A different kind of benchmark of G4 against Pentiums
Gang, I don't mean for this to be a bash against Intel or the Pentium, but a post in one of the Mac newsgroups had an interesting set of numbers for a distributed key-cracking task (crypto is something I have a long-standing interest in, and I was probably the first to propose, back around 1993, that a "key-cracking screen saver" be distributed to soak up idle CPU cycles--this is now a "peer-to-peer computing" approach supported by the SETI project and even by Intel).
The numbers show a point both Dave Budde and I have made, that the apparently "slow" clock speed of the G4 doesn't tell the whole story. (Which surprises none of us, given issues of pipeline design/depth, types of instructions, etc.)
It will be interesting to see how the P4 does in this test. Also, the current 1.4 GHz Athlons should surpass the 500 MHz G4.
(Apple ships a dual-processor G4 system, but they are otherwise _way_ behind PC and server vendors in supporting multiprocessor configurations. Apple's focus is almost exclusively on personal computers, not on servers, of course.)
Here's part of the article (John Wasser, comp.sys.mac.system):
----
The Distributed.net client software for the RC5-64 code cracking
challenge takes advantage of Altivec. A 500 MHz G4 goes through keys about three times as fast as a 500 MHz G3, mostly due to Altivec.
(It also goes over 50% faster than a 1 GHz Pentium III)
n0cgi.distributed.net
rc5
Power PC 7400 (G4) 500 4,383,581 (Keys per Second)
Power PC 750 (G3) 500 1,660,172
Power PC 604e 200 617,764
Power PC 603e 200 623,348
Power PC 601 66 113,586
n0cgi.distributed.net
rc5
Intel Pentium III 1000 2,813,958
Intel Pentium III 500 1,385,637
Intel Pentium II 500 1,398,777
Intel Pentium (P5) 200 274,996 |
