From: "Robert Comer" 

> It seems to be something in the protocol and/or sockets implementation.

Could well be...

- Bob Comer


"Gregg N"  wrote in message
news:428a448f$1{at}w3.nls.net...
> Robert Comer wrote:
>> I thought maybe that was the case, but I got lots worse numbers using
>> a smaller -b on my 100Mbit ether, especially with larger packets.  the
>> difference between larger packs and smaller ones wasn't near as big
>> with -b set to 100M or higher.
>>
>> fragmentation and collisions from the extra traffic is probably what's
>> making the difference, but I'm not losing near the performance you
>> are... (90Mbit for 1024 vs 80Mbit for 1025)...
>>
>
> There shouldn't be fragmentation with a 1500-byte MTU. To eliminate the
> effect of collisions, you can use the loopback interface and see similar
> results. I got:
>
> -b 1000m -l 1024 ==> 250 Mbps
> -b 1000m -l 1025 ==> 200 Mbps
>
> It seems to be something in the protocol and/or sockets implementation.
>
> Gregg
>

--- BBBS/NT v4.01 Flag-5