| 1 |
Am Mon, 20 Mar 2017 18:16:55 +0000 |
| 2 |
schrieb Alan Mackenzie <acm@×××.de>: |
| 3 |
|
| 4 |
> Hello, Dale. |
| 5 |
> |
| 6 |
> Apologies to Thelma for hi-jacking the thread so early, but this point |
| 7 |
> is too interesting just to pass over: |
| 8 |
> |
| 9 |
> On Mon, Mar 20, 2017 at 12:57:00 -0500, Dale wrote: |
| 10 |
> |
| 11 |
> [ .... ] |
| 12 |
> |
| 13 |
> > I might add, when I buy power supplies for my puter rigs, I try to |
| 14 |
> > buy one that will only be loaded at around 40 to 50%. |
| 15 |
> |
| 16 |
> Do you mean 40 to 50% when the computer is going full blast (?like |
| 17 |
> building libreoffice, or something like that). |
| 18 |
> |
| 19 |
> > One, it is lightly loaded relative to what it can handle. Two, it |
| 20 |
> > will most likely handle heat better at those levels. Third, it |
| 21 |
> > allows for upgrades, hard drive additions etc without having to buy |
| 22 |
> > another one. |
| 23 |
> |
| 24 |
> Just how does one calculate the amount of power a box will use? |
| 25 |
> Processors proudly say "95W tdp", or whatever, but how much power does |
| 26 |
> RAM use, or the motherboard, or SDDs, or HDDs? |
| 27 |
|
| 28 |
I think that is pretty much well documented somewhere. HDDs use much |
| 29 |
more power during spinup, after that they should be somewhere around 10 |
| 30 |
watts - depending whether you access them or not. SSDs have no spinup |
| 31 |
but also vary in power draw if accessed or not accessed. But whatever |
| 32 |
you do, they should stay far below HDDs. |
| 33 |
|
| 34 |
Take care of the PSU efficiency. Most are at 80%. The rest of the watts |
| 35 |
is just converted to heat. |
| 36 |
|
| 37 |
Similar with CPUs, tho most of the watts are converted to heat. |
| 38 |
Actually, CPUs draw more power than the TDP value, especially during |
| 39 |
peaks (for example, when turbo boost activates). |
| 40 |
|
| 41 |
A system producing less heat automatically uses less watts. For you CPU |
| 42 |
I'd suggest to undervolt it (start with 0.1 to 0.2 volts), it should |
| 43 |
reduce heat by a good degree, and thus also power drawn. I undervolted |
| 44 |
my i5 2500k by 0.15 volts (offset mode), and instead raised the turbo |
| 45 |
multiplier from 37 to 40, plus added a better cooler. The result is: No |
| 46 |
more thermal throttling, temperature is 10-20°C lower, idle temp is |
| 47 |
much lower. I didn't measure power drawn but expect it to be some watts |
| 48 |
lower which I expect to improve my PSU live. |
| 49 |
|
| 50 |
> I'm anticipating building a new rig in the coming weeks/months |
| 51 |
> (depending on how soon the motherboard makers start producing Ryzen |
| 52 |
> MBs in quantity), and I'd be interested in getting an optimally sized |
| 53 |
> PSU. Most of the time, my PC is just idling along, with sporadic |
| 54 |
> bursts of activity like building libreoffice. |
| 55 |
|
| 56 |
I wouldn't call that a sporadic burst... Building LO is a heavy |
| 57 |
long-time job in terms of CPU cycles and power drawn. And it forces |
| 58 |
heavy duty on all components at once: HDD, RAM, CPU. No no, not |
| 59 |
sporadic... ;-) |
| 60 |
|
| 61 |
> So how do I work out the electricity consumption of all these |
| 62 |
> components? |
| 63 |
|
| 64 |
Look at the specs of each component (peak values), add some safety, |
| 65 |
plan for undervolting (CPU, maybe also RAM), and buy a PSU with |
| 66 |
efficiency above 80%. |
| 67 |
|
| 68 |
-- |
| 69 |
Regards, |
| 70 |
Kai |
| 71 |
|
| 72 |
Replies to list-only preferred. |
Archives