[hashcash] Obtaining speed results for different CPUs (collision tests/sec)
- From: John Honan <jhonan@xxxxxxxxxxxxxxx>
- To: hashcash@xxxxxxxxxxxxx
- Date: Fri, 20 Aug 2004 00:00:25 +0100
I have been attempting to compile a list of average collision
tests/second across different speed x86 CPUs. I have been trying to work
out some spammer breakeven type analysis and 'Moores law' CPU inflation
forecasts.
In order to get a like-for-like comparison, I've been using
hashcash.exe, just typing "hashcash -s" at the command prompt, and then
averaging over a number of results. The version of hashcash I've been
using is quite an old one, version 0.28 (Nov 03) obtained from
http://www.hashcash.org/binaries/win32/ I know I'm taking a very
simplistic approach to this as I'm using non-optimised code. I just need
some rough numbers for these forecasts.
Example results;
PII 400MHz = 221,300 collision tests per second
PIII 750MHz = 401,000 collision tests per second
P4 2GHz = 650,000 collision tests per second
P4 3GHz = 1,267,800 collision tests per second
486 25MHz = TBD
PII 266MHz = TBD
PII 233MHz (Mandrake) = TBD
My questions at this stage are;
1) I would like to run the timing test on a dos 486/25. Is there a
16-bit version of hashcash.exe available (or one that will work on a dos
486/25, running Win3.11 believe it or not...) The v0.28 exe was giving
me the 'cannot run in DOS mode' error, no matter how I tried to run it.
2) I have a PII 233MHz running Mandrake 8.0 (Linux 2.4.3_20). I'm not
much of a Linux expert, but I installed the rpms from
http://www.hashcash.org/binaries/rpms/hashcash-1.02-1.src.rpm However,
I think I might need a later version of Mandrake, as I was getting the
error "/lib/libc.so.6: version 'GLIBC_2.3' not found'. Would a later
version of Mandrake solve this?
3) Does anyone else have any collisions/second timings from current
versions of the code running on different chips (PPC for example) that
they could share with me? I need to get a 'fair' comparison across
CPUs. I'm at a bit of a disadvantage here, as most of the latest
compiled binaries appear to be for Linux... I'm doing my tests by simply
booting to dos and running the v0.28 hashcash.exe. Judging from Jonathan
Morton's speed results, things have moved on considerably since then.
I'd like to keep the comparisons fair. I don't really want to report
'1,267,800/sec' on the P4 3GHz, and then run the optimised code on the
P4 2GHz and report 4 million/sec (that might kind of skew my numbers!)
Any help would be much appreciated.
Thanks,
John.
Other related posts:
