Johannes Berg wrote: > Hi, > > Great update, thanks! Good to hear that you got that laptop working. > >> Performance testing is accomplished using Iperf with the server connected to >> my Linksys WRT54G V5 >> AP/router via a 100 BaseTX wired connection. If I boot Windows XP and run >> Iperf from it, I get 19.5 >> Mbs reported. As the Iperf results show considerable jitter, I always run at >> least 5 trials and >> report the maximum. The distance from my test computer to the AP is >> approximately 2m. All bcm43xx >> tests were done with the latest patches applied. > > The iperf tests you're running are downstream tests, right? I mean, > you're doing downloads from that server to the machine with the cards.
No, I was testing the upload speed to the server. My experience is that the cards have always been able to receive better than transmit. Iperf has two switches that are supposed to handle bi-directional testing, but I have not yet made either of them work. >> I'm surprised that amplitude of signal does not correlate with performance; >> however, the 4311 >> amplitude does diminish a lot at the 48 Mbs rate, which probably explains >> the falloff in performance >> at that rate. In addition, the 4306 showed a much poorer separation of the >> individual subchannels in >> the spectrum. > > I think these are probably related. If you are indeed doing downstream > tests (I guess so) then all this means is that now we've reached a power > level where the 4318 can successfully ack frames (which would previously > lead to retransmissions and timeouts of packets etc). I'd think that at > that point the transmission characteristics become less important than > the reception characteristics. I suppose that we need to work on those > for the 4306, but most likely newer chips also just have better > characteristics due to hardware improvements. It is likely a combination of better hardware and more stuff done in firmware, which isn't bothered by the mistakes in the driver. With that ancient pre-g 4306, the quantities known as rotor and retard, which I suspect to be involved in the OFDM subchannel encoding, need to be loaded. Later revisions of the PHY have these built in. > I think it could be quite interesting to run the tests the other way as > upstream tests. I will do the reverse tests soon. Larry _______________________________________________ Bcm43xx-dev mailing list [email protected] https://lists.berlios.de/mailman/listinfo/bcm43xx-dev
