> Neither of these has any chance of being your problem. > > Its likely the hardware; you really dont know whats going on with the bus > on the machine. It could be the driver (which is clearly designed and > optimized to run on an i386 platform), or just some OS related delay in the > interrupt processing. You cant put a rocket engine on a bicycle and expect > rocket level performance. > > Also make sure you are negotiating a gig link. It almost sounds as if you > have a 100Mb/s link somewhere along the line. Hi, Dennis, Yesterday, I borrowed a somewhat faster bicycle (an MVME5100 with a 500 MHz PowerPC 7410 CPU) and put my "rocket engine" on it. Here is a comparison of its benchmark numbers with the slower CPUs: MVME5100 series, 500 MHz 7410, 100 MHz memory bus: MTU [bytes] Non-zero-copy [10^6 bytes/s] Zero-copy [10^6 bytes/s] 1500 65.8 94.1 9000 80.4 119.5 16000 78.9 92.8 MVME2300 series, 333 MHz 604r, 66 MHz memory bus: MTU [bytes] Non-zero-copy [10^6 bytes/s] Zero-copy [10^6 bytes/s] 1500 13.5 20.9 9000 14.3 21.3 16000 15.1 19.5 MVME2600 series, 200 MHz 604e, 66 MHz memory bus: MTU [bytes] Non-zero-copy [10^6 bytes/s] Zero-copy [10^6 bytes/s] 1500 10.4 11.8 9000 12.8 14.4 16000 13.0 14.7 As you see, the MVME5100 board is able to saturate the gigabit link, at least with large frames and zero-copy. However, the performance doesn't seem to scale down in any reasonable way. So, it's not a problem with the test setup, and it's not simply the fact that it's a PowerPC. Thanks, -- Fred -- Fred Gray / Visiting Postdoctoral Researcher -- -- Department of Physics / University of California, Berkeley -- -- fegray@socrates.berkeley.edu / phone 510-642-4057 / fax 510-642-9811 -- - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
