[RC5] Win95/98 users take note:
nfo at deskmail.com
nfo at deskmail.com
Fri Jan 29 13:21:03 EST 1999
From: "Somebody Else" <elsesomebody at hotmail.com>
To: rc5 at lists.distributed.net
Subject: Re: [RC5] Win95/98 users take note:
Date sent: Fri, 29 Jan 1999 12:23:35 EST
Send reply to: rc5 at lists.distributed.net
> Someone please correct me if I'm wrong, but it seems to me that
> increasing the priority of the client relative to other applications
> won't increase the number of total keys processed. This is how I see
> it: Say you want to use your computer while rc5 is cracking in the
> background. Lets consider the example of launching netscape. There is
> a certain number of cycles/processes that need to take place to launch
> netscape. This will reduce the number of available cycles/processes
> available for rc5 to use.
see explanation below to see how wstart helps correct the reduction in cycles
to rc5des client. You are correct, however, that the max number of cpu cycles
available for all programs is constant and no program can make any other program
run faster than it would with 100% of the cpu cycles going to it.
However if decrease the priority of netscape
> it will take more time to launch but the same number of
> cycles/processes need to take place, it will just take longer to do
this is an excellent explanation of why if you give one program more priority
with wstart, the others will lag.
So in a 24 hour period lets say you have a certain number of
> cycles (24 hours * 3600 seconds per hour * your processor speed *
> 1000000 cycles per second). Netscape still needs the same number of
> cycles leaving the remaining cycles for rc5 and therefore yielding
> exactly the same number of blocks. Like I said if I need correcting
> please do so. And if I'm right I'd like to know for sure as well.
you are correct about the number of cycles my cpu produces being constant.
what wstart does is to devote more of those cycles to a specific program.
In a perfect OS, (there are none and windows is prolly the farthest
away...) all the cpu cycles would be divided equally among the
running threads. Even under this perfect system, as you opened
more programs, the number of cycles devoted to each thread would
be less and less. This would cause each consecutive program to
run slower as well as all the ones already open. This program
merely tells windows to pay more attention (ie give more cycles) to
a specific program. The program getting the most cycles will run
the fastest. Windows is a self serving OS, it will give itself cycles
before any other program. This program allows you to tell windows
to give a specific program the same amount of attention as it gives
itself. On my computer, this caused my dnet client to maintain a
more consistant output than it did when windows was allowed to
take cycles from it whenever deemed necessary. It is true that it
cannot make the program run faster than it does with the system
idle. I mentioned in my stats that my system runs at ~92kk/s
when idle. Lets consider that max output. I know that no program
can help me exceed that figure. In the daytime, (many other
programs running and sharing cycles) my output would fall to
~~70kk/s. This output drop was caused by windows giving some of
the constant number of cycles to other programs. Wstart basically
does this: It tells windows that its ok to share cycles with other
programs, but, you need to give more of them to this program.
Using Wstart, I was able to stabilize and increase my output
slightly to ~82kk/s when other programs were running. Note: it
does not exceed the max output value, it merely increased my
output in a highly tasked environment by giving it more cycles to
process with than windows normally would. Also, as I mentioned in
my original letter, giving more cycles to one program does cause
the others to lag. This is why, before using this program, you must
decide which is more important: more blocks per day or a system
that tasks smoother. I leave the choice to the individual. I am
merely suggesting a way to increase the output for those who run
the client on windows 95/98 systems that are also used for other
tasks. I did also include the disclaimer that these were my results
and that your may vary.
> >Do you miss the priority function that was available last in ~ version
> >417? A program called "wstart" available from
> >http://www.winfiles.com can give it back. It's a program launcher
> >app that lets you use command line switches to set the priority of
> >Win95/98 programs just like you can with the task manager in NT.
> >On busy systems like mine, it can give you a substantial increase
> >in keyrate while you're doing other things.
> >The speed increase is not without tradeoffs. The latency will be
> >slightly increased on your other apps. If keyrate is more important
> >to you than minor system slowdowns then you might want to
> >consider this program.
> >Personal results on my 486/dx4-100:
> >night rate (system idle) ~92 kkeys/s
> >day rate (without wstart) ~70 kkeys/s
> >day rate (with wstart set to "high") ~82 kkeys/s
> >Thats roughly a 15% increase in keyrate while system is loaded.
> >Your results may vary.
> >Something to think about :)
> Get Your Private, Free Email at http://www.hotmail.com
> To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
> rc5-digest subscribers replace rc5 with rc5-digest
Put your idle CPU cycles to good use:
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest
More information about the rc5