The use of salts renders pre-generated rainbow tables useless, hence my comments about how unlikely they are to get cracked. Salting combines a random string of bits with the passwords to create an entirely different string of characters which is then hashed. Basically it's like scrambling them and then hashing them.
OPHcrack isn't really an accurate measure as to what would be required to do this as you are likely using a rainbow table and attacking the LM hash. Try bruteforcing an NTLM hash using a CPU-based utility and you'll quickly see what I'm talking about. NTLM CPU bruteforcing typically generates <20 million passwords/sec whereas my 5770 generates 1.6 billion passwords/sec. An 8 character small/digit password takes about 20 minutes to bruteforce. That's why I use a 21 character small/capital/digit/special windows password, >10000 years
Salted, SHA-1 Oracle and mySQL passwords cut that speed by 2/3 and an 8 character small/digit password takes about an hour to crack on my GPU. That's assuming you know the salt. I don't know of any CPU or GPU-based hashing tools that can bruteforce a salted password in a fully automated fashion if the salt isn't known. Without the salt, the crack program can only attack the "outer" MD5/SHA-1 hash which means no password it generates a hash for will match the overall hash being bruteforced. Most forums tack on a 5 character small/cap/digit/special random salt on every user which means if you don't know the salt you'll have to generate hashes for the full range of characters at a length of 5 digits longer than the password itself.
TLDR: If the passwords were salted they can't use a rainbow table and it'll take an impractically long time to crack each one individually.
