[RC5] Re: Pentium4: Benchmark and real keyrate are different

Mark Renfer mri at netsec.ch
Thu Sep 27 16:42:05 EDT 2001


> Did you leave it idle for a while to complete a few packets?
> If so it could just be the system processes stealing a bit of
> CPU time.

Yes, I ran the client for hours without anyone using the machines,
the CPU usage of dnetc is most of the time at 99%. The impact of
the system services on an idle machine is less than 1%, not 20%,
even on much slower machines. I've been running RC5 on various
machines and operating systems for more than 3.5 years, this has
nothing to do with this huge 20% difference.

> When you run the client at raised priority, does this
> raise it to the same priority as it runs with in benchmark
> mode?

I think the benchmark runs at normal priority. I have raised the priority
to higher than normal level, but this has no impact. As said above,
the dnetc client gets almost all CPU time when run at the default
level, so increasing the priority is not likely to help in the first place,
but I still tried without success.

> It sounds like the RC5 application is not running the same core that the
> benchmark is.

It is, I tested all cores manually and the benchmark as well as the
standard running mode select the fastest core for the machines,
namely core #8.

I wonder if some person with access to P4 machines could run the
benchmark and then start the client and let it crack say 10 2^28
blocks and check the keyrate reported on the console.

> Do you actually use this machine, or did you just buy it as a heater?

Well, we do software development work with VC++. Our main product
is quite a large application and compilation times are an issue. The
P4s are brandnew and we don't have much experience yet, but they are
definitely much faster than our current P3 machines. We also have
a few Athlons, one of them being an 1.4GHz doing 5MKeys, but the heat
issue is really worse with the Athlons. The P4s with the Intel standard
coolers go slightly above 50°C. We tried several coolers for the Athlons
and we now use ThermalTake's SuperOrb which manages to keep the
1.4Ghz TBird at ~60°C. I know that this is still 35°C below its
specification
but those are really large dual-fan coolers.... So if heat is an issue,
the Athlon isn't exactly much better than the P4. I think both the
P4/1.7Ghz and the Athlon 1.4GHz have top power ratings of more
than 70W.

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