My standard wireless test rig is an Acer Aspire One netbook into which I have installed an Intel Centrino Ultimate N Wi-Fi Link 5300AGN mini-PCIe card. As I explained in my original post about this upgrade, I've also added the third antenna to get the full 3-stream MIMO (450Mbps) capability.
Until now, for reproducibility reasons, I've been using the driver that I first installed, which is the Intel driver version 13.5.06, dated 19/1/2011 (released just a few days before I upgraded). But 18 months is a long time in the driver world, so I thought I'd see what new drivers were available and give them a try.
The latest is version 15.2, dated 20/7/2012. I installed this and ran my usual tests in Passmark Performance Test 7, using the advanced networking test module. I was puzzled to find that performance on the 3-stream, 450Mbps router I was testing dropped significantly, but more worryingly, the CPU utilisation as measured in Passmark was pegging itself at 100% for most of the duration of the tests.
This effect of this was to cap throughput at about 30Mbps on both 2.4GHz and 5GHz frequencies - the throughput graph was mostly a flat line, with few of the usual peaks and troughs. With the original driver, depending on the router and local interference, I rarely see more than about 60-70% CPU utilisation (the Aspire One only has a 1.6GHz Atom CPU, so this is not surprising), and I have measured wireless throughput of well over 100Mbps using this adapter (with a Linksys E4200 v1 router).
I downgraded to the previous driver version, 14.3, and found the same problem. So I went back through the drivers one by one, and found that this behaviour seemed to start with the 14.x series of drivers, which first appeared in January 2012.
I have no idea why this is happening, so if anyone has a clue I would be interested to know. But for now, I have reverted back to a known quantity, which is the 13.5.06 driver.