Can 12VHPWR issues be fixed?

[droid99]

Try to get this 12V-2X6 / 12VHPWR 90° CABLE PCI-E

 

[iameatingjam]

Try to get this 12V-2X6 / 12VHPWR 90° CABLE PCI-E

I would like something like that... only problem is these connectors ( well not necessarily this one, but ones like it) seem to fail just as often or more. And I have a release model tuf 4090 with the original 12VHPWR that is indeed on an angle ( made it as straight as I possibly could, but there's not a lot of room to work with. Ugh this whole situation is so frustrating.

 

[freeagent]

I just hope and pray I hear the click when I plug it in, since I am waay over 30 cycles lol..

 

[harm9963]

Over a year with my 4090 , at first used CableMod EVGA cable , last three months with a new PSU and WireView GPU from Thermal Grizzly , WireView fits very tight and run my 4090 stock .

 

[lexluthermiester]

What do you think?

No - Just give me more 8-pin connections

I'm with this crowd. There is no reason why they couldn't create a 10pin version of the standard PCIe power jack. 2x10pin PCIe? Yes please and thank you.

All reviews suddenly forgot about the 30 mating cycle this connector has

Also, there's this, something PCIe jacks do not suffer from.

So does 8-pin.

Um, what? I have old PCIe cards that have gone way, WAY past 30 connect/disconnect cycles and there are zero signs of wear on the card jack nor any of my cables.


12VHPWR is a piss-poor design. 12V-2x6 solves nothing, lipstick on a pig.

 

[RJARRRPCGP]

I just hope and pray I hear the click when I plug it in, since I am waay over 30 cycles lol..

This reminds me of the socket pins of LGA CPUs. Been told that I'm possibly doomed if I swap CPUs again. I usually have been switching CPUs on socket 775.

 

[LabRat 891]

This reminds me of the socket pins of LGA CPUs. Been told that I'm possibly doomed if I swap CPUs again. I usually have been switching CPUs on socket 775.

It's totally slipped my mind over the years, that LGA sockets have limited insertion/removal/re-insertion life, too.
Pretty sure that would explain the seemingly unexplainable I've seen on my EP45 board:
>Socket looks good
>carefully and correctly insert CPU
>no POST and/or 'dead' RAM channel
>go to reseat CPU, notice ounc a couple pins are 'collapsed'.
>*CAREFULLY* use eyeglasses' flathead to 'lift' the bent pins
>reinsert CPU, POSTs. G2G, OC.

Also, the 8pins on my Rosewill Hive 1kw are starting to get a little too easy to insert and remove. I mean... it's only been used for mining and at least 2 other builds before mine.

 

[Ruru]

I just want to know the reason what's wrong to NOT use traditional PCIe connectors? They've been in use for ~20 years and they work hella fine.

It's totally slipped my mind over the years, that LGA sockets have limited insertion/removal/re-insertion life, too.
Pretty sure that would explain the seemingly unexplainable I've seen on my EP45 board:
>Socket looks good
>carefully and correctly insert CPU
>no POST and/or 'dead' RAM channel
>go to reseat CPU, notice ounc a couple pins are 'collapsed'.
>*CAREFULLY* use eyeglasses' flathead to 'lift' the bent pins
>reinsert CPU, POSTs. G2G, OC.

Also, the 8pins on my Rosewill Hive 1kw are starting to get a little too easy to insert and remove. I mean... it's only been used for mining and at least 2 other builds before mine.

Remember when there were a batch of Asus mobos with bent pins on stock?

 

[A Computer Guy]

I just want to know the reason what's wrong to NOT use traditional PCIe connectors? They've been in use for ~20 years and they work hella fine.

Perhaps newer generations of PC builders and engineers were emotionally scarred from their Molex experiences and felt they had to do something to improve PCIe connectors.

 

[Ruru]

Perhaps newer generations of PC builders and engineers were emotionally scarred from their Molex experiences and felt they had to do something to improve PCIe connectors.

Molex was also fine with GPU power connectors on the AGP days, they were frustrating on HDDs

 

[lexluthermiester]

I just want to know the reason what's wrong to NOT use traditional PCIe connectors? They've been in use for ~20 years and they work hella fine.

Right? It needed expansion and improvement, not replacement.

 

[Shou Miko]

Using dual 8 pins and will probably buy an AMD card with 8 pins. If it aint broke dont fix it.

I dont want to buy anything with that shit on it.

This is why I went RTX 4070 even I am a big AMD fan, but I made the which this time because the power efficiency is really good on the RTX 4070 compared to the AMD models sadly

I can also understand why having 3 or 4 x 8pins might not be enough for Nvidia on the high-end but going out and blame their customers they cannot figure out how to use their "new" power connector is so wrong but they do not care.

Personally I never really had a card if I remember correct that uses more than 2x8pin I am too cheap to go any higher, I remember when the 6pin came, than suddenly 2x6pin, 1x6 and 1x8pin than 2x8pin which have been solid no real issues and I am afraid going with the 12vhpwr because no warranty when things go wrong it's just something I do not want to deal with.

Using dual 8 pins and will probably buy an AMD card with 8 pins. If it aint broke dont fix it.

I dont want to buy anything with that shit on it.

