The following has be reported by IBM LTC: [perf][netperf] poor throughput of e1000 in RHAS 3.0 B1 Hardware Environment: (2) x335's (2.0 GHz) NIC on client cabled directly to like NIC on server specific NIC defined in results section Software Environment: RHAS 3.0 B1 netperf-2.2pl3 TCP_STREAM test results are averages of 5 60-second runs per send msg size all standard deviations were acceptable (typically much less than 0.20%) MTU 1500 Steps to Reproduce: 1. Install RHAS 3.0 B1 on client and server 2. Install netperf-2.2pl3 on client and server 3. Run netserver on server side 4. Run 5 60 second netperf TCP_STREAM tests from the client 5. Average the throughputs of all 5 runs, check for acceptable std dev Actual Results: Throughput (Mbits/s) per send message size 1K 2K 4K 8K 16K 32K 64K tg3 530.62 684.11 809.83 911.62 938.21 940.49 940.50 e1000 475.92 551.26 556.32 554.73 552.14 547.37 470.44 Expected Results: The throughput for the e1000 driver included in RHAS 3.0 B1 should more more closely match that of tg3. Additional Information:Glen / Greg Talked to Barry, he wants problem reported to RedHat. Thanks
------ Additional Comments From barndt.com 2003-29-08 10:34 ------- kernel version - vmlinuz-2.4.21-1.1931.2.349.2.2.entsmp
------ Additional Comments From barndt.com 2003-29-08 15:11 ------- Bug does not exist in RHEL AS 3.0 beta2, results are very good.
yay! Based on your comment, I assume there's no objection to closing this issue as resolved. If this issue needs to be reopened feel free to do so.