> Timo Schoeler <[EMAIL PROTECTED]> wrote: On an semi-related note, I recently tested the vpn1411 in a significantly faster (2.8GHz P4 Celeron D):
des3/3des: w/ acceleration: # time dd if=/dev/zero bs=1m count=100 | openssl des3 -pass pass:test -engine cryptodev -out /dev/null engine "cryptodev" set. 100+0 records in 100+0 records out 104857600 bytes transferred in 4.724 secs (22195998 bytes/sec) # sysctl kern.usercrypto=0 kern.usercrypto: 1 -> 0 w/o acceleration: # time dd if=/dev/zero bs=1m count=100 | openssl des3 -pass pass:test -out /dev/null 100+0 records in 100+0 records out 104857600 bytes transferred in 5.556 secs (18870270 bytes/sec) Note: A difference of about 4MBps -- it would most like be more-profound on a VIA C3 or AMD Geode platform. The little old CeleronD 2.8GHz can carry its own weight. aes-256-cbc: # time dd if=/dev/zero bs=1m count=1024 | openssl aes-256-cbc -pass pass:test -out /dev/null 1024+0 records in 1024+0 records out 1073741824 bytes transferred in 20.563 secs (52215443 bytes/sec) # sysctl kern.usercrypto=1 kern.usercrypto: 0 -> 1 # time dd if=/dev/zero bs=1m count=1024 | openssl aes-256-cbc -pass pass:test -out /dev/null 1024+0 records in 1024+0 records out 1073741824 bytes transferred in 45.238 secs (23735297 bytes/sec) A 2x difference: The big question is differential power supply amp draw and ambient temperature on the CPU, which I will try to measure later. Don't forget to set the sysctls! Don't forget "-engine cryptodev". Also note that "openssl speed" is not a valid method for bechmarking /dev/crypto apparently. ~BAS ---- hifn0 at pci8 dev 11 function 0 "Hifn 7955/7954" rev 0x00: LZS 3DES ARC4 MD5 SHA1 RNG AES PK, 32KB dram, apic 4 int 20 (irq 9)
