12 Volts Only Power Supply impact on power usage

[FucaBala]

Hello folks,

This is my first post here so apologies if I did on wrong place.

I watched the video from Wendell about possiblities for home lab and I believe there is one point that might be worth going a bit deeper.

Some time ago I got a HP Prodesk 400 G6, with a i5 9500, 64GB RAM, 1TB NVME and 3 SSDs. Iddle power usage was between 8 and 9 watts (no undervolt was performed since it is running SOPHOS firewall).

Now I included a quad ethernet NIC intel i350-T4 and is iddling around 12 watts.

When checking how much power a similar machine like this would draw from the wall on the internet, I found figures similar to what Wendell mentioned (between 40W and 50W).

There is nothing on the HP machine that I believe would make it so efficient, apart from their power supply which only provides 12V. The connectors from SSD are provided by the motherboard.

Does anyone have a regular i5, built with standard parts so we can have a comparison of how much these machines are drawing from the wall for comparison?

Thanks and wish everyone a awesome day.

Bill

[FunnyPossum]

I can’t see how a 12v only PSU wouldn’t help, its got to, right?

Out if interest, how did you measure power?

[rcxb]

@FunnyPossum

I can’t see how a 12v only PSU wouldn’t help, its got to, right?

12VO doesn’t improve efficiency much, it’s mostly a gimmick that just moves the inefficiency from the PSU and onto the motherboard. All PCs need 5V both when powered on and when in standby, so that (inefficient) conversion is always happening, whether it’s done in the PSU or on chips on the mobo.

12VO PSUs make sense for OEMs because everybody started looking-up the efficiency numbers of their PSUs, while nobody knows the efficiency of their mobos, so the inefficient voltage conversion can be hidden on there.

Of course 12VO offers some efficiency benefit in dropping all the other voltages that aren’t needed at all. But there are alternatives to ATX12VO, like:

…Seasonic’s CONNECT module. This does effectively the same as the ATX12VO standard, removing the 5 V and 3.3 V rails from the PSU and moving them to an external module, off of the mainboard. It can be fitted into the area behind the mainboard in many computer cases, making for very clean cable management. It also allows for increased efficiency.
Intel’s ATX12VO Standard: A Study In Increasing Computer Power Supply Efficiency | Hackaday

[Ruklaw]

I’m not sure it’s just a gimmick.

While certainly you could use this approach to move inefficient components from a PSU onto a motherboard in order to print good numbers on your PSU while wasting power on the motherboard I’m not sure that’s necessarily what happens.

I’ve had a Lenovo TS140 server since new (2015), and this uses a (mostly) 12v power supply standard - all the sata power leads come from the motherboard. It has a xeon E3-1226V3 quad core processor - equivalent to a fourth gen i5, TDP 84w - and is running 3x 2.5" 5tb hard drives, 1x 3.5" 5tb hard drive and a 1TB sata SSD for boot.

Total system power consumption at the wall, when idle as it is most of the time, is about 28w. Booting up it uses about 54w, peak I’ve managed to get is 94w when scrubbing the disks and running a handbrake encode.

I happen to have another Intel desktop of similar vintage and broadly similar spec, running an i7 4790 (not K, also 84w TDP), with a bit more RAM (16gb instead of 8gb) but just the one sata ssd drive, no graphics card or anything else installed, motherboard is an MSI Z87-G41 PC Mate.

With that idling on the desktop I’ve not seen it below 50w, boot is around 65w, peak about 100w.

Certainly the power supply on the second computer isn’t anything special (it’s a modular unit, perhaps a coolermaster?) but wasn’t scraping the bottom of the barrel either.

The power supply in the Lenovo is an FSP with an 80 plus platinum rating, it’s mostly 12v but does also have a 5v and -12v rail

Both are rated over 400w but are obviously only using a fraction of that.

It’s notable that between the two, peak power consumption is pretty similar (or at least, what you’d expect given the addition of hyperthreading) but the idle consumption on the Lenovo is much better.

Clearly, the platinum power supply has not just had inefficient components moved onto the motherboard, as the total system power consumption is markedly less that the other, otherwise similar system.

To add a couple more data points, my main desktop, an i5 10500 with an RTX 2070 super, 80 plus bronze PSU doesn’t get below 54w idling (and that’s with both screens off), both screens on browsing the internet it draws around 90w (single screen usage is in the middle around 75w), with peaks near 300w when gaming.

Polar opposite to that I have an M1 mac mini, which can run dual screens, video on one browsing on the other for a mere 8w (peaks about 25w when encoding etc). Certainly it helps that it’s basically running a souped up phone platform with minimal compatibility requirements!

I should be clear that I’m in the UK on 240v which helps efficiency a bit if comparing these numbers.

Overall, my impression is that a typical ATX desktop PC is not a very power efficient platform, and it’s likely that a 12V DC power supply standard could improve things, both from simplifying efficient power supply design and from encouraging components to standardise on voltage used.

Problem right now is it’s hard to assess how much there is to be saved until you spend the money on a different power supply, and the price of more efficient power supplies is such that you could be waiting (many) years to profit from switching to a more efficient model, even with the recent surge in energy prices.

This is compounded by the fact that most power supplies are ridiculously over-specced unless you are running an absolutely top end system - if you aren’t gaming (or are doing so at entry level) a cheap and efficient 200w power supply could be an absolute no-brainer purchase, but such things don’t exist.