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

521–526 of 526 posts

Re: Nvidia's RTX 5090 power connectors are melting

#521

Post by fennecfoxy »

ChrisRR wrote: post #444
bongodongobob wrote: post #374

Earlier quoted context omitted.

Idk but if computers are drawing enough power to melt themselves just to play ray traced Minecraft or write boilerplate Python code, perhaps we've gone wrong somewhere.

> write boilerplate Python code This is my biggest issue. I'm still impressed that we need GHz and GBs to run simple apps that were achievable with MHz and MBs 20+ years ago

Because we only need hours to write such things, rather than months or years. The market wants rapid development, not necessarily developers. I usually just see the old-heads implying younger devs can't hack it.

Re: Nvidia's RTX 5090 power connectors are melting

#522

Post by fennecfoxy »

kllrnohj wrote: post #378
Joel_Mckay wrote: post #367

Earlier quoted context omitted.

In general, poorly designed cables will use steel instead of tin plated copper conductors. The differences become more relevant at higher power levels, as energy losses within the connectors will be higher. Thus, the temperatures rise above the poorly specified unbranded plastic insulator limits, and a melted mess eventually trips the power supply to turn off. I would assume a company like NVIDIA wouldn't make a powe…

derbauer did an analysis of the melted connector. It was nickel & gold, not steel. buildzodes analysis showed Nvidia removed all ability to load balance the wires which they used to have with the 30xx connector that introduced it, and they had even fancier solutions for the multi 8/6 pin connectors before that. The specification also has an incredibly low safety factor, much much lower than anything that came before.…

Engineering, or executive management?

Re: Nvidia's RTX 5090 power connectors are melting

#523

Post by Joel_Mckay »

fennecfoxy wrote: post #522
kllrnohj wrote: post #378

Earlier quoted context omitted.

derbauer did an analysis of the melted connector. It was nickel & gold, not steel. buildzodes analysis showed Nvidia removed all ability to load balance the wires which they used to have with the 30xx connector that introduced it, and they had even fancier solutions for the multi 8/6 pin connectors before that. The specification also has an incredibly low safety factor, much much lower than anything that came before.…

Engineering, or executive management?

The fact remains a custom after-market cable is not an NVIDIA design problem. If you didn't use the cable in the OEM box, than ethically one should eat the loss if something broke.

I considered publishing a mod tutorial for amateurs (including why crimped-copper is so much better than soldered-steel), and changed my mind given GamersNexus is already looking into the issues. Have a nice day =3

Re: Nvidia's RTX 5090 power connectors are melting

#524

Post by deelowe »

Schiendelman wrote: post #511
deelowe wrote: post #479

Earlier quoted context omitted.

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.

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.

https://www.digikey.com/en/products/detail/texas-instruments...

There should be plenty. 48-54 VDC is the standard for OCP powershelf designs. Hyperscalers such as Google have been working for nearly two decades now to eliminate voltage conversion steps. When I left, the power plane within the server PCB ran at the busbar voltage, which could float up to 54VDC. Given this, I'd expect them to convert from 48-54VDC down to 3.3 directly or at most something like 5VDC and then use smaller VRs near components such as ram and cpu.

Re: Nvidia's RTX 5090 power connectors are melting

#525

Post by fennecfoxy »

Joel_Mckay wrote: post #523
fennecfoxy wrote: post #522

Earlier quoted context omitted.

Engineering, or executive management?

The fact remains a custom after-market cable is not an NVIDIA design problem. If you didn't use the cable in the OEM box, than ethically one should eat the loss if something broke. I considered publishing a mod tutorial for amateurs (including why crimped-copper is so much better than soldered-steel), and changed my mind given GamersNexus is already looking into the issues. Have a nice day =3

Oh, you're the uwu guy, I've seen you around.

I mean that's very true, but that the same time it would be remiss not to chastise Nvidia for selecting a standard that teeters on the edge of design constraints even while using the provided cable. Especially when removing any sort of balancing feature to even things out.

It's the same with the USB standard and various USB cables' current/data carrying capabilities which are often completely transparent to consumers; just totally bad design.

Others have suggested new a new standard that supports 48v or whatever out of the PSU. It does seem a little ridiculous to bring this in for the occasional card that hits a new high but at the same time, why _shouldn't_ our PSUs follow a standard like USB PD has where voltage/current limits budget is negotiated?

That said I'd hope that they include a speck of silicon in cables as well, that outlays the cable's capabilities to the controller on each side - no capabilities present then you only get 5v1a, bub.

What I would hope that this would mean is that we'd no longer have the various different cables, just a single type varying by current/power carrying capabilities.

Re: Nvidia's RTX 5090 power connectors are melting

#526

Post by Joel_Mckay »

fennecfoxy wrote: post #525
Joel_Mckay wrote: post #523

Earlier quoted context omitted.

The fact remains a custom after-market cable is not an NVIDIA design problem. If you didn't use the cable in the OEM box, than ethically one should eat the loss if something broke. I considered publishing a mod tutorial for amateurs (including why crimped-copper is so much better than soldered-steel), and changed my mind given GamersNexus is already looking into the issues. Have a nice day =3

Oh, you're the uwu guy, I've seen you around. I mean that's very true, but that the same time it would be remiss not to chastise Nvidia for selecting a standard that teeters on the edge of design constraints even while using the provided cable. Especially when removing any sort of balancing feature to even things out. It's the same with the USB standard and various USB cables' current/data carrying capabilities which…

48v supplies are not practical for certified end-user products due to National Electrical Codes which limit the risk of electrocution on UL/CE/IC certified power supplies. Note, commercial Telecommunications equipment standards have a very different design market goal, so do use this DC standard already.

The current NEC code (pun intended) effectively limits most supplies to around under 32vDC in most jurisdictions for amateurs (equipment does not need retested in a lab, or signed off by a municipal ticketed EE.) Thus, a standard 24vDC (28vDC actual) equipment rail makes more sense in terms of design economics for the CPU/GPU buck-converters, and is indeed already very common to see in industrial/military rated hardware.

"speck of silicon in cables as well", Maybe... note USBC uses resister values to identify the plug initial behavior... but practically there comes a point where IT workers/hobbyists will have to know first-year ohms law again. At bare minimum people still need to calculate when the wall outlet power-limits are exceeded to avoid posing a fire-hazard in older homes.

I will spare you the xkcd cartoon about standards (and rants about ESD in USBC), but a power-requirement sticker would likely be just as effective for novices.

We all have mixed opinions on NVIDIAs silly GPU design kludges, but the industry has reached a sort of stagnation years ago. Cheers =3

Post reply on HN