[mlsu]

Connectors are actually extremely difficult to make.

- you have to ensure that the metal connectors take shape and bond to the wire properly. This is done by crimping. Look up how much a good crimping tool costs for a rough approximation of how difficult it can be to get this right.

- one plastic bit has to mate with another plastic bit, mechanically. This needs to be easy enough for 99.99% of users to do easily, yet it needs to be 99.99% reliable, so that the two bits will not become separated, even partially. Even under thermal expansion.

- the electrical contacts inside must be mechanically mated over a large surface area so that current can pass from one connector to another.

- it must be intuitive for people to use. Ideally user pushes it and it clicks right in. No weird angles either, it could be behind a mechanical component that's tough to reach. Also, user has to be able to un-mate the connector from the same position. It should be tough for a user to accidentally plug in an ill suited connector into the wrong slot.

- has to cost peanuts. Nobody will pay $3 for a connector. Nobody will even want to pay $1 for a connector. BOM cost is 15-20% finished goods cost. Will the end user pay $8, $10, $12 for a good connector? No.

- repeatable to manufacture (on the board and on the cable) at high quality. User might take apart their PC a dozen times, to fix things, clean, etc for the lifetime of the component. So the quality bar is actually very high. Nothing can come loose or break off, not even internal parts.

- physically compact. PCB space is at an extreme premium.

- your connector design has to live across many product cycles, since people are going to be connecting old parts to new boards and they'll be upset if they can't do this. So this increases risk by a lot as redesigning a connector means breaking compatibility for existing users.

Connectors are actually a very very deep and interesting well.

I'm not surprised at all that they are running into issues here, these cards are pulling 500+ watts. That is a LOT of current.

I think next gen we will begin seeing 24V power supplies to deal with this.

[pja]

> I think next gen we will begin seeing 24V power supplies to deal with this.

May as well go the whole hog & jump to 48V.

(50V is as high as you can go whilst still being inside the “low voltage” electrical safety regime in most countries IIRC.)

[eqvinox]

General SELV limit is 60V, that's why PoE is 54≈56V at the source (it's calculated at roughly 10% tolerance so it can be built cheaply.)

[Schiendelman]

Then the graphics card would have to have a transformer on it to step down to the voltage that the chips can handle.

[dcan]

They already do - most of the components buck the 12V down to the 1.3ish volts that the GPU core needs

[xxs]

They are not transformers, though. The coil/chokes are not galvanically isolated which makes them (more) efficient. Stepping down from 48V to 0.8V (with massive transient spikes) is generally way harder than doing it from 12V. So they may ended with multi step converters but that would mean more PC with more passives.

[eqvinox]

3.3V from 48V is a standard application for PoE. (12V intermediate is more common though.) The duty cycle does get a bit extreme. But yes, most step-down controllers can't cover both an 0.8V output voltage and 48-60V input voltage. (TI Webench gives me one - and only one - suggested circuit, using an LM5185. With an atrocious efficency estimate.)

You'd probably use an intermediate 12V rail especially since that means you just reuse the existing 0.8V regulator designs.

[xxs]

Aside the step down, the transients can be quite crazy, which might make the power consumption higher (due to load line) calibration. 48V fets would have much worse RDSon compared to lower voltage spec'd ones. So it does make sense no single smart power stage to have such transistors (presently).

There are other issues, too. 48V would fry the GPU for sure, 12V often time does not even with a single power stage failure.

In the end we are talking about a stupid design (seriously 6 conductors in parallel, no balancing, no positive preload, lag connectors, no crimping, no solder) and the attempted fix is a much more sophisticated PCB design and passives.

[Schiendelman]

So then it would need to be significantly larger.

[lazide]

Likely smaller actually.

[SV_BubbleTime]

This isn’t how it works.

Your SMPS needs sub-2V output, cool. That means it only needs to accept small portions of the incoming.

But, if the incoming is 48V, it needs 48V tolerant parts. All your caps, inductor (optional typically), diodes, the SMPS itself.

Maybe there isn’t a sides difference in a 0603 50V capacitor and 10V 0603 capacitor, but there is a cost difference. And it definitely doesn’t get smaller just because.

Your traces at 48V likely need more space/separation or routing considerations that they would at 24V, but this should be a quickly resolved problem at your SMPS is likely right next to your connector.

