Running a 12v PCIE card from the motherboard

I am making a custom enclosure for the Framework and I’m looking to integrate a PCIE card that needs 12v input on the riser. The card only needs about 10w

At the moment the best solution I have is to step up the 5v3a I can get from one of the USBC ports to 12v. Not a terrible solution, but I am just wondering if there are any other ways.

The only other one that occurs to me is to ‘tap’ the battery input and step this down to 12v… however I am not sure if this would interfere with the charging logic if the 12v load caused any voltage drop and/or what might happen as the motherboard switches from drawing power to charging the battery.

Anyone have any ideas on this front?

That’ would only work for very few pcie devices, most may not draw more than 1A continuously but will go over it occasionally.

I have accidentally ran some 10gbit nics with 5v instead of 12 and it worked though XD

That is definitely a better source of power, though why not just tap into the main power bus while you are already soldering stuff. It’s the place where power is supposed to be grabbed from and won’t interfere with charging.

I can test the device, but it is listed as “10w” for its power consumption by the manufacturer. The USBC port should be good for 13a or so after conversion loss, I figured that was enough overhead but I guess testing wouldn’t hurt.

Where might I find the main power bus on the motherboard?

10W on average, could work or could not.

Ok, noted. I’ll run some power draw tests on it.

About that ‘main power bus’ though… IT had occurred to me, but no idea how to locate it on the board.

Should be somewhat easy to locate near the cpu vrms, but having schematics would certainly be neat XD

Hmm, not something I really want to screw around with without knowing more tbh. At least at the battery I know exactly where everything is. Would be fairly trivial to tap into there.

Why not just pull components from an eGPU enclosure? Or the aliexpress eGPU route? Surely it would be easier to integrate that way and I’m sure some madlad has gunned up a way to covert ATX PSU to USB-PD somewhere?

Egpu enclosure has a power supply, I would like to be able to run this pcie card without an external power supply and with the motherboard running off of the battery.

At the moment i am running the card from the internal nvme slot with a riser. The system ssd will either be over USB or with an adapter from the WiFi slot as i don’t need WiFi.

1 Like

But if you run the ssd over usb, you probably won’t have the full 3A available for the pcie adabter XD.

Might want to hook into the main power bus anyway. Just find the input caps for the cpu vrms and hook into that and slap a buck converter on that.

I hadn’t thought of that… and that is a fair point. Although this would be resolved by using the wifi slot for the SSD drive.

Honestly I want to avoid soldering tings directly to the mainboard. Even if I find what I think is the right part on the motherboard, and manage to check this with someone who can confirm my educated guesses… I just don’t like the idea of it.

One thing I for some reason had missed is that the BMS for the battery is on the battery itself. So voltage drop before the BMS actually shouldn’t be a problem (I don’t think). Splicing in a power outlet to the battery wires appeals to me much more than doing so on the main board caps. Plus if I screw it up I only lose a battery… not a mainboard.

The question remains what happens when the motherboard switches from power out to power in and if there is any power loss for any period of time when this happens. I would assume not given the laptop itself has to stay powered… but I guess only testing will tell.

How would you tap into the battery? Soldering “directly” to the battery is a lot more sketchy than the board imo.

It should not infulence battery protection or culoumb counting but it may infulence charging.

And that would not be a problem if you tapped the power bus where that is already dealt with.

Probably best to check the usb power method first though.

Or option 3 tap into the power input, get one of those c to c breakout boards and grab the power off that. Of course you would not have power while on battery but I am not sure if you usecase needs that (I am guessing it’s like a 10gbit nic or an hba or something which would make your setup relatively stationary).

Or option 4, get one of those usb “UPS” modules, that would smooth over the spikier bits if your card mostly works with the 5v boost option but not quite.

From the wires, if it was the right route then I would make a small PCB, snip the positive and negative wires and solder them onto this.

Yea, you’re right. I guess the laptop would get the voltage of the battery wrong and I even if the BMS has a high voltage cut off (which it may not) if the laptop was continuously thinking it was below the right voltage this would be problematic. I could set a lower charge limit… but it’s definitely sketchy.

Yea, I totally get what you are saying, I’ll be looking into it I’m just not a fan of the idea

It’s definitely the ‘cleanest’ option. It’s also the first one for me to test as I have the parts to do so.

When I first designed the enclosure it didn’t have a battery, and tapping input power would be WAY simpler! But I really liked the idea of having the battery for a variety of reasons, so unless it’s a complete road block I can’t tap input power as I do need it to work when only on battery.

I don’t think I would fit the UPS module in the case, everything is already quite tight. But a good idea if the card ends up having spikes that I can’t otherwise deal with.

The PCIE card is a SDI video in out - A Blackmagic Decklink Duo 2.

A bit of a curiosity:

When you mentioned the SSD power consumption, I looked up the drive I am using at the moment (a WD SN530 2230) and dismissed the figure from the manufacturer - 75mW - as it seemed impossible to me… so I figured they must be idle or power saving mode figures.

However their documents on their website say the average active power consumption is 75mW… even if we assumed this meant general file access, as opposed to sustain load. How can it possibly be 0.07w when most SSDs in this class are 3+w of power?

I can’t find any actual reviews or tests. It just doesn’t seem to add up.

Smartctl can show you the power state table for your drive, the 75mw are probably the deep sleep values.

Here’s an example (some samsung oem drive), every ssd is a bit different though and what power states it actually uses also depends on your usb enclosure.

The usb3 nvme enclosures I have usually don’t exceed 1A of draw on 5V.

You would think right… But here are the specs from the Manufacturer:

I’ll have a look at actual values when I get a sec, its just these docs that seem hard to believe

That’s just a bit of a garbage datasheet.

Whilst true, it just seems bizarre that WD would mis-report power consumption by a factor of almost 50.

It’s not misreported, they clearly state what they are testing XD

I just find it hard to believe that it only consumes 75mW whilst being used in any capacity… even web browsing/office work. As you said in your first reply to the question: 75mW sounds like deep sleep figures.