http://www.psc.edu/networking/projects/hpn-ssh/ has a number of patches to OpenSSH that increases performance in some situations. We should evaluate if it is a good idea to build our openssh with these patches.
They've also added support for multi core support for the AES CTR mode. Might need to poke upstream and the distributions to get them to update sshd as well though as the client might not be the bottle neck.
One should not though that modern SSH also like AES GCM, which is possible to run in parallel, but probably isn't implemented that way right now.
ssh isn't even close to being a performance bottleneck at the moment, so investigation bugs like these aren't productive.