In message <200112241624.fBOGOlB18524@guild.plethora.net>,
Peter Seebach writes:


>In message <009e01c18c78$fdba6560$84469280@Y5F3Q8>, "Robert Schaefer" writes:
>>Would you like to share that configuration?  I can't say as I currently
>>*need* to saturate a gigibit segment, but it would be nice to know how...


[ supermicro 370de6, p3tde6-g, 3com Tigon-IIs ...]
>The
>observed limitation is that it's possible to peg the CPU.  (Annoyingly,
>you can peg one CPU while the other is idle.  This may get fixed in a future
>release; I know that BSD/OS will, in the forseeable future, fix a similar
>limitation in how multiple CPU's are used.)

One idle CPU wouldn't be terribly suprising for a giant-lock kernel,
if one CPU is spending most of its time in the kernel computing
TCP/UDP checksums or servicing interrupts.

If you can spare time to tell the curious: which version of FreeBSD
what timeframe; with or without checksum offload, and which driver(s)?
(wx, with or without the interrupt-coaleace fixes; gx, with or without
jlemon's off-by-one fix...)

Its always nice to have a well-defined target to reach :).