duplex technology.
Sckipio suggested last year they were close to vectoring 48 lines using dedicated processors. Now, Huawei's Jack Zhu tells me they have a 96 port unit on the roadmap for Q4 of this year, based on Broadcom chips.
A master vectoring engine will control four boards with 24 ports each. Zhu spent nearly a decade supporting access networks at BT for Huawei and closely watches what the carriers require. Speaking from China, he tells me the Europeans want both more ports and higher performance at 300-500 meters.
Interoperability remains a problem, although Swisscom has had some success using CPE from a different vendor than the DSLAM. The Broadband Forum continues interoperability testing with the University of New Hampshire, but the carriers need to be more demanding to get the chipmakers to work together well.
Britain has the greatest diversity, testing DSLAMs from Alcatel, Huawei and Adtran. I believe the Alcatel and Huawei units are using Broadcom chips and Adtran Sckipio chips. Broadcom's chip supports VDSL as well as G.fast, a major advantage in mixed networks. Although Sckipio's chip may have better performance, many customers prefer the flexibility of continuing support of VDSL.
Once 48 and 96 ports ship, watch the big telcos ask for more.