[HARDWARE] Solaris on SPARC keyrate

R. Bruch rbruch at u.washington.edu
Fri Oct 15 15:12:31 EDT 1999


On Fri, 15 Oct 1999, Aaron Kramer wrote:

> Has anyone else noticed that the benchmark keyrate (./rc5des -benchmark) is
> way, way higher than the actual keyrate?

This gives me a chance to put together a couple of messages I've been
meaning to post the last couple of days.  First, thanks to everyone who
contributed advice about putting together an Intel, specifically Celeron,
rc5 cracking machine.  I've now got a Socket 370 366 Mhz Celeron box up
and going.

The other thread/flamewar I wanted to contribute to is the overclocking
debate.  I've now got a couple of data points to share.  The first is one
that won't be very relevant to today's technology:  I took a Pentium 90
(60 Mhz bus speed x 1.5) and OC'd it to 120 (60 x 2.)  The keyrate went
from 127 Kkeys/sec to 169 Kkeys/sec, almost exactly a 33% increase, which
is exactly the increase in processor speed.  That box has been running
perfectly for a couple of months now.

The more relevant point is my Celeron 366.  I have got that box running at
550 Mhz (100 Mhz bus speed x 5.5) rather than the stock 366 (66 Mhz bus x
5.5.)  I have a fat heat sink on the processor (Global Win FDP32) and it
runs very cool.  The keyrate has gone from 1.03 Mkeys/sec to 1.55
Mkeys/sec, an increase of 50%, again matching the increase in processor
speed.  I haven't had any problems with the box, however I will be honest
and say it's only been about a day since I cranked it up.  I really think
this was a good way to go.  The processor cost $45, even with the
fan/heatsink combo price of $20 added on, it's still cheaper than a 400 or
433 Celeron.  And from what I've been reading, the current Celerons max
out between 550 and 600, so it's unlikely that a 400 will go to 600.  You
could probably get it to around 550, but here the 366 comes out on top (I
think) because it gets the benefit of the 100Mhz bus speed, while the 400
or 433 will be at 75 or 83 or something.  Also, the PCI bus will be at the
correct speed (33 Mhz = 100/3) as opposed to something higher, and
potentially unstable/damaging to peripherals (either 37.5 = 75/2 or 42 =
83/2.)  I'm not claiming to have all the answers, this is just my
interpretation from reading a lot of stuff for the past couple of months.

And finally to the original question:  Does the benchmark program report
significantly different speeds than you actually get?  Well, on my Celeron
box, rc5des -benchmark gave a speed of 1,716,866 but even when it's the
only thing running on my box the client runs about 1,551,700 max, a
difference of about 11% of the lower speed.  I've never noticed that much
disparity on any other box, but the numbers were never this big before...

Anyway, any questions or disputes are welcome.

-Rick

--
To unsubscribe, send 'unsubscribe hardware' to majordomo at lists.distributed.net



More information about the Hardware mailing list