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

211–220 of 526 posts

Re: Nvidia's RTX 5090 power connectors are melting

#211

Post by diggan »

wing-_-nuts wrote: post #136
slightwinder wrote: post #92

Earlier quoted context omitted.

> So glad I went with 7900 XTX - should be all set for a couple of years. Really depends on the use case. For gaming, normal office, smaller AI/ML or video-work, yeah, it's fine. But if you want the RTX 5090 for the VRAM, then the 24GB of the 7900 XTX won't be enough.

Honestly, the smart play in that case is to buy 2 3090's and connect them with nvlink. Or...and hear me out, at this point you could probably just invest your workstation build budget and use the dividends to pay for runpod instances when you actually want to spin up and do things. I'm sure there are some use cases for 32gb of vram but most of the cutting edge models that people are using day to day on local hardware…

> most of the cutting edge models that people are using day to day on local hardware fit in 12 or even 8gb of vram.

I'm not sure what your idea of "day to day" use cases are, but models that fit in 12GB of VRAM tend to be good for like autocomplete and not much more. I can't even get those models to chose the right tool at the right time, even less be moderately useful. Qwen2.5-32B seems to be the lower boundary of a useful local model, it'll at least use tools correctly. But then for "novel" (for me) coding, basically anything below O1 is more counter-productive than productive.

Re: Nvidia's RTX 5090 power connectors are melting

#213

Post by numpad0 »

Schiendelman wrote: post #145
numpad0 wrote: post #66

Earlier quoted context omitted.

I'm suspecting it's really less than half a dozen people at NVIDIA, like guys in purchasing division or PCB designers not wanting to make a drastic parts/footprint change. M8 SMD lug terminal in a gamer accessory is crazy, but not rocket science.

That's what I wondered, you may not understand all the players. I believe the PCI standard specifies this Molex connector. Somewhere between what Nvidia ships and the power supply itself, that standard is the only common connection.

No, NVIDIA's use of the connector and first reports of melting predates the spec. They were never had hands tied to use it.

Gaming GPUs are having sagging problems for years too, and little is done to solve it. The cards are bending in their own weight. They're not products of proper engineering.

Re: Nvidia's RTX 5090 power connectors are melting

#214

Post by cesarb »

baq wrote: post #171
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.

At these wattages just give it its own mains plug.

> At these wattages just give it its own mains plug.

You might think you're joking, but there are gamer cases with space for two PSUs, and motherboards which can control a secondary PSU (turning both PSUs on and off together). When using a computer built like that, you have two main plugs, and the second PSU (thus the second mains plug) is usually dedicated to the graphics card(s).

Re: Nvidia's RTX 5090 power connectors are melting

#215

Post by eqvinox »

xxs wrote: post #189
dcan wrote: post #53

Earlier quoted context omitted.

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

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.

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.

Re: Nvidia's RTX 5090 power connectors are melting

#216

Post by FirmwareBurner »

ziddoap wrote: post #160
FirmwareBurner wrote: post #146

Earlier quoted context omitted.

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.

Presumably, the point is that a working car is more complicated & cheaper (in this case) than the graphics card, while the graphics card can't figure out how to make a connector. I read it as a kind of funny comment making a broader point (and a bit of a jab at nVidia), not a rigorous comparison. I think you might be taking it a bit more seriously than was intended.

An old legacy car is definitely not more complicated than designing and manufacturing a cutting edge silicon made for high performance compute.

The price difference is just the free market supply and demand at work.

People and businesses pay more for the latest Nvidia GPUs than for an old car because for their use case it's worth it, they can't get a better GPU from anywhere else because they're complex to design and manufacture en-masse and nobody else than Nvidia + TSMC can do it right now.

People pay less for an old beater car than for Nvidia GPUs, because it's not worth it, there's a lot better options out there in terms of cars and cars are interchangeable commodities easy and cheap to design and manufacture at scale at this point, but there's no better options easier to replace what Nvidia is selling.

Comparing a top GPU with old cars is like comparing apples to monkeys, it makes no sense that doesn't prove any point.

Re: Nvidia's RTX 5090 power connectors are melting

#217

Post by diggan »

engineer_22 wrote: post #123
Kirby64 wrote: post #112

Earlier quoted context omitted.

I’m aware we have 240V outlets. They are just not used in a place where you would put a PC. Until there is a shift in need (I.e., every normal user would need more than a 120V plug could handle), you won’t ever see 240V outlets in offices. I suspect it will never happen. In server areas and extremely specialized stuff? Yea, sure. But we’re talking desktop PCs here.

I envy my European friends' 240v electric kettles

> I envy my European friends' 240v electric kettles

... do you not have electric kettles in the US? Foolishly, I thought this was a standard kitchen appliance all over the world, I've even seen it in smaller cities in Peru.

Re: Nvidia's RTX 5090 power connectors are melting

#218

Post by dylan604 »

Joker_vD wrote: post #177
baq wrote: post #171

Earlier quoted context omitted.

At these wattages just give it its own mains plug.

Also put it in a separate case, and give it an OcuLink cable to attach to the main desktop tower. I suspect that's exactly where we're heading, to be fair.

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.

Re: Nvidia's RTX 5090 power connectors are melting

#219

Post by kookamamie »

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...

The GPUs require additional power, as the PCI-e slot they're connected to can only carry so much.

Obviosly there are GPUs without aux power connectors, but they're considered low-tier.

Post reply on HN