Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: PSU requirements  (Read 2263 times)
Rhossydd
Sr. Member
****
Offline Offline

Posts: 1965


WWW
« on: July 08, 2012, 04:02:02 AM »
ReplyReply

How does one accurately estimate the PSU requirement for a new system ?

With the new Ivy Bridge CPUs power requirements being lowered from previous CPUs can we lower our PSU requirements ?
Greater integration of connectivity(more ports on the motherboard) means less extra cards are needed.
I know that SSDs confuse the situation somewhat, whilst needing lower power their speed of access means that the CPU can now work harder, so there's no real power saving there.
Manufacturers often don't specify a maximum current draw on their components to further complicate things.

From what I can find on the web, there's no utility that can actually measure the power being drawn from the PSU and inefficiencies in conversion mean that relying on actual current draw on the mains isn't a good way to see what the PSU is delivering. Although there are some useful estimation tools.

Being able to use a lower output PSU would be ideal, less power used, less heat, less cooling, quieter, better for the environment and cheaper to run and buy too. However I don't want the system to crash first time everything is being fully utilised.

Any thoughts ?
Logged
Farmer
Sr. Member
****
Offline Offline

Posts: 1631


WWW
« Reply #1 on: July 08, 2012, 05:52:26 AM »
ReplyReply

http://www.tomshardware.co.uk/power-supply-protection-calculate-consumption,review-32356.html
Logged

Rhossydd
Sr. Member
****
Offline Offline

Posts: 1965


WWW
« Reply #2 on: July 08, 2012, 06:34:47 AM »
ReplyReply

Interesting article, but the range of possibilities from the calculation chart gives me anything from 200w - 800w for my system. So frankly isn't a lot of use in choosing a PSU.

More useful is http://extreme.outervision.com/PSUEngine
« Last Edit: July 08, 2012, 06:40:50 AM by Rhossydd » Logged
PierreVandevenne
Sr. Member
****
Offline Offline

Posts: 510


WWW
« Reply #3 on: July 08, 2012, 08:06:30 AM »
ReplyReply

Efficiency varies depending on the load. Energy Star compliant power supplies should constantly be above 80% efficiency, regardless of the actual load. It could very well be that a good 800W power supply will consume less than a bad 400W power supply at a 200W load. The loss of efficiency at low load is not dramatic (see for example http://www.tomshardware.com/reviews/750-watt-psu-80-plus-gold,2927-7.html - a random link I picked). In typical scenarios you'll get at most a few percents of difference in efficiency. In order to really optimize your power consumption, you should then profile your typical use of the system and write an optimizer in a spreadsheet taking into account the relative times spent at different consumption level. It should be noted that graphic cards are the most power hungry components of a typical relatively high powered machine and that the trend is to use GPU acceleration everywhere now, including in web browsers.

OTOH, low quality, low output power supplies are a frequent source of problems. Most consumer desktop PCs in the $300-$600 range die from power supply failures (the second trouble spot is inadequate cooling due to undersized, underpowered fans and dusty heatsinks). Power supplies used at levels above 80% also have a very high tendency to fail or become too unstable for the MB voltage regulators...

I grew tired of blown power supplies and weird symptoms around 2005-2006 and since that time, I've continuously had a good dozen of PC running with good quality, high efficiency, power power supplies in the 650 - 1000w range. In fact, most of my power supplies have seen three video card/cpu/mb upgrades. IMHO, a long lasting power supply also has environmental benefits, if only in terms of raw materials.
Logged
Farmer
Sr. Member
****
Offline Offline

Posts: 1631


WWW
« Reply #4 on: July 08, 2012, 05:52:35 PM »
ReplyReply

I think the key point from the article, and one which is mirrored above by Pierre, is that it's more important to get a good quality, high-efficiency PSU than anything else.  That will save you power, heat and improve performance and lifetime of components (a good quality UPS also goes a long way in this regard, providing filtered power and ensuring better consistency as well as protection).

