News RTX 4090 16-Pin Cable Surveys Indicate Some Use 'Under Spec' Wiring

[Admin]

RTX 4090 16-pin adapter surveys indicate that many use AWG14 wires but a few use AWG16. Meanwhile, a popular TechTuber has not been able to reproduce the cable melting problem even when purposely damaging cables and overclocking for many hours.

RTX 4090 16-Pin Cable Surveys Indicate Some Use 'Under Spec' Wiring : Read more

 

[btmedic04]

zotac doing zotac things i see

 

[hannibal]

Or it is just gamble and you get what you get...

 

[mfellows]

If the adapter ideally uses 14g, should we be worried about CableMod using 16g for their direct to PSU 12vhpwr cables?

 

[-Fran-]

As I said in the Discord, someone (company or individual) really screwed nVidia with this one.

Regards.

 

[TechieTwo]

From a technical standpoint a 16 AWG stranded wire of one foot can easily and safely carry 18 amps continuously, i.e. 216 watts per wire @ 12v. Four 16 AWG power supply wires thus can safely supply 864 watts - far beyond the power consumption of the RTX 4090. Thus the 16 AWG wire should not be an issue in overheating. Poor / high resistant connection of terminals to the wires or terminal-to-terminal contact however can cause overheating issues.

 

[logainofhades]

Yea this stuff is melting due to poor connections. A broken wire, as long as said wire isn't close enough to arc, it also will not cause a melting of the plug. I believe jay proved that one, when he clipped the wires, and nothing melted. A broken solder joint, though, is most likely going to arc and melt the connector.

 

[Kamen Rider Blade]

I thought nVIDIA was sole sourcing the Adapters for QC reasons.

Why is there so much manufacturing variance in the parts including some 150V & 300V cables?

Astron being shoddy / shaddy again?

 

[DavidLejdar]

From a technical standpoint a 16 AWG stranded wire of one foot can easily and safely carry 18 amps continuously, i.e. 216 watts per wire @ 12v. Four 16 AWG power supply wires thus can safely supply 864 watts - far beyond the power consumption of the RTX 4090. Thus the 16 AWG wire should not be an issue in overheating. Poor / high resistant connection of terminals to the wires or terminal-to-terminal contact however can cause overheating issues.

So technically, could a 300(+75) Watt GPU be easily supplied via a 6-pin connector?

 

[PlaneInTheSky]

$1600 GPU with cables that can catch fire.

The future is here ladies and gentlemen.

 

[Amdlova]

When a power supply dies I cut the cables to use in other projects and the reality same model has bad cables, chineses manufactures tend to make the profit way and fake the products. Need to measure the wire instead of the printing. I see a lot of aluminum cables or something like a fake cooper.

 

[JarredWaltonGPU]

So technically, could a 300(+75) Watt GPU be easily supplied via a 6-pin connector?

If you look at the 8-pin connector vs. the 6-pin, the two extra wires are just ground pins, meaning all the power is still coming over six wires.

 

[cia1413]

If the adapter ideally uses 14g, should we be worried about CableMod using 16g for their direct to PSU 12vhpwr cables?

It looks to me that the 14 gauge adapter is only 4 wires, and the cablemod cable looks to be 6 wires. Sounds fine to me. 14 gauge is 10-15 amps and 16 is 10ish.

 

[InvalidError]

The only difference between 150V and 300V wiring is insulation thickness, which doesn't matter at 12V apart from the extra cut/abrasion/pinch resistanceafforded by more insulation between the conductors and whatever is touching the wires.

It looks to me that the 14 gauge adapter is only 4 wires, and the cablemod cable looks to be 6 wires. Sounds fine to me. 14 gauge is 10-15 amps and 16 is 10ish.

15A is insulated wall rating for #14. For a ventilated enclosure like a PC case usually is, it is 32A for #14, 22A for #16 and 16A for #18.

So even 6x#18 would still technically have a 100% safety factor at 50A if GPUs enforced current balance across all six 12V pins. One positive upshot to increasing wiring resistance is it it reduces the proportion of total wiring resistance contributed by connectors which would passively help enforce current balance, albeit at the expense of increasing wiring losses by ~1.5%.

 

[TechieTwo]

So technically, could a 300(+75) Watt GPU be easily supplied via a 6-pin connector?

