Hi Mitch, I will have to have someone here to help me get both lspci -vvv for you.
To be clear, let me state it again, the 82573 NICs based system is dual core 1.6GHz CPU system; The 82546 based on Dual Xeon 3.6GHz with PCI-X 133 66it buses designed to have PCI-X buses on the north bridge. I hope I am not that stupid to expect 33bit PCI bus to get many gigabits throughput. When I apply the load, the 82573 based setup scale up well. When I apply 1GB/s load, it will get 1GB/s throughput; when I apply 2GB/s load, it will pass 2GB/s throughput. For the 82546 chip, When I apply 1GB/s load, it gets 1GB/s load at 25% CPU usage. When I apply 2GB/s load, it still passes 1GB/s throughput at 50% CPU usage. When I apply 3GB/s load, it still passes 1GB/s throughput at 80% CPU usage. Software setup on both system are the same way. Only difference is 82573 uses E1000E driver and 82546 uses E1000 driver. Thanks! Wayne -----Original Message----- From: Williams, Mitch A [mailto:[email protected]] Sent: Friday, March 27, 2009 2:47 PM To: [email protected]; 'Jay Vosburgh' Cc: [email protected]; Ronciak, John Subject: [work] RE: [E1000-devel] Bug report E1000 driver bonding in 802.3ad mode can not go beyond 1GB/s throughput >is why E1000E driver and 82573 NIC bonding can get 2GB/s with two NIC >bonded, but Pro/1000 MT server >adapter cannot break the 1GB/s limit. We actually tried also >with CentOS >5.2, it seems true for that >also -- E1000 driver with 82546 NICs bonding in 802.3ad cannot >get more than >1GB/s through it. the tests >are with same clients, same servers, same switch, and same >network cable >wires. Please send a dmesg log and the output of lspci -vvv on both configurations. The 82573 is a PCI-Express part, and the 82546 is a PCI-X part. Most newer systems that support PCI-Express have fairly limited bandwidth for their PCI slots. It may be that you're running into a bus bandwidth issue. If your two 82546 boards are in 32-bit slots in a shared bus you will never get 2 GB/s out of them no matter what you do. -Mitch ------------------------------------------------------------------------------ _______________________________________________ E1000-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/e1000-devel
