Soldato
Shoot i've put it in the wrong forum. Could a mod move it please?
edit: Thanks
Last edited:
Linus takes a look at an ATX 12V0 PSU
- Thread starter james.miller
- Start date
Shoot i've put it in the wrong forum. Could a mod move it please?
edit: Thanks
I hope this catches on so that we end-users can buy them.
I can imagine a few possible reasons. It could be that 12V is too entrenched; the 12V0 standard only requires PSU and motherboard manufacturers to cooperate, but moving to 24V would also require graphics cards to be changed, as well as fans and any peripherals that use just 12V from Molex connectors. It's also possible that it would be more expensive overall; saving money on copper would be good, but a lot of cheap components would need to be changed for ones with higher voltage ratings.
Maybe just a 24v rail to the main board.
You'd probably be able to power the GPU via the pcie slot, then the main board can step down to 12v for fans and other stuff.
It's definitely time it got changed the 24pin connects is a bit of a joke. But so is the upwards of 12 pin GPU connections.
An integrated sata/power cable would be ideal, power drives via the main board.
So many things you can do to tidy things up.
I don't think a single move to a solitary 24v psu is really plausible. It would need new GPUs that run at 24v (and other add in cards of course, but GPUs are the obvious and most power hungry ones), unless you expect the motherboard to include the step down components necessary to power GPUs which in some cases you're look at north of 700w of 24>12v stepdown (to support the majority of dual GPU configs), which is more than your average ATX psu can output. That's bad for a number of reasons, cost being one of them.
A split 12/24v psu could work. But then how much capacity do they throw at 24v considering there's exactly no hardware output there that supports it right now? It's a bit of a chicken and egg situation.
If 12V0 gains traction, i can see revisions being made or new standards being born of the back of this - remember 12V0 is only aimed at system integrators right now - so maybe down the line 24v could be adopted as part of a new generation of ATX PSUs. But, even if that happens, it would be likely years before we see them on the market.
A split 12/24v psu could work. But then how much capacity do they throw at 24v considering there's exactly no hardware output there that supports it right now? It's a bit of a chicken and egg situation.
If 12V0 gains traction, i can see revisions being made or new standards being born of the back of this - remember 12V0 is only aimed at system integrators right now - so maybe down the line 24v could be adopted as part of a new generation of ATX PSUs. But, even if that happens, it would be likely years before we see them on the market.
It’s an interesting concept that’s for sure. How much of a cost saving are we looking at for one of these? Particularly against those PSUs that already do this internally.
The obvious downside is motherboard complexity increasing significantly. uITX obviously being a no go, I can’t see how they would fit it on an enthusiast grade mATX either should it become mainstream.
Are any SIs doing anything similar already? I know my Dell T20 server doesn’t follow the ATX power standard despite it otherwise being a standard mATX board.
What’s left that uses the 5v and 3.3v lines? Just disk drives?
The obvious downside is motherboard complexity increasing significantly. uITX obviously being a no go, I can’t see how they would fit it on an enthusiast grade mATX either should it become mainstream.
Are any SIs doing anything similar already? I know my Dell T20 server doesn’t follow the ATX power standard despite it otherwise being a standard mATX board.
What’s left that uses the 5v and 3.3v lines? Just disk drives?
Pretty certain that extra cost of added complexity in motherboard and possible extra DC-DC modules would be clearly costlier than currently in PC with multiple drives etc.
Drives (in fact SATA SSDs only use 5V), M.2, PCIe slots, USB.
Associate
Saw this the other day. It's one of those things that makes so much sense when you hear about it. Was quite surprised by power savings to be had too.
Quite likely yes
I suspect that the increase in cost from extra DC-DC converters on the MB will be less than the reduction in cost from removing them from the PSU. A lot of things on the MB already need them to generate lower voltages, notably the CPU, memory and chipset.
SATA, M.2 and PCIe need 5/3.3V, but the MB only needs to supply as much power as the number of connectors it has, whereas a PSU needs to supply enough for arbitrary devices that might be connected to its Molex and SATA power connectors. That means the conveters on the MB can be smaller and cheaper.
USB already needs DC-DC converters for USB-PD.
That's unrelated though, they won't be the same parts. Very different requirements.
That's the point though. In the past, ATX PSUs had most of their output capability on the low-voltage rails. Over time, more and more components switched to local regulation from the 12V rail, so the low-voltage rails could be made smaller. There is now very little on the MB that actually needs either 5 or 3.3V. As demand for lower power consumption drives voltages lower, the next logical step is to have those last few parts use local regulation too, rendering the intermediate conversion in the PSU redundant.
Lots of greenwash marketing and inaccurate/misleading methodology there.
If hardware feature set/amount of stuff doesn't change, there's very little to be saved.
First of all those DC-DC modules in PSU won't need much any idle power.
Just for scale idle draw of whole ATX PSU can be 5W!
(80+ Gold should be also below 10W)
So there's simply no way on earth, or hell, for moving lower rail DC-DC from PSU to motherboard to save that much.
And as major difference maker instead of high end board that tested board is basic features OEM board for market PCs.
Basic chipset Mini-ITX board would have been more honest reference point.
Or board from cheap market PC.
Pretty certain power draw would have "magically" dropped closer to 40W from already that.
Also that 32W consumption reading has somewhat questionable accuracy:
Cheap basic power meters require high power factor/very linear load to show always accurate power reading.
Otherwise they usually start showing too low reading.
And even active PFC PSUs have power factor dropping when going to 10% load level and lower.
So actual power draw could have been handfull of watts higher.
You still need controller and monitoring chip(s). (or then some more expensive for single component integrated thing)
And power output can be sized small only in that low end market PC board with little connectivity.
So for somewhat equal output it really won't be smaller, or much cheaper than inside PSU.
While motherboard's PCB design becomes more complex and you need connectors for drive power cables.
And more cables sticking out from motherboard for powering drives no doubt go well for those with glass side panel.
Trying to cram angled connectors to edge would just double PCB complexity increase.
And as you can see from that video even that full size ATX but minimal feature set market PC board was really full of stuff.
So for higher end PC you would fast need separate DC-DC dongle/block what ever hanging around somewhere inside case.
So we fast have more complexity and overall cost than in current situation.
Reason wasn't making lower voltage rails smaller.
Reason for increased use of 12V was huge rise in power consumption of CPU and graphics card from not needing any kind cooling to needing big active coolers.
Currents necessary for those high powers would have caused major transfer losses at low voltages, or needed lot thicker cables/bigger connectors.
Parts whose power consumption haven't increased have stayed on lower rails.