The 2621XM only lists PPS because that is what the CPU is doing. Passing packets... the number of Mbps depends on the size of the packet.
You can overload the CPU using only 1Mbps of bandwidth worth of PINGs. Check if you have any ACL or policy routing on the Interface(s).
When you do a "show process cpu extended" (if using 12.4 IOS or up) and it will show you an exact output of what is draining your CPU.
Here is an output from my 2651XM with 1Mbps of large packet 1472 size ICMP. You can see it's already 39% with "IP Input" at the top of the list.
Router#sh process cpu extended
################################################################################
Global Statistics
-----------------
5 sec CPU util 39%/14% Timestamp 4d01h
Queue Statistics
----------------
Exec Count Total CPU Response Time Queue Length
(avg/max) (avg/max)
Critical 1 0 0/0 1/1
High 8 0 0/0 1/1
Normal 2046 1458 0/40 1/7
Low 13 0 0/0 1/2
Common Process Information
-------------------------------
PID Name Prio Style
-------------------------------
65 IP Input M New
CPU Intensive processes
-------------------------------------------------------------------------------
PID Total Exec Quant Burst Burst size Schedcall Schedcall
CPUms Count avg/max Count avg/max(ms) Count Per avg/max
-------------------------------------------------------------------------------
To sum it up, I think you are hitting peak capacity of the router. Keep in mind there is overhead in the packets that you are not using in your calculations to determine 80-90Mbps of traffic!
