12VHPWR has a higher number of reported failures than 6/8 pin does despite being much younger

[AusWolf]

Look at this sequence:
year 1: 1/20
year 2: 2/30
year 3: 4/40
year 4: 7/48
year 5: 9/56
year 6: 8/62
That is the sort of growth in failure rate that we would expect to see as the connector saturates the market and the failure rate grows with age and then stabilizes. Adoption rate slows down while failure rate lags behind, grows faster, and then stabilizes.

This is exactly what I was talking about. If you get such data from every year, and look at the number of connectors that failed each year, then it's logical to draw a conclusion like you did. However, if you only look at one single point of data, then there's no correlation to be seen, as a single point of data never correlates with anything and never bears any significance.

 

[eidairaman1]

New cable is beefy .

Do you have the old 1 to compare it with?

 

[ty_ger]

This is exactly what I was talking about. If you get such data from every year, and look at the number of connectors that failed each year, then it's logical to draw a conclusion like you did. However, if you only look at one single point of data, then there's no correlation to be seen, as a single point of data never correlates with anything and never bears any significance.

I agree.
So why is your conclusion that the data won't change and age has no relevance? Your hypothesis was that since these are percentages, the 8pin is actually currently the worse product, when in actuality, when taking into account age, the 12vhpwr connector is the worse product.
Like the post in post #7 and the discussion which follows it, your hypothesis is essentially that age has no relevance, the data is meaningless, no conclusions can be drawn, and if anything, the 8pin is currently the worse product. You and them throw out age as a relevant factor and then come to an illogical conclusion. This thread is about age and product maturity. It's in the title.
The OP's hypothesis is that given the age and lack of product maturity, reaching failure percentage parity at such a young age is a bad sign. It matches expectations given the lower product safety factor, so it is not a surprise, but does confirm that there isn't some redeeming factor that we forgot to consider. I agree.

 

[AusWolf]

I agree.
So why is your conclusion that the data won't change and age has no relevance? Your hypothesis was that since these are percentages, the 8pin is actually currently the worse product, when in actuality, when taking into account age, the 12vhpwr connector is the worse product.

I have no idea where you got that from, it is not my conclusion at all. All I said was, the diagram screenshotted into OP is one single data point in time which you can't use to draw any conclusion.

Like the post in post #7 and the discussion which follows it, your hypothesis is essentially that age has no relevance, the data is meaningless, no conclusions can be drawn, and if anything, the 8pin is currently the worse product. You and them throw out age as a relevant factor and then come to an illogical conclusion. This thread is about age and product maturity. It's in the title.

No conclusion can be drawn from one single data point. No one said that the 8-pin was a worse product.

I don't remember articles popping up about connectors catching fire when the 8-pin was introduced, so it's a better product at least in that regard. Time will decide about the rest.

 

[ty_ger]

You continue to say that no conclusions can be drawn. And you said, number-wise, the 8pin is currently worse since it has a greater total failure quantity; as if that somehow has any affect on this discussion.

lol

I think I analyzed your opinion and illogical conclusion just fine.

 

[Vincero]

The only problem here is there are no data points at all. The GN graph has started an idea / argument with people thinking they can attribute some real world statistical data to it.
They can't.

To quote myself:

All GN can say for sure is that, statistically, people who might have some experience with either or both types of connector have a higher experience of some kind of failure of either the card or intermediate connection cable related to the newer 12VHPWR type. The problem with that is, as GN touched upon, if the connector happened to be the Nvidia supplier 3/4xPCIe>1x12VHPWR adapter, part of the problem could easily be a bad connection on the PCIe cabled/connector side having a knock-on effect to the 12VHPWR side... or maybe the PSU maker supplied cable and connection point which may not even be 12VHPWR style... this doesn't help statistics and perception.

 

[AusWolf]

You continue to say that no conclusions can be drawn. And you said, number-wise, the 8pin is currently worse since it has a greater total failure quantity; as if that somehow has any affect on this discussion.

That was some example math, not a conclusion. A conclusion can only be drawn when the issue has been investigated from multiple points of view, and with adequate amount of data.

I think I analyzed your opinion and illogical conclusion just fine.

You certainly did not.

 

[freeagent]

On my low power, 300w GPU it is great. In a tight case like mine it makes routing a ton easier. I can’t speak for the higher powered cards though..

 

[harm9963]

Do you have the old 1 to compare it with?

 

[eidairaman1]

View attachment 367453View attachment 367453

I presume the 1 on the right is the older cable

 

[harm9963]

I presume the 1 on the right is the older cable

Yup , and both are 3.0ATX OEM ,for my PSU , wasn't expecting a updated cable , will ask for information at some point ,still testing the new cable , will not remove or open the case , want to observed 12v for deviation thru many heat cycle's , will update in a few months .
Did a 30 min loop , max v1.100 from 1.050 ,1000 memory and 160 OC on core.

https://gpuz.techpowerup.com/24/10/13/dnd.png

https://gpuz.techpowerup.com/24/10/13/n4g.png

 

[Vario]

I am using a modDIY cable, has been fine for the past 4 months. Hopefully it will continue to be fine. This cable has a right angle formed by bending the wires rather than a rigid right angle, and it plugs into my SuperFlower's 2x8 pin ports. I think it is this model (https://www.moddiy.com/products/Ang...ual-8-Pin-to-16-Pin-Power-Cable-for-EVGA.html).
Also my 4070 Ti Super barely draws any current, it is REALLY efficient.

