My point here is that you cannot accurately determine the required specification for your power supply simply by adding together all the maximum power ratings for all the separate components - it simply doesn't work like that.
Maybe not, but for the vast majority of people who don't have access to the required tools to measure the power draw it's a reasonably quick way of estimating.
If you measured the power in realtime on any component in a PC (CPU, hard drive, graphics card, etc etc) you would see spikes in the power draw and those spikes may be close to the quoted maximum. However, the average power draw would be considerably lower than this, easily 50% lower on average.
True, but the is good reason why firms tend to work with theoretical maximum's althrough admittedly with Seasonic they are a bit more exact... I'll go into this below.
The possible exception is the CPU which can be made to run at 100% load with special software (eg Prime and other benchmark software) but even then I dont believe you would see a straight line in the power graph. Certainly, in normal usage under a Microsoft OS running music software, I dont think you'll be able to get the CPU to pull the maximum power rating for anything other than the occasional spike.
If your maxing out your CPU on a project to the point the audio is breaking up then I've noticed in past testing that your running the the CPU harder than prime does. Fire up Dawbench and take it to failing point and you'll see that Core temp will run hotter than a full on Prime 95 session and you'll be pullng the same wattage from the wall. Obviously depends on the cpu but I tend to end up working the 50% - 80% range on a i7 series quite a lot towards a projects end, so that's a 130w/140w pull (140w as it's overclocked) stright off just for the cpu for extended time periods.
As a direct result of this philosophy, when you put all the components together in a real system you will typically get a maximum power consumption much lower than the theoretical maximum and this is the general findings of the article I posted.
Yeah, true and given power saving advances over the years throttling of componants is expected but you don't prepare for the average you need to spec for the maximums or the first time you spike it you'll receive a nice BSOD for your trouble.
Sure, it's getting older now and some components in a PC draw much more power now (especially Graphics cards) but the principle is still valid. You could also argue that some components draw less power now than then did 4 years ago (some CPUs for example which have become far more efficient).
Like for like based on what as tested there, and what we'd use now cpu pull has doubled in 4 years.
I'm not an expert in computer PSUs, but I am an electronics designer who has some experience in designing both low and high current switched mode PSUs. In all my designs I've always ensured some 'overhead' for the required power supply - in other words, I never push it to the max. The reason is that you always want to ensure that the components remain within their 'safe operating area'. The safe operating area will typically vary alot with temperature and a multitude of other parameters and you can never really model exactly how stressed they will become as a result. Hence, as a designer you cover yourself by over-engineering the solution and adding overheads. The amount of overhead is simply a factor of your confidence in the components, which is normally very closely related to the price/quality of the components. In essence, the more expensive the component the less overhead you need.
All of which is very true, and something a lot of people would never consider and this is where we get back to my point defered from up top...
A few factors here: (1) Seasonic design good power supplies (2) They use good components (3) They will already have factored in overhead as part of the desing stage in order to cover themselves (ie a 460W PSU will run at 500W no problems).
I know this thread/artical is in reference to Seasonic but as a side bar for anyone else reading this that isn't you or I, is that a hell of a lot of firms don't work like this. PSU bench sessions are choc full of PSU's that advertise a figure and fail to meet it out of the box. PSU manfactures seem to like bouncing around maxmimum figures in the same way hi fi makes rate speakers PMPO rather than RMS, and that's why you see crazy figures getting quoted for some computer hardware as far as wattage draw goes. It's simply the company (gfx card makers normally) building in a overhead because they know at least some of thier customers spent £30 on a 800w psu rather than the £100 you'd require to get a good one that acturly delivered the ampage that is required for purpose.
Back to the case in hand. Seasonic do meet their rated figures out of the box and in most cases I've seen tend to exceed them by 10%, but I'd still prefer not to have them run at their limits to ensure a longer life span of the caps inside.
Given this, my thought is that I'd have no problems running a 460W Seasonic in a system whose proven (ie measured) maximum instantaneous power consumption was 400 - 420W. My 2nd thought is that I believe that a system with a theoretical maximum power draw of, say, 800W will never exceed 460W real world tests.
That would fully depend upon how the system is being used. One usage spike would hang the system if that arguement was used to fit a lower capacity psu in the first place.
In summary, I'd have no problem using the Seasonic X-series 460W PSU in the majority of averagely-specced modern music PCs.
Just to get this back on the topic of Martins original artical
Quote Matin Walker SOS 12/10:
The PSUs are initially available in X400FL (400W) and X460FL (460W) versions, but these wattages are probably not sufficient for the musician whose PC is stuffed to the gills with DSP cards and RAID hard drives. However, they are excellent choices for the more typical i5/i7-based audio PC, as long as you can afford to pay the premium price: the 400W model typically retails at around £115, which is roughly double the price of a fan-cooled device of similar capacity.
Martins full quote suggested the same as you, as in the psu's in question would suit the average user, and it was only those pushing their machines with loads of bolted in kit that might have issues.
One piece of advice in this (far too) long post, if in doubt them buy a cheap current clamp from RS, Maplin, Jaycar etc etc and physically measure the current being drawn from your AC socket when your PC runs. Measure it when it's doing musical stuff and then compare it to when benchmarks are running. You can use the current to work out the real world average power consumption. Add 10 to 20% (depending on how much confidence you have in the target PSU) and this is the real world value that you need to use to buy the PSU.
I do this already.
Machine in front of me:
1 X SSD
3 X WD 2B
Nvidia 210 GFX
180w at idle on desktop
290w Cubase running at 85% ASIO load
325w Prime 95 running
No other add in cards currently present. A couple of UAD's, a more capable cuda solution through, or even just overclocking the chip as this one is currently stock and I'd expect that 400w Seasonic to drop off the usability list given those figures, and if the PCI slots were fully loaded up and you've added another 80w to those above figures above, then by the time you factor in the 10% - 20% overhead figure we've broken the available output from the upper model as well.
ScanProAudio & 3XS Audio Systems