Yes it can.

 

[Phaaze88]

Wha?
So we're back to user error until more info is available?

 

[PEnns]

Damage control is in full swing and blaming the user for this fiasco is still priority # 1 for Nvidia and its shills.

 

[InvalidError]

Wha?
So we're back to user error until more info is available?

If nobody is able to intentionally reproduce the issue even using combinations of hypothesized failures, then the people who really got melted connectors either did something unusually bad or were very unlucky.

From reports so far, it seems Igor's 1/2/1 150V cable is the unicorn variant that nobody else has seen, maybe all of the failures are from a batch of those that slipped into retail boxes.

 

[daworstplaya]

zotac doing zotac things i see

And yet, knock on wood, we haven't seen any melted connectors on Zotac cards.

 

[hotaru251]

isn't close enough to arc

can 12volt even actually arc?

I know it could spark by brushing against stuff but idk if its powerful enough to actually have energy to arc across an air gap.

 

[Phaaze88]

can 12volt even actually arc?

I know it could spark by brushing against stuff but idk if its powerful enough to actually have energy to arc across an air gap.

Buildzoid was corrected about that in his video on the adapter - he edited one of his posts in the comments:
EDIT: "my statement that 12V can't cause arcing isn't entirely accurate. While it's technically true that 12 VOLTS can't cause an arc. Inductive loads(like motors) powered by 12V can cause arc when suddenly disconnected due to the inductance causing a temporary high voltage spike. This however is not relevant to the discussion at hand as you aren't gonna get an inductive voltage spike when one or several pins in the 12VHPWR lose contact. (you might get a inductive voltage spike if the connector is entirely disconnected)"

 

[Exploding PSU]

zotac doing zotac things i see

I've always considered Zotac as one of the better, upper-tier NVIdia AIBs out there. They look solid and their designs are generally on the cute side.

 

[InvalidError]

can 12volt even actually arc?

The minimum voltage required to create an arc over an air gap large enough to prevent electron tunnelling directly across under normal atmospheric conditions is about 300V according to Paschen's law. The sparks that come out when you live-make contact with a substantial resistive load or almost anything with input capacitors is the initial point of poor contact getting hot enough to get vaporized into plasma and the arc going through the metal plasma like a scratch-start welder, not an arc striking first through air.

 

[Endymio]

From a technical standpoint a 16 AWG stranded wire of one foot can easily and safely carry 18 amps continuously, i.e. 216 watts per wire @ 12v.

It's not nearly so simple. The ampacity of a certain gauge wire depends on the ambient air temperature, the maximum temperature the wire itself is allowed to rise to, the proximity of other current-carrying conductors, and a few other factors. A 16 AWG wire carrying 18 amps is going to exceed 170F degrees F -- and hotter still bundled together in a 4-conductor cable. I'm quite sure there's a reason that 14 AWG is specified ... that will have an operating temperature some 20 degrees cooler.

 

[edzieba]

150V vs. 300V rating should be an absolute nonissue for devices running on 12VDC (i.e. if they ever see appreciable more than 12V then your device is already dead long before the connector could cook). The voltage rating is to do with the insulation properties of the wire, and has effectively nothing to do with the current capacity*.

The assembly process changes found, extremely limited sample size of failures, lack of commonality of failures, and inability to replicate the failures, all point to an atrociously bad batch of parts that have gotten out into the environment. A huge QC fail for someone, but not the sort of fundamental design flaw that many have been crowing about. With the indication that the overheating failure was documented prior to release, there's a good chance this was a bad batch that was found during QC, the cause diagnosed and fix rolled out, but somehow some faulty runs did not make their way straight to the scrap bin but instead were packaged for final shipment. Could be anything from and unscrupulous manufacturer somewhere along the chain (not Nvidia, possibly not even a card OEM, e.g. could be a subcontractor hired to distribute the adapters) to an ambiguously labelled parts bin in the wrong place at the wrong time.

*Caveat only in that conceivably in a specialised ultra-high-temperature application where wiring is expected to operate at several hundred °C as standard, the additional heating from a high current load could push an insulator from marginal to failure at the same voltage. This is not in any way the case for a GPU in a consumer case, nobody is gaming on the surface of Venus.