[RC5] Pentium4: Benchmark and real keyrate are different

Mark Renfer mri at netsec.ch
Wed Sep 26 16:09:10 EDT 2001

Howdy folks,

Has anyone of you noticed that the RC5 keyrate reported
by the benchmark code on a P4 is actually about 20% higher
than the real speed that the client will be cracking at? I can
reproduce this on several P4 machines running W2K with both
the latest client as well as an older client. I have reinstalled
the OS on one machine which didn't change the behaviour.
Raising the client's priority doesn't help either. The P4 has
hardware features that allows the processor to reduce it's own
clock speed to prevent it from overheating if not properly cooled.
I have checked that this is not the problem by watching the
CPU temperature as well as the current clock speed using
the Intel utility. The processors are well below their
specifications regarding temperatur and the Intel utility
doesn't indicate that the processors reduce their speed. So
I am pretty sure that this is not an issue of this kind. I can run
the benchmark over and over again which drives the processors
to the same temperatur as they go when running the client, but
still the keyrate reported by the benchmark is 20% higher.

I get ~2MKeys on 1.7GHz machines while the benchmark
reports ~2.5MKeys. In the speed database of DNet, the P4/1.7
has 5 records, all at 2.4+ MKeys. I wonder if there is actually
something wrong with my machines or if the benchmark reports
a wrong keyrate. This could be the case even when taking
into account the entries in the speed database since the people
reporting the keyrates probably used the benchmark feature to
get the numbers.

Also, does anybody know of any work regarding an optimized
core for the P4 using SSE2?

Kind regards,       Mark

To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest

More information about the rc5 mailing list