[EMAIL PROTECTED] wrote:
> Flemming
>
> Have done MANY tests with every combo of boards you can think of plus
> version of code with Linux. I can tell you the numbers below are right on
> with the tests you are doing. I am sure you can play with the settings on
> iperf to push these numbers a bit higher to around 28-29Mb but that's about
> it.
>
> The issue is that the Atheros wireless driver relies heavily on the CPU
> where the Ethernet does not, this is one reason why the cards are lower cost
> to make these days. For example we tested and use a XScale 533Mhz core
> design and we maxed out at about 41Mb with a 54Mb link (all Atheros advanced
> options on using a pre-compressed 100Mb file) which is great using iperf,
> doing the same test via Ethernet was 95Mb. With this test CPU load never
> maxed out so we know we maxed the drivers performance.
>
> To answer your question we tested a Prism54 chipset card a while back and
> these used less CPU then Atheros, but because of costs I don't think you are
> going to see this going forward. For example cards from Ralink we are
> testing also put load on the CPU via the drivers.
>
> Anyone can easily do the same test with two X86 computer running a Linux
> build etc so the CPU is not the issues.
ath cards do use more cpu cycles in certain cases (e.g. beacon
processing) but you can easily get everything in the channel which this
person isn't even seeing. All I can say is that there are many people
running freebsd and netbsd that get better performance than has been
reported here. Hell I was getting 20+ Mb/s bridged as an ap on a 4511
(under freebsd) and that's a gutless platform compared to a 4801. So
stop with this FUD. If there are issues it's with the linux code and
this is not the proper forum.
Sam
_______________________________________________
Soekris-tech mailing list
[email protected]
http://lists.soekris.com/mailman/listinfo/soekris-tech
