On Tue, Sep 13, 2005 Joe Orton wrote: > On Sat, Sep 10, 2005 at 06:57:02PM +0200, Johannes Stezenbach wrote: > > Hm, that's not the real fix, there seems to be another bug. > > Can you try the following patch: > > I had a play with this since I also see poor quality with my 761 card, > trying cpump values of 1, 2 and 3 for the 470..750Mhz range with your > patch to also do cpump << 6. > > I ran tzap for 60 seconds on a mux at 665Mhz and recorded all the ber > values, extracted into a spreadsheet: > > cpump 1 2 3 > average 13804 6071 5365 > median 9618 5494 4758 > > which is still not exactly great but it seems cpump=3 gives the best > results; I'll do some more testing with this patched driver and report > back later in the week. If there's anything else it would be useful to > test let me know. I committed my patch now unchanged. The cpump=1 value from the Windows i2c dump was apparently wrong, 2 vs. 3 doesn't seem to make that much of a difference so I'd rather not mess with it. Johannes
