Wireless LAN vendors using the same silicon claim different results. I thought I'd ask the silicon vendors what's going on.

As we all know by now, dual-radio Wi-Fi uses more power than 802.3af can deliver (unless we believe Siemens' claims to have fixed the problem - and I don't). Everyone has a different answer, but everyone agrees that future silicon will make the problem much better, as power demands will be reduced.

I paid visits to both Marvell and Broadcom at the Mobile World Congress and got into quite technical discussions with people at both stands. I'm still not 100 percent sure I understand where we are with this, but I think the consensus is that there is going to be a lot of improvement with future generations of silicon.

The move to 65nm CMOS will reduce the power requirements in the baseband part of the chip, both companies agree. But, I asked, there must be a certain increase in the power requirements for amplifiers, which is unavoidable? Dual band APs and MIMO both mean more antennas - and each one needs to have a signal amplified enough to get it out and across the air.

This is true, said Chris Bergey at Broadcom, but as the output signal strength is only 300mA, there must be a lot of scope to trim down the power put in, before the output signal is cut down. He reckons, I think, that a dual band AP can be done with only about 15 percent more power required over a single-band one. I trust him, since Broadcom is not currently doing any significant business in connected enterprise APs that I know of, so the company probably doesn't have customers demanding this.

I had a longer talk at Marvell, and the upshot there is that overall, for a given throughput, 802.11n will cut energy use, not increase it, since each bit is sent faster, and fewer Joules are used. That's not particularly relevant, since peak power could still be higher, and anyway, throughput will go up.

Marvell - and I'm sorry but I lost the card of my contact there - reckons that there's plenty of scope to improve the amplifier performance - including integrating the amplifier with other chips - so, again, we expect 802.11n to come in under the 802.3af power.

So the higher power, 802.3at spec may not be useful. Which will be useful, because in another conversation this week, Gigabit switch vendor Force10 told me that 802.3at may be a bit of a red herring since most wiring closets won't have enough power to support many ports of it!