Description of problem: "openssl speed" on an Athlon64 test server using the RH beta for AMD64 reports 922 1024-bit RSA signings per second. This is a tad less than optimized code on an 800MHz Itanium will produce on a CPU running at less than 1/3 the speed of the Athlon64. How reproducible: Fully reproducible. Actual results: 922 1014-bit RSA signatures per second. Expected results: At least 3000 1024-bit RSA signatures per second.
How does it look with the latest upstream openssl?
Current Fedora Core devel package uses the optimized assembly code for BN operations.