I agree here with @natr0n if it ain't broke don't fix it Nvidia and if 3x8pin ain't enough well maybe optimize the power efficiency to make the card use less power at the same performance.

 

[lexluthermiester]

This is why I went RTX 4070

Yeah, I'm hitting a memory barrier with my 3080 and I want to upgrade, so I'm going to find a 4070 with standard connectors. The difference is only 2GB VRAM but that 2GB is the difference between my programs taking 4 hours to complete a task and taking 7 hours. It's not the GPU performance, it's the VRAM limit.

 

[LabRat 891]

Right? It needed expansion and improvement, not replacement.

Agreed.
We (the community, at large) do not know the specifics but, there were ulterior motives at play in +12VHPWR's coming to be.
As I (and Igor) recall, 8-pin EPS was on-path to replace PCI-e 8-pin. However, I can also see why that would've been a bad idea.
IIRC, I've read more than a few stories of people managing to He-Man an EPS8 into PCIE8 and visa-versa (which, are inverted in polarity/hot-gnd positions from e/o)

IMHO, It would've made more sense to:
A. Create a new 'optional' (over)spec, built upon PCIE 8-pin with heavier gauge wire, and removing 'sense pins' from the spec entirely. Thus, giving PCIE 8-pin near-equal current carrying ability to EPS 8-pin
(PCIE 6-pin has 2x +12v and 2x gnd. PCIE 8-pin has 3x+12V and 3x gnd, specified to carry current.
Notably, PCIE 6-pin is spec'd, identically to 4-Pin +12vATX; PCIE 8-pin retains 'sense pins', not specified for current-carrying)

B. Add +2 +12V to the existing 8-pin or 6+2(gnd,sense) spec. Making a fully backwards compatible 6+2+2 / 6+4 / 8+2 / 10-pin PCI-E
Preferably: 6 (3x +12v, 3x gnd) + 2(gnd) + 2(+12v), no sense pins.

 

[Massdeth]

That plug is the dumbest thing theyve done with hardware in the pc space for awhile, to this degree. They used smaller pins for higher power, , then moved the plug to the middle of the card on the outside where you cant hide it at all. I suspect this whole thing was because of the size of the chips getting so small,, squeezing more into them with more interconnects and ai etc... to get competitve results from now and in the future they need that power connect as close to the cpu/vram as possible, which does help and will cut down some latency. Plus you got less copper for traces and less chips needed for smoothing,maybe even less layering which all means more money in profit. On top of a proprietary plug with licensing. Greed got us here and it will continue till they destroy themselves with it.

 

[Shou Miko]

Yeah, I'm hitting a memory barrier with my 3080 and I want to upgrade, so I'm going to find a 4070 with standard connectors. The difference is only 2GB VRAM but that 2GB is the difference between my programs taking 4 hours to complete a task and taking 7 hours. It's not the GPU performance, it's the VRAM limit.

Well I never had the chance to try a RTX 3080 I tried the RTX 3070 and 3090 but changed them both out for a RX 6800 XT it was the right choice for me.

As much as I enjoyed my RX 7900 XT with the current power prices I rather loose about 30% performance for 50% less power under full load.

so I am hoping to see CPU 65W and GPU 185W instead of CPU 65W and GPU at least 300W

 

[Zach_01]

I just want to know the reason what's wrong to NOT use traditional PCIe connectors? They've been in use for ~20 years and they work hella fine.

I'd say nVidia chose to go to this compact connector because of the also compact choice of FE PCB design on 4000series to then serve its need of the cooler design



No room for multiple PCI-E connections there is it?
So what about AIB then?



Didn't do it exactly the same but they kept the original power delivery layout most likely for cost reduction.
Because you know... AIB profit margins are so high for them.../s

 

[rv8000]

I can also understand why having 3 or 4 x 8pins might not be enough for Nvidia on the high-end but going out and blame their customers they cannot figure out how to use their "new" power connector is so wrong but they do not care.

Both 3x8 and 4x8pin can supply more power than the ridiculous 12vhpwr connector.

 

[lexluthermiester]

I'd say nVidia chose to go to this compact connector because of the also compact choice of FE PCB design on 4000series to then serve its need of the cooler design

View attachment 334568

No room for multiple PCI-E connections there is it?
So what about AIB then?

View attachment 334569

Didn't do it exactly the same but they kept the original power delivery layout most likely for cost reduction.
Because you know... AIB profit margins are so high for them.../s

The laws of electro & thremo dynamics are very clear about energy densities. This connector was poorly(read badly) designed. It puts too much power through too little a space.

So to answer your question, they should chosen a connector that has a proven track record for both functionality AND safety. Changes to the design of the PCB to accommodate the larger/safer connectors would have happened naturally.

This new connector was made and used because of penny-pinching aholes that care more about profits than the use of common sense.

 

[Panther_Seraphin]

Both 3x8 and 4x8pin can supply more power than the ridiculous 12vhpwr connector.

4x8 pin meets the same power capacity while staying in spec.

3x8 CAN do it but you are technically out of spec and are into the safety factor in the design.

This new connector was made and used because of penny-pinching aholes that care more about profits than the use of common sense.

I would argue it more of the "Mac Effect", form over function