 

[ty_ger]

That was some example math, not a conclusion. A conclusion can only be drawn when the issue has been investigated from multiple points of view, and with adequate amount of data.

Example math was used because you didn't seem to understand. You still can't understand. The numbers are an example, but the conclusion is fact. For reasons explained, new adoption rate decreases while fail rate increases. I thought an example would help you see the numbers, but it seems you are wilfully blind.

We had adequate data from the start. It's in the datasheet. It is inferior and will have a higher failure rate. Guaranteed. They even warned us that this product is not as resilient as we are used to and wears out faster than we would normally expect.

 

[eidairaman1]

Yup , and both are 3.0ATX OEM ,for my PSU , wasn't expecting a updated cable , will ask for information at some point ,still testing the new cable , will not remove or open the case , want to observed 12v for deviation thru many heat cycle's , will update in a few months .
Did a 30 min loop , max v1.100 from 1.050 ,1000 memory and 160 OC on core.

https://gpuz.techpowerup.com/24/10/13/dnd.png

https://gpuz.techpowerup.com/24/10/13/n4g.png

View attachment 367457

After this fiasco i would be keeping an eye on it every so often

 

[evernessince]

Except that AI has almost no use in the consumer space, unlike crypto.

I think you are under-estimating just how many people buy hardware for running local AI. There are 557K members on the stable diffusion reddit and 227K on the Local Llama reddit. The former is in the top 1% size wise of all subreddits. Both of them focus on running LLMs on consumer hardware.

There's a reason the 4090 has sold so well, it's a good card for running AI. The 8 and 12 GB cards lower in the stack aren't great for running LLMs but they are good enough for some people while for others the card might be pulling double duty for both gaming and LLMs. Regardless, that's more people buying more cards who are also willing to pay more due to the added functionality. That plus Nvidia having the enterprise market on lock doesn't really help gamers as it greatly reduces their levarage to push for lower prices.

At the end of the day it's a simular situation to crpto. The only thing that's going to help is the AI market switching to dedicated AI accelerators.

I'd have to agree with @R0H1T here, people in the local LLM community would absolutely push for that if it meant more performance / VRAM. Heck if you head over to local llama you can see people already running 2-3 cards already.

 

[AusWolf]

Example math was used because you didn't seem to understand. You still can't understand. The numbers are an example, but the conclusion is fact. For reasons explained, new adoption rate decreases while fail rate increases. I thought an example would help you see the numbers, but it seems you are wilfully blind.

Please continue your personal crusade against me to prove how wrong I am based on things I never said. I'm sure it'll make you feel like the proud keyboard warrior that you are in the end. I'm out.

We had adequate data from the start. It's in the datasheet. It is inferior and will have a higher failure rate. Guaranteed. They even warned us that this product is not as resilient as we are used to and wears out faster than we would normally expect.

The data is probably there somewhere, just not in that one diagram that OP posted. For a start, I'd look at the connector and wire gauge which made it visible how inferior the new standard is right from the start. More power through thinner cables and a smaller connector... You don't need (flawed) statistics to see where things are going.

See? We're not on different sides here. All I'm saying is that one single diagram with that one single data point is inadequate to prove anything.

 

[londiste]

The data is probably there somewhere, just not in that one diagram that OP posted. For a start, I'd look at the connector and wire gauge which made it visible how inferior the new standard is right from the start. More power through thinner cables and a smaller connector... You don't need (flawed) statistics to see where things are going.

You are not quite correct on wires part. PCIe 6/8-pin wires are usually 18AWG, sometimes on quality PSUs 16AWG. IIRC the spec allows 20AWG but nobody is going that far these days. 12VHPWR requires 16AWG from get-go.

 

[AusWolf]

You are not quite correct on wires part. PCIe 6/8-pin wires are usually 18AWG, sometimes on quality PSUs 16AWG. IIRC the spec allows 20AWG but nobody is going that far these days. 12VHPWR requires 16AWG from get-go.

That would only put it on par with the old connector, except for two things... 1. The new connector is smaller and flimsier, 2. The old connector is rated for 7-8 A per pin, the new one at 9.5 A per pin (Wikipedia). At a similar wire gauge, I'd naturally expect it to be hotter and get damaged sooner than the old one.

Source:

 

[dgianstefani]

On my low power, 300w GPU it is great. In a tight case like mine it makes routing a ton easier. I can’t speak for the higher powered cards though..

Fine here at 400 W.