If the high end of your calculation suggests 800w, then I'd get a high-quality, high-efficiency, 800w supply (and, frankly, that sounds very reasonable for a modern machine whereas 200w-400w sounds very underbaked and possibly risks maxing out the PSU at periods of high load).
Logged

Rhossydd
Sr. Member
****
Offline Offline

Posts: 1965


WWW
« Reply #5 on: July 09, 2012, 12:59:27 AM »
ReplyReply

If the high end of your calculation suggests 800w, then I'd get a high-quality, high-efficiency, 800w supply (and, frankly, that sounds very reasonable for a modern machine whereas 200w-400w sounds very underbaked and possibly risks maxing out the PSU at periods of high load).
My point was that the range of values on that particular page is so wide it becomes useless.
My new system will have less components, and the CPU should draw less power too, so why would I need to increase the PSU capacity ??

I'm now waiting for a power meter to arrive and it will be interesting to see what my existing system is actually drawing. It should help in the choice of PSUs available ( all of which are decent, high efficiency ones anyway).
Logged
lfeagan
Full Member
***
Offline Offline

Posts: 208



« Reply #6 on: July 09, 2012, 06:01:56 AM »
ReplyReply

Power meters are a great diagnostic tool.

Regarding the "right-sizing" you are trying to do with PSU output capacity. Make sure to err on the side of too much rather than too little. Instability due to voltage sagging will cause you nearly endless headaches with weird symptoms coming from all over your system, both hardware and software. Estimating consumption is a fairly simple process. Just add up the numbers for your various chips (CPU, north/south bridge, memory/DIMMs, and other controllers on the motherboard), add-in cards (video, sound, firewire, etc), and devices (hard drives, CD/DVD, etc), and then tack on 25% or so for error margin. More if you think you are going to expand significantly in the future. For most single-socket desktops with a mid-range graphics card 600W should be sufficient. For a high-end dual-socket systems with high-end graphics cards, 850 to 1000 watts would be the way to go.
Logged

Lance

Nikon: D700, D800E, PC-E 24mm f/3.5D ED, PC-E 45mm f/2.8D ED, PC-E 85mm f/2.8D, 50mm f/1.4G, 14-24 f/2.8G ED, 24-70 f/2.8G ED, 70-200 f/2.8G ED VR II, 400mm f/2.8G ED VR
Fuji: X-Pro 1, 14mm f/2.8, 18mm f/2.0, 35mm f/1.4
Rhossydd
Sr. Member
****
Offline Offline

Posts: 1965


WWW
« Reply #7 on: July 09, 2012, 04:46:25 PM »
ReplyReply

Power meters are a great diagnostic tool.
Indeed, it's interesting to compare real numbers with the estimates of the web sites for my existing system;
http://www.tomshardware.co.uk/power-supply-protection-calculate-consumption,review-32356.html suggests anything from 200-800w
The Asus site mentioned in the above site suggested 650w
http://extreme.outervision.com/PSUEngine suggests 450w
Actual measured maximum draw = 333w

I'll install the meter in the new system and keep a close watch on what it's actually using. Initially at rest it's hardly getting off the bottom end of the scale @ 87w, even running Windows Experience testing it's only just exceeding 220w, but still has some HDDs to add.
In reality no task will pull maximum power from every component at the same time.

Logged
lfeagan
Full Member
***
Offline Offline

Posts: 208



« Reply #8 on: July 21, 2012, 09:46:21 AM »
ReplyReply

Just in case anyone wants some other numbers on power usage.

My system specs:
2x Xeon X5550 (2.66GHz, Quad-core)
24GB DDR3 1333 ECC (6x4GB DIMM)
2x Intel 160GB SSD
2x WD Caviar Black
1x PNY nVidia GTX 580

BIOS Notes: Hyper-threading is disabled. I have enough cores that I prefer to avoid HT's performance hit.

"Off" = 10 W
During POST/ECC bit setting = 225 W
Idle with C-States enabled in Linux = 230 W
Idle with C-States disabled in Linux/Mac OS X = 285 W
Prime95 running on all 8 cores = 400 W

Logged

Lance

Nikon: D700, D800E, PC-E 24mm f/3.5D ED, PC-E 45mm f/2.8D ED, PC-E 85mm f/2.8D, 50mm f/1.4G, 14-24 f/2.8G ED, 24-70 f/2.8G ED, 70-200 f/2.8G ED VR II, 400mm f/2.8G ED VR
Fuji: X-Pro 1, 14mm f/2.8, 18mm f/2.0, 35mm f/1.4
Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad