Windows 2008 R2 / NUMA with >64 cores

I have a a Windows Datacenter 2008 R2 Server configured with 4 nodes, each
with 8 six core processors for a total of 98 cores. I am running into a
whole mess of performance problems so serve that a basic 8 core server with
12 GB of memory is running twice as fast.

By default, when a server has more then 64 cores, it assigns them evenly
into groups with a max of 64 CPU per group. So in my case I have 2 CPU
groups with 48 CPUs each.

Now the problem I am running into is that window OS applications, like
perfmon that will only see the CPU assigned to group 0. In this case , I
want to use perfmon to monitor all the cpu, but it will only show be the
first 48. What do in need to do to be able to monitor all the cpu?

I also noticed that group 0 has cpu from node 0 and 2, and group 1 has cpu
from node 1 and 3. Now if applications are going to be access the processors
by their groups, I would want the group to contain nodes that were direct
neighbors to each other (i.e node 0 and 1) so that the application/threads
can access the memory that is closest to the CPU it is running on.

Thx

--
-----------------------------------------
William F. Kinsley
NextGen Healthcare Info. Sys. Inc.