Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Nvidia's RTX 5090 power connectors are melting

theverge.com

Post reply on HN
Print view View original

271–280 of 526 posts

Re: Nvidia's RTX 5090 power connectors are melting

#271

Post by KeplerBoy »

fooker wrote: post #247
dylan604 wrote: post #218

Earlier quoted context omitted.

I've built video rigs that did just that. An external expansion chassis that you could put additional PCIe cards when the host only had 3 slots. The whole eGPU used to be a cute thing, but it might have been more foreshadowing than we realized.

Have you measured latency? In modern(last 4 years approximately) GPUs, physical wiring distance is starting to contribute substantially to latency.

latency due to wiring distances is far from being an issue in these scenarios. The signals travel at the speed of light. 186 miles per millisecond.

The problem you will encounter with pcie gen5 risers is signal integrity.

Re: Nvidia's RTX 5090 power connectors are melting

#272

Post by immibis »

Dylan16807 wrote: post #240
cesarb wrote: post #227

Earlier quoted context omitted.

> they'd have to convince the whole industry to agree on new PSU standards. We already have a new PSU standard, it's called ATX12VO and drops all lower voltages (5V, 3.3V), keeping only 12V. AFAIK, it's not seen wide adoption.

It's not great. Dropping 5V makes power routing more complicated and needs big conversion blocks outside the PSU. I would say it makes sense if you want to cut the PSU entirely, for racks of servers fed DC, but in that case it looks like 48V wins.

There are already huge conversion blocks outside the PSU. That's why they figured there's no need to keep an extra one inside the PSU and run more wiring everywhere.

Your CPU steps down 12 volts to 1 volt and a bit. So does your GPU. If you see the big bank of coils next to your CPU on your motherboard, maybe with a heatsink on top, probably on the opposite side from your RAM, that's the section where the voltage gets converted down.

Re: Nvidia's RTX 5090 power connectors are melting

#273

Post by kersplody »

12vhpwr has almost no safety margin. Any minor problem with it rapidly becomes major. 600W is scary, with reports of 800W spikes.

12V2x6 is particularly problematic because any imbalance, such as a bad connection of a single pin, will quickly push things over spec. For example, at 600W, 8.3A are carried on each pin in the connector. Molex Micro-Fit 3.0 connectors are typically rated to 8.5A -- That's almost no margin. If a single connection is bad, current per connector goes to 10A and we are over spec. And this if things are mated correctly. 8.5A-10A over a partially mated pin will rapidly heat up to the point of melting solder. Hell, the 16 gauge wire typically used is pushing it for 12V/8.5A/100W -- that's rated to 10A. Really would like to see more safety margin with 14 gauge wire.

In short, 12V2x6 has very little safety margin. Treat it with respect if you care for your hardware.

Re: Nvidia's RTX 5090 power connectors are melting

#274

Post by perching_aix »

nubinetwork wrote: post #212

Someone remind me again why GPUs need 600 watts? I never liked the concept of having to plug a power cable into a GPU, but these new connectors are just terrible...

Because it's a competitive industry, and further efficiency gains are either not available ("can't do" option) or were deemed strategically unsound to roll out for now ("won't do" option), possibly both. It's an active dimension of sprawl.

Re: Nvidia's RTX 5090 power connectors are melting

#275

Post by masklinn »

superjan wrote: post #266
DannyBee wrote: post #97

Earlier quoted context omitted.

This is because they are trying to parallel like 50amps (it's 12 volt IIRC) over a few conductors to get to 600watts. If it becomes unbalanced due to any number of reasons, none of those individual cables can come close to handling it - they will all generate enough heat to melt lots of things. Conservatively, they'd have to be 8awg each to be able to handle the full load without melting if they ended up taking the f…

When you work with normal AC power, it is considered unsafe practice to use parallel wires to share load in a circuit. Reason: one might get decoupled somehow, you don’t notice, and when fully loaded the heat causes a fire risk. This problem sounds similar. A single fat wire is the easiest, but I guess it’s not that simple.

