Regarding collaborative simulations on my server

srynznfyra

Distinguished
Dec 18, 2008
178
0
18,690
I have a server that runs pretty much idle all day (although there are alot of resources on it being used such as media streaming, printing, so on and so forth). What I mean by idle is that the CPU isn't really being utilized - it's an AMD Athlon at 2.2 GHz so I should imagine there is quite a bit of potential for scientific projects such as folding@home etc, especially as the server's running 24/7 not like my desktop computer(s).

There are two things that are stopping me from running these projects:

1) Would 24/7 100% CPU utilization mean that the server will use considerably more power? I've measured the whole computer to use about 50 watts and I'm wondering how much (roughly) this would change if the CPU was always at 100% utilization, as of course power = money + environmental costs of course.

2) Also, would 24/7 100% CPU utilization have a considerable negative impact on the lifetime of the components? The ones I'm especially worried about, in descending order are: PSU, CPU and other motherboard components such as northbridge, HT link etc etc etc.

I'm not worried about the PSU being over-utilized as it's rated at 450W and as previously mentioned the server currently runs about 50W of power, I'm more worried about capacitor aging due to high load for long periods of time, coil aging in the transformer, and other such things that can break/age.

I would have thought that CPUs were designed to withstand constant usage but I'm not totally sure how well my server's CPU would handle it - remember this isn't anything fancy and it would normally go in a desktop. Don't they use special CPUs in supercomputers/servers that have extra good quality silicone?

About the motherboard, it's a very (very) cheap gigabyte mobo, quite basic really. Not so worried about it, the main thing would be heat I reckon. There aren't any fancy heatsinks on the northbridge and other components (not even the RAM actually) and I'm wondering how necessary they are - they could be just for show judging from what PC gamers tend to buy! :lol:

Another (minor) issue would be heavy network load - although I'm pretty sure it would be quite easy to throttle network traffic for a certain CPU thread or threads (operating system is Debian Linux 5 and I'm open to any suggestions on this).

Thanks alot in advance :)
 

wuzy

Distinguished
Jun 1, 2009
900
0
19,010
Server components don't any special treatment besides higher QC/QA before being shipped/packaged for lower DOA rate and better (oftenly longer too) software support. Naturally that means desktop components will last just as long.

The whole thing about capacitor aging in PSU is just heat. Maintain an excellent case airflow and you'll be able to load a PSU 80-90% constantly 24/7 no problem (for PSUs with a well known ODM internals). Most 80 Plus certified PSU achieve their highest efficiency at ~65-70% load.

The idle-load power usage difference is relatively small compared to components that's has added voltage for overclocking. You can view data from sites like Anandtech for idle-load power difference (@wall) with modern quadcores.

As mentioned earlier, good overall case airflow does wonders for all the passively cooled components. If you look in pre-built rackmount servers you'll find they'll all have heavy front->back airflow and ducted air guide (I use cardboards) to divide up airflow areas on the motherboard.
 

srynznfyra

Distinguished
Dec 18, 2008
178
0
18,690
@wuzy:

Thanks alot :) I think I have quite good airflow as the fan in the PSU is REALLY powerful, and there isn't much air to move through (the case is micro ATX form factor). There aren't any fans apart from that but the PSU fan is so powerful that I don't think it needs any.

I'll try leaving it running with SETI@Home or something, and I'll measure the power usage increase. If it's below 10 watts I'll leave it on.

Thanks again for the advice, much appreciated.