I dont think another agent handler would help, the handler 'handles' communication between the SQL server and the end points. If SQL is your bottleneck I dont see how this would help? Each agent handler is supposed to be able to handle 100,000 end points, but the best bet is 50,000. So well above what you need.
I have a separate SQL and ePO server (and agent handler in my DMZ) and that seems to give me the best performance.
Also might want to check your not keeping events you dont need and your purge jobs are running as expected. Deleting some unwanted events might improve performance, i.e. EEPC was set to report everytime the machine was locked and unlocked and turned on. Useful to keep a few days of this but 90 days was using way to much space in my database.
Pierce is absolutely right. With that size environment, you should be able to run on much lighter hardware.
Another thought would be to ensure that you have the proper VirusScan exclusions in place for both ePO and MS SQL. It's easy enough to overlook, and you could shoot yourself in the foot that way.
The only other suggestion that comes to mind, is the virtual disk performance. If you have access to vCenter (assuming VMWare), try looking to see what other VMs share the same datastore. Also, you could try increasing the resources shares of the ePO/SQL VM. We used to randomize our ODS tasks, but we have been moving more and more into the MOVE-AV world.
To follow up on Joe - usage of the McAfee Profiler (Available from your grant downloads) is absolutely critical for proper VSE exclusion policy - you won't know what your issues are until you uncover them. Profiler will give you a succinct list of processes and files which may actually be benign processes. Usage of low-process executables is preferred to on-access scan path exclusions.