On Wed, Dec 28, 2005 at 09:22:09PM +0100, Aritz Bastida wrote: > Hi.
Hi, > First of all, thank you. It seems the problem may be in there. The PCI > buses are actually quite slow. As I told before I have a Pentium 3 > with a e1000 card which achieves 350kpps at most. I have also a > Pentium 4 with another e1000 card (the specific models were said in > the prev mail). This can't either send packets faster than 350kpps. > > But may be the reason is the PCI bus: With the patch applied I found this: > 0000:00:0c.0: e1000 (PCI:33MHz:32-bit) 00:0e:0c:76:d3:75 > > Both NICS are attached to 33MHz pci buses. Comparing with the results > in the pktgen paper, this is the slowest possible (the fastest would > be 133MHz). Could be that the reason for such a bad throughput? Sure, very well possible. > Shouldn't the Pentium 4 send faster than the Pentium 3 anyway? Nope :) If anything, the P4 would be slower than the P3 since it most likely has higher MMIO latency (in English, the 'distance' between the CPU and the NIC is greater, in terms of clock cycles.) For ages, Robert's top performing pktgen machine was a Pentium 3 box. > As I told you also, I have another machine, a dual AMD Opteron with a > tg3 which sends at 650kpps at most. Is there a similar patch for tg3 > which lets me know the bus types/speed? Doesn't it do that already? If not, check out ethtool -- there might be a way to extract the data with that. cheers, Lennert - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
