You need to make a pool for fastcgi, see the loadbalance example http://wiki.nginx.org/LoadBalanceExample
This solution is the same for any OS suffering from upstream timeouts.
With a pool of 4 you can easily handle 1000 requests/sec, and yes even with nginx win32.