MB> Performance problems in what way?  I was shocked that I could push the
 MB> server to 85% utilization with a single client doing a lot of file
 MB> copying,
What's your segment utilization? I have no problem getting a Pentium class 
computer to generate 1100K/sec traffic, which is a fully utilized segment 
anyway. The only antidote I've found for heavily-used piggies (MS Access, 
Foxpro) on Pentium networks is 100-TX. And, on a particular server I work on, 
I can push file copies at 100Mb at 5MB/sec with only 3% utilization. The 
server uses PCI (obviously) network cards and PCI/SCSI hard disks. Even 
small-packet benchmarks (64-byte disk reads) don't affect this server.
As for your W95/W3.11 comparison, that seems most likely a difference between 
the low level drivers or other fundamental factors, since there's nothing in 
the W95 stack that would radically change the efficiency of talking to IO 
registers. Also, my testing hasn't shown any marked throughput differences 
between Dos, WFWG and W95.
 MB>  where the same operation would not have broken 20% under
 MB> Windows 3.11.  I don't know if this was packet burst or large packets or
 MB> what, but the difference was striking.  If the Microsoft client is any
 MB> faster, the server would be pegged.
A Netware server? I can pipe 10000 packets per second at under 10%. I don't 
think this is a W95 issue.
Jeff
--- GoldED/2 2.42.G0615
---------------