How long has your system been running without a reset?
 

I was just curious what the longest period of time that anyones GIMPS computer out there has been running without a reset. Here at home, I don't think that I have ever exceeded 30 days due to the fact that I either install a new piece of software that requires a reset, or upgrade Windows with one of the numerous patches which also require a reset. I have also had one power failure. I can imagine that some of you out there have a system sitting off in a corner that you never touch that has been running MUCH longer.

I had a long (somewhat over 6 hour) power outage early this month.

At that point, several of my LINUX boxes had been running GIMPS 24/7 dedicated for 7-8 months - or since I first moved them over TO GIMPS from RC5/OGR.

My experience with GIMPS and distributed.net using LINUX dedicated cruncher boxes is that is the NORM - they keep running unless I get a hardware failure or a power outage too long for the UPSs to handle. It's one of the reasons I picked GIMPS as a change from RC5/OGR - the native client support for LINUX - and *one* of the reasons I geve up in disgust on United Devices - even though most of my boxes are NOT optimal for GIMPS work.

I have a P-II box in the corner of a lab at work that's been up for well over a year. The project for which it was purchased is long since completed, and at 300 MHz it's a bit of a wimp, so no one has any interest in touching it.

My PC at home needs to reboot at least once a week. It's mostly because I download anti-virus updates.

Too bad we can't run mprime on it...

Our forum server:

[code:1]%uptime
12:38PM up 145 days, 16:56, 2 users, load averages: 0.36, 0.52, 0.45[/code:1]

I think the longest continuously running client has been on one of my FreeBSD dual processor machines:

11:20AM up 146 days, 10:03, 1 user, load averages: 2.00, 2.00, 2.00

Currently, my NT4SP6a server has been up for 124 days; would have been longer but for the building repairing the electrical system. The same system was up for 229 days prior to that.

My Linux server was up over 170 days, but mprime wasn't running the whole time because mprime crashed for a few days when the primenet server went down. A power outage last Thursday destroyed my uptime.

Is a reset good for you?
 

I'd been running my P4 under Win2k for several weeks without any apparent problems and knocked off my first exponent. I then started on a 10,000,000 digit baby and had been crunching that for some days, generally getting an iteration time of 0.105 or 0.106 when I wondered if Linux might be faster. So I rebooted into Linux and miserably failed to get mprime installed (boy am I dumb when it comes to Linux! :rolleyes: ). But here's the interesting thing. When I rebooted into Win2k and ran Prime 95 again on the same exponent I was getting 0.097 to 0.098 iterations! And I had changed nothing in my setup.

Can a system reset actually improve times by freeing up some resources or other? or had I just frightened Windoze into trying harder in case I replaced it? :)

P.S. If anyone has the patience to explain to me how to get mprime working I'd appreciate it.

Well there could always be a process stealing cycles. The taskmanager is a good place to look. Add a column for cputime - cpu-usage should be on by default. Look for process having a large cpu-time after the machine has been up for a day or so.

Now to the linux stuff.

I will use the console to do all this.

How far have you got? First of all you need to make a directory where you want mprime to run - something like md /usr/local/mprime - then put the tar.gz archive there. (mv mprime.xxx.tar.gz /usr/local/mprime) - next we get to unpacking the archive, use tar xvfz mprime.xxx.tar.gz to do that - after that start it up with ./mprime -m - you have about the same menu as in windows, just that it is text-based. Set everything up and you are done.

Hth, Matthes

Re: Is a reset good for you?
 

[quote="kwstone"]Can a system reset actually improve times by freeing up some resources or other?[/quote]
My experience says this is quite possible. I run under Windows 2000 and I have seen this happen plenty of times. It is not always necessary even to go as far as a reboot: commonly just stopping Prime95 (Test > Stop) and then immediately restarting it has this effect. You might have to do this several times to achieve the best iteration time!

Using Task Manager I have established that there is no consistent drain on CPU time from other processes causing this effect, and I think there was an earlier discussion somewhere that tentatively led to a conclusion that the way system RAM (i.e. [i]not[/i] cache) was allocated led to this effect - if it was allocated more sequentially/contiguously or something like that the access was quicker. I am not personally sure if this is true (or even possible) but I have certainly demonstrated this phenomenon many times.

George: would you care to comment on this phenomenon?