McKinley does much better on Open SSL marks

Details below:

"The numbers you posted for Itanium 800MHz are for an old version of OpenSSL. I just tried the benchmark myself (openssl-0.9.6g) and on a Debian-based 733MHz Itanium machine, I'm getting these results:

$ openssl speed

"Note: I simply downloaded openssl-0.9.6g ran "config" and "make". No special configuration or anything of that sort.

"The McKinley numbers are of course signficantly higher. Please see this URL for more info:

HP page"

And another reader writes:

"Off the record, RedHat/OpenSSL does not do either of the CPUs justice, RedHat does not seem to ever bother compiling the ASM code that comes with OpenSSL. The Itanium always looks bad unless some work is put in. These are some times for linux boxes

signs/sec verifies/sec
1024 bit 66.7 1197.5 ia64 800 OpenSSL
1024 bit 1845.5 24700.0 ia64 800 RSA

1024 bit 52.1 1092.6 Celeron 400 OpenSSL
1024 bit 117.7 1676.0 Celeron 400 RSA

1024 bit 57.0 994.2 P4 1700 OpenSSL
1024 bit 556.5 9034.0 P4 1700 RSA

1024 bit 263.0 5508.7 Athlon 1666 OpenSSL
1024 bit 656.0 9042.0 Athlon 1666 RSA

1024 bit 205.0 3980.0 Clawhammer 800 OpenSSL

"So I would say that the Clawhammer 800 should really be able to get around 510 and 6530 if the ratios for the Athlon stay the same.

"As you can see, RedHat/OpenSSL make the P4 and itanium look very bad, while in reality, they are very good.

"I believe that the Clawhammer in 64bit mode should be at least four times faster, so in theory, clock for clock, it should be faster than the IA-64 which may upset a few Intel feathers."

[That's enough benchmarks, Ed.] µ