University Computer Labs

[NookieN]

I've measured the power draw on my UPS with Prime running vs idle on my P4 3.3Ghz. When prime is running it uses an extra 130VA over idle power draw. That translates to an extra 90W. Keep in mind this is the change in total system power (i.e. due to CPU, memory, and chipset).

So an extra 50W is probably a good estimate for your typical ~2.5Ghz system.

[robreid]

Don'y tell my flatmates this,
I live in a shared flat and I run four computers 24-7 in prime95 but we still split the bill evenly
But hey, they are all girls, they use the oven, washing machine, hair drier more than me.

[TauCeti]

On a 3GHz P 4 it's about 70W more. And i mean W and not VA here.

On my system at work, prime95 does also mean that the northbridge-fan, the PSU-fans and the CPU fan go to full power which may decrease their life expectancy also.

And perhaps even the capacitors for the onboard voltage regulators wear out faster?

But maybe you can tell the admin that he should install some devices that _really_ power off the systems in the 16 hours they are not used at all? A typical PSU draws about 5-7W standby power that you really don't need to waste if you don't need Wake on LAN. And 100 PCs _really_ switched off at night might get you better rates with the fire insurance - just call them :)

Don't get me wrong here, _i_ think the costs are justified. It's just fair to tell the boss. I will do that also.... errrr... sometimes ;)

[lycorn]

Originally Posted by TauCeti

prime95 does also mean that the northbridge-fan, the PSU-fans and the CPU fan go to full power which may decrease their life expectancy also.

And perhaps even the capacitors for the onboard voltage regulators wear out faster?

Computer systems tend to get outdated very quickly now a days. I think that chances are that a PC will become obsolete well before it becomes worn out by extensive use...

[trif]

Originally Posted by lycorn

Computer systems tend to get outdated very quickly now a days. I think that chances are that a PC will become obsolete well before it becomes worn out by extensive use...

Not fans, especially in laptops. They're not built for 100% duty cycle. Fans in full size computers can take it, but variable speed fans are out of my experince, so I can't comment there. But it's very common to lose laptop fans to DC projects, especially Prime95. I recommend against using laptops for DC because of this. It would be very cool to have a setting that would run the DC program on a duty cycle just below where the fan comes on.

It's also my experience that DC projects extend the life of non-moving parts such as CPU, memory and motherboard systems by preventing thermal stress (stop-and-go syndrome). This is not a big deal if you are replacing your system every year, but many businesses and educational institutions run their machines for at least three years before replacement, and it becomes an issue then.

[Xyzzy]

Originally Posted by trif

It would be very cool to have a setting that would run the DC program on a duty cycle just below where the fan comes on.

http://www.mersenneforum.org/viewtopic.php?t=397

[markhl]

Some admins ask users to leave their systems on overnight and weekends (for virus scans etc.), making the PCs even better for DC projects!

If power is a concern, admins should ask users to switch off monitors.

If I go away for a while, I also switch off my speakers & printer.
I could even unplug my USB devices and Palm docking cradle!

These measures will offset the power requirements of a DC project, but I don't have hard figures on how much power these ideas save.

[trif]

Originally Posted by Xyzzy

http://www.mersenneforum.org/viewtopic.php?t=397

That doesn't keep the fan from coming on. If you figure out how much delay to put in to keep the fan from coming on when the box is idle, then when a user starts doing something, the fan will come on, even though what the user is doing is not sufficient to have started the fan alone. The DC program should adjust and simply not run when the fan is on.

I also don't have any statistics or references for "DC prolongs the life of components". I've been in charge of or worked on large fleets of computers over the past couple of decades, and it's much more rare to see a computer that has been running DC just up and die, than one that has not been running DC. Typically businesses will have 3 year warantees on their boxen and then replace them at the end of that (frequently they're leased anyway), but when a box dies, there are more costs than simply replacing it. Hassle and interruption of work are important factors, and Murphy dictates that the box will die during the most inconvenient time. Prime95 can even give one a heads up on a box that is about to die, as you can monitor the logs for hardware errors (not all errors, some are reproducible and due to inherent limitations of the FFT implementation used), and swap the box out at night or over the weekend if it is that critical that you don't want to wait to see if it fails altogether.