Sounds like you got an even better deal out of HP!
That server, above, initially looks attractive, but after some back-of-envelope calculations, its power consumption 24/7 would be a killer for a professional scrooge like myself :o
There is a server project in Health & Efficiency, our Wayne's favourite environmental magazine. To help save the planet, they are designing a server at the very lowliest end of the spectrum. Based on the ultra-efficient socket s1g1 (AMD64) platform, the effort below runs fanless and with SSD.
Not yet undervolted or underclocked, and using an AMD Sempron 2100+ intended for netbooks, the system ticks over at 17 watts and maxes at 22 watts.
Though, according to great uncle Ebenezer Asbo, that still adds sixpence a day to our fuel bill. Or £20 a year! Not to be sniffed at: it's enough to buy our Wayne the 60 B&H he always gets for his birthday :D
cheers, a
(https://forum.kitz.co.uk/proxy.php?request=http%3A%2F%2Fwww5.picturepush.com%2Fphoto%2Fa%2F8664158%2F480%2Fd2703%2FD2703-A23-2b.jpg&hash=c90b9449083959611cac52b0f80dd09b93f3b22c)
The 20 watt webserver? (http://picturepush.com/public/8664158)
(https://forum.kitz.co.uk/proxy.php?request=http%3A%2F%2Fwww4.picturepush.com%2Fphoto%2Fa%2F8664182%2F480%2Fd2703%2FDSC-0308b.jpg&hash=d463e3000eeeda4c76d8998f6bdcbdab141f112b) (http://picturepush.com/public/8664182)
Installed in a Novatech Vision (http://www.novatech.co.uk/products/components/cases/cases/nov-vision.html) half-height case
(https://forum.kitz.co.uk/proxy.php?request=http%3A%2F%2Fwww5.picturepush.com%2Fphoto%2Fa%2F8664178%2F480%2Fd2703%2FDSC-0306b.jpg&hash=7e71cae76249904e34d7a6588d73fda96d0e25d8)
16.91 watts - Sell your shares in Southern Electric! (http://picturepush.com/public/8664178)
My main PC runs 24/7 anyway !! :o Mainly because most components fail at powerup so if I leave it running ..... I have absolutely no idea how much it costs !
It doesn't bear thinking about! ???
A large proportion of failures of modern consumer electronics are attributable to aluminium electrolytic capacitors, the lifetime of which is governed by a chemical process which is a function of temperature. As a rule of thumb, life of these caps is halved for every 10 degree increase in temp.
Equipment that is switched on usually creates a warmer environment, leading to shorter life expectancy of these components.
I always conduct postmortems when my own equipment fails, and it is quite rare these days to find failure that is attributable to any component other than these cap's. For that reason, even if not for energy saving, I like to power things off.
Browsing the documentation for Fujitsu industrial motherboards has uncovered some interesting 'thermography' tests. These tests (see example below) use an infrared thermal imaging camera to monitor a motherboard for signs of hotspots ("What is a Hot Spot not?!" "A good spot!"):
(https://forum.kitz.co.uk/proxy.php?request=http%3A%2F%2Fwww1.picturepush.com%2Fphoto%2Fa%2F8697074%2F640%2F8697074.png&hash=f5c3726519eac5a9cf6e8a8cf273684b5384b905) (http://picturepush.com/public/8697074)
Motherboard Thermography
Professional cameras for IR thermography cost a fortune ($10,000 and upwards (http://www.flir.com/thermography/americas/us/view/?id=44983)). Then it struck me that perhaps there are low-cost PC-based IR-sensitive cameras that could at a pinch be used instead?
The sensitivity of the camera, and specifically its true spectral response to infrared light, would presumably be the factors critical to success. Camera makers don't seem too keen to divulge this data.
Does that sound like it could work?
Also mentioned above is the undervolting of a CPU to reduce the core temperature and extend the MTBF of the system.
Most CPU have a voltage tolerance built into their specification. For example, the low-powered TK-42, an AMD Athlon64x2 core designed for netbooks, has a listed Vcore of 0.8000 volts.
Yet by very gingerly dropping the Vcore by just a few millivolts at a time (until the system crashes) it's possible to discover the point where it still runs stably but at a much lower voltage than officially listed for the CPU.
Due to manufacturing differences, each core is different, but I got one TK-42 to run at 0.6750v, about 15% less than the listed VCore for that CPU. In that case, undervolting the CPU also lowered the core temperature by roughly 15°C, even with the CPU at full utilisation.
On a website called elrepo.org (run by a frightening bunch of bearded *Nix fanatics) there are some excellent instructions on using Linux tools for (k8/k10) CPU undervolting :)
http://elrepo.org/tiki/kmod-powernow-k8
Next stage is to discover other appliances that can be undervolted!
cheers, a