We are starting to form a picture that Broadcom’s launch into G.Fast last week was not nearly so strong as Sckipio’s and that it has made the dreadful mistake of trying to buy itself more time.
The G.Fast launch which Broadcom orchestrated at Broadband World Forum seemed simple on the surface, until people started getting home from the show and began to compare notes. The demonstrations that were on offer are at the heart of the matter and the rest may well be supposition, cynicism and vested interest, but it’s all worth reporting.
It turns out that Lantiq in Germany has pretty much thrown in its lot with Sckipio, and that makes it a virtual certainty that the reference platform that it launched around the Israeli chip, three weeks ago, is the forerunner of the Adtran devices which will make up the bulk of Deutsche Telekom’s broadband strategy as it goes into vectoring – both VDSL and G.Fast, during 2015.
The demonstration at the stand was typical of how Sckipio expects the chips to be used. There was one with 16 Sckipio CPE devices which were all 50 meters away, attached to a single device with four of the Sckipio central office chips in, each of which supports 4 CPEs. Initially it was run without cross talk cancellation, and then when the cancellation was switched on it ramped up to 700 mbps for EACH of the lines.
But the two other G.Fast demonstrations were nowhere near as convincing. A single device at one end, another at the other, and a needle that went up to just 500 Mbps in between.
Now we know that G.Fast chips are supposed to go all the way up to 1 Gbps not just 700 Mbps and certainly not 500 Mbps. But the two that resulted in a single line at 500 Mbps certainly did not show cross talk actively being cancelled, because there were no other lines to throw up cross talk. And it is clear that the Sckipio chips could go up to 1 Gbps, except all 16 had a load on them at once, and we suspect that when a line is tested on its own it would reach that speed, and in fact we were told that some demonstrations showed the 1 Gbps being achieved on this rig, when some of the devices were turned off.
This is the type of performance that gets better as the technology gets more mature, but the lower level 500 Mbps demonstrations without cross talk, included the Broadcom device and another by Alcatel, and there is the suspicion that these solutions are nowhere near as close to being ready for deployment as the Sckipio solution.
Analysts and journalists at the show were shown around the Sckipio booth, but a number of them told us that were not allowed into the “customer only” viewings in the other booths, and that immediately raised hackles and suspicions.
The feeling is that the field trials that Alcatel certainly has in play are being conducted with a pre-production chip, likely an FPGA of its own design, and there is a rumor that the Broadcom chip is a chip that was used for central office termination of broadband lines in VDSL vectoring systems, which has simply been adapted to reach the entire 106 MHz, without re-engineering the rest of the silicon to keep up, for instance, with the more rapid cross talk cancellation that will be needed. In other words a holding action, not the real deal.
So while we spent part of this week trying to find out which chips Alcatel and Huawei in particular will use, the probably answer is that no-one knows, because these companies do not know.
A video of the two CEOs of Sckipio and Lantiq talking at the show has been put on YouTube and it refutes the Broadcom statement really clearly when it comes to co-existence between VDSL and G.Fast. Broadcom last week said that it is the only company that can do this, while the two CEOs, David “Dudi” Baum of Sckipio and Dan Artusi of Lantiq, made it clear that they felt they were also able to make this claim insisting that this was the only way that Telcos could economically install it, by upgrading one line to VDSL and another to G.Fast, in the same bundle.
Another issue was the power consumption. Sckipio has clearly said that its system requires 1.5 watts per line. Broadcom said its chip was “lowest” power, but declined to say what power was needed to run it. Most critics say that this is what needs to be measured, since its power drain will give it away if this is a temporary fill in product. The Broadcom solution uses two chips – a DSP and an analog front-end for VDSL in its central office system, and we can only assume that this is a temporary measure and that it probably uses more, rather than less power, than the Sckipio system.
On the other hand, given that Broadcom said that it was shipping in volume as of now, then someone will presumably order some, and tell us what power they use? And if it turns out that you can’t (order them) – then perhaps the announcement was just a spoiler to keep Sckipio at bay.
That’s not game, set and match for Sckipio by any means, but the impression we may have given readers last week was that with Broadcom announcing just a few weeks after Sckipio, then it was still in pole position to take dominance in this market. Now that doubts have emerged over these demonstrations, we’re not quite so sure.