Page 2 has an error: The 2500K should be listed as 32nm with a 95-watt TDP.

Interesting article

corrected.

What is the i7 system using all that 3.3V current for?

I look forward to the day when PCs require only +12V (or +24V, or any other voltage, as long as it's only one).

Thanks for the update!

Thank you for the analysis.
It's good to have something to point to when people want advice on a new system and choose a 600W PSU for a system that's mostly the same as the i5 6870 combination from the article.

Agreed. It's also showing that there's a definite lack of quiet, high-efficiency PSUs on the market for the 300-400W range. From a gamer perspective, it would be interesting to check the power consumption under gaming situation, to balance it against a full load situation, and help decide what PSU range once should get.

Unfortunately, every game has a different load power. One could lightly stress a system and another could crush it...see my signature for an example.

While I appreciate what you have tried to do with this article I think it is very misleading. The reason I think so is because while the i5-2500K + HD 6870 is less efficient in power usage it is highly more effiecient in terms of processing power. If you were to run some benchmarks for the computers such as 3dMark 11, SiSoftware Sandra 2010, Blender, etc. and normalize them a set score you would find that the new generation of processors (xxxx models from intel and fx series from AMD) will perform quicker while drawing less power so your energy usage is lower overall. Plus with new zero power states on the processors and now HD 7000 series gpus the idle power is much lower. I am not saying I disagree with the fact that most of the computers (even gaming) need the high power output (750+) psu that are available but that it needs to be tempered by the ability of the system to complete the tasks. Howerver I could be wrong but if you would be so kind to test my theory I would appreciate it.

I'm not sure it's misleading. Almost every article comparing gear can be misleading to those who jump to unwarranted conclusions (most people, that is).
This article performs a useful service in documenting quality measurements of power use in real conditions. If you use it to determine which CPU is more efficient or something, you're using it wrong.
Obviously idle power is hardly a function of the CPU. The mobo is a lot more important for one thing.

The article is clearly looking purely at the distribution of power consumption and NOT power efficiency.

This article confirms my suspicions that most power supplies sold these days are extreme overkill for most users. Your i5 system would be a typical mid-high range system in today's market and you show that it needs no more than about a 300W PS.

Why didn't you run the same test with the i5-2500K system using the onboard GPU rather than a discrete GPU? I suspect that in that case your conclusion would be that a 200W PS would be sufficient.

A modern desktop system with a SB (or soon Ivy Bridge) CPU using on board graphics with an SSD really doesn't need much power. And unless you are a gamer I don't know how this isn't a sufficient system for everyone, assuming that you have one PC acting as a media server for all of your large files.

Isn't this a function of the motherboard rather than the CPU?



+5V looks like a good candidate to drop first. What's the purpose of the negative voltages?

Might be, but extremely interesting nonetheless!
In multi-rail PSUs the 24-pin connector is often bunched together with some other potentially power hungry connectors, while the 8-pin connector get a rail of its own.
Given that the 24-pin is also used to feed most/all PCIe slots with up to 6A each, this connector really deserve its own 12V rail.

Yeah, nobody use USB these days...