> This problem sounds similar. A single fat wire is the easiest, but I guess it’s not that simple.

The problem is the 12V architecture, so the only way you can ramp power up is to increase amperage, and sending 50A over a single wire would probably require 8AWG. That's... really not reasonable for inside a PC case.

Then again, burning down your house is somewhat unreasonable too.

Re: Nvidia's RTX 5090 power connectors are melting

#276

Post by immibis »

dragontamer wrote: post #82
Havoc wrote: post #77

What’s wild is that this is a company clearly capable of designing highly complex things with numerous tradeoffs, challenges and unknowns. And then the fuckin cable is the issue. Repeatedly.

Both of my previous cars had door recalls. Ford Focus Mk3 and Prius Prime 2024. Yup. The door had weird failure cases that needed a recall. -------- Connectors and cables is damn near Masters level in knowledge and application. Its a rarely studied and often ignored piece of engineering. The more you learn about them, the crazier it gets. That being said, this news that the 4090 and 5090 are using but one shunt resis…

A card that draws 600 watts likely already has more than 6 phases of power conversion, so it could put a separate phase on each pin, and one more on the PCIe slot power, and then guarantee load balancing as well as being able to detect any single broken connection.

Re: Nvidia's RTX 5090 power connectors are melting

#277

Post by revnode »

Kirby64 wrote: post #243
RetpolineDrama wrote: post #231

Earlier quoted context omitted.

I pulled a fresh 20A (120V) circuit just for my 5090 build.

What power supply do you have that even has a 20A inlet? 20 amp breakers are common for outlets (especially in newer builds) but the outlets are still 15A outlets. And there is essentially no desktop power supply that exists that would exceed a 15A outlet currently.

https://www.amazon.com/dp/B09PJYMK77/zcF9kZXRhaWxfdGhlbWF0aW...

I'm sure there are power supplies for servers that go above 1600 watts too. If you really want to, you can ... but you really shouldn't.

Re: Nvidia's RTX 5090 power connectors are melting

#278

Post by KeplerBoy »

short_sells_poo wrote: post #254
KeplerBoy wrote: post #119

Earlier quoted context omitted.

If only we had connectors which could actually handle such currents. Maybe something along the lines of an XT90, but no Nvidia somehow wants to save a bit of space or weight on their huge brick of a card. I don't get it.

This is the top end halo product. What's wrong with pushing the envelope? Should we all play tetris because "what's wrong with block graphics?". I'm not defending the shitty design here, but I'm all for always pushing the boundaries.

Pushing the boundaries of a simple connector is not innovation, that's just reckless and a fire hazard.

Re: Nvidia's RTX 5090 power connectors are melting

#279

Post by masklinn »

cesarb wrote: post #227
opencl wrote: post #196

Earlier quoted context omitted.

Because nobody makes 24V power supplies for computers, they'd have to convince the whole industry to agree on new PSU standards.

> they'd have to convince the whole industry to agree on new PSU standards. We already have a new PSU standard, it's called ATX12VO and drops all lower voltages (5V, 3.3V), keeping only 12V. AFAIK, it's not seen wide adoption.

It's also of no use for the problem at hand, PCIe already uses 12V but that's way too low for the amount of power GPUs want.

Re: Nvidia's RTX 5090 power connectors are melting

#280

Post by perching_aix »

elorant wrote: post #105

All these incidents are from aftermarket cables.

The cable and the connector are standardized, and no evidence of noncompliance has been demonstrated yet. ¯\_(ツ)_/¯

Also note that the only first-party cable NVIDIA provides is an adapter for converting between four 8-pin PCIe power connectors and the 16-pin 12V2x6 power connector. They do not provide a 12V2x6 to 12V2x6 cable nor a 12VHPWR to 12V2x6 cable, you're left with your PSU manufacturer's or other third parties' cables in those scenarios (which would include the reporter's).

I will say that it's regrettable that even derbauer's analysis was basically "idk man looks good to me". He did use a cable from the manufacturer of his PSU however at least for his own testing, so if you'd consider that a first or second party product, there you go.

Post reply on HN