[deelowe]

No one uses transformers anymore. VRRs are basically mini PCs now. They run firmware and report telemetry and are crazy efficient.

[Schiendelman]

I'm not familiar with this! I've tried to investigate but I just get variable refresh rate. Tell me more?

[deelowe]

Voltage regulators. Voltage regulation technology is extremely advanced as even very small efficiency gains can save billions for hyperscalers. Unfortunately, I don't know of any specific products to share as power isn't my domain. I'm only familiar with the space because we sometimes have to pull telemetry directly from the VRs when doing system level RCAs. Some of our BMCs can do this directly via I2C.

[Schiendelman]

No that's okay, thank you for the pointer in the right direction!

It doesn't look to me like anything out there can take voltage from 48v to 2-3v; at least not obviously.

[readingnews]

>> Connectors are actually extremely difficult to make.

While your points listed are valid, we have been making connectors that overcome these points for decades, in some cases approaching the century mark.

>> I'm not surprised at all that they are running into issues here, these cards are pulling 500+ watts. That is a LOT of current.

Nonsense. I used to work at an industrial power generation company. 500W is _nothing_. At 12VDC, that is 41.66A of current. A few, small, well made pins and wires can handle that. It should not be a big deal to overcome that. We have overcome that in cars (which undergo _extreme_ temperature and environmental changes, in mere minutes and hours, daily, for years), space stations (geez), appliances, and thousands of other industrial applications that you do not see (robots, cranes, elevators, equipment in fields and farmlands, equipment in mines, equipment misused by people)... and those systems fail less frequently than Nvidia connectors. But your comment would lead one to think that building a connector with twelve pins on it to handle a whopping (I am joking) 500W (not much, really, I have had connectors in equipment that needed to handle 1,000,000Watts of power, OUTDOORS, IN THE RAIN, and be taken apart and put back together DAILY) is an insurmountable task.

[userbinator]

One word: cost.

Look up how much industrial/automotive connectors cost, and you'll see the huge difference in quality.

[drdaeman]

Those GPUs aren’t particularly cheap, even a $100 connector and cable wouldn’t be a huge deal breaker for a $2000-3000 device if it means it’s reliable and won’t start a fire (that’ll cost way more than $3100)

[vollbrecht]

Yes cheap connectors exist and there is a marked for it, like everything "cheap". But to what point one wants to "defend" a trillion dollar company, on a product that was never marketed as "cheap", that actually comes with a hefty price tag, to skimp on something that is 0.01% of there BoM cost. If you sell for a premium price you should better make sure your product is premium.

[Manozco]

I've bought cars that cost me less than a nVidia card (and they were running).

[FirmwareBurner]

Which new cars cost less than 2000$-1000$?

[ziddoap]

They didn't say new cars.

[FirmwareBurner]

Then what's the point of such an arbitrary comparison? It's normal that plenty of commodities that were expensive when new have been devalued by age and can cost less on the used market than the top of the line BRAND NEW cutting edge GPU today, which itself will be worthless in 10-20 years on the used market and so on.

[numpad0]

used objects and imports from economically isolated land are traded at meme value, doesn't count.

[janalsncm]

That would be relevant if the margins on GPUs weren’t astronomical.

[broeng]

No, not for a connector for 500W, on a $2000 GPU from one of the worlds biggest companies. They can do better.

[Henchman21]

Well surely they can take that cost out of the $5090 people are paying for these cards.

[state_less]

They could use a common XT90 or something similar. You find high amperage connectors on all the RC lipo batteries and they are cheap enough, you find them on $100 products (batteries).

I regularly work with 100amp+ at 12v. It’s obvious the connector NVidia is using is atrocious and we all know it.

[jnwatson]

Nvidia is clearing 4 figures on each 5090. They can afford another few dollars on connectors.

[ImHereToVote]

"Nobody will pay $3 for a connector"

I would pay $10.

[Henchman21]

This whole conversation seems absurd! Of course you'd pay for the right power connector for your multi-thousand dollar card!

You don't buy a $200k sports car and then take it to Jiffy Lube for oil changes. You pony up for the cost of proper maintenance!

[Modified3019]

A quote I found the other day and saved (forgot where from):

>Like the classic trope says, it's not about affording to buy the Ferrari, it's whether you can afford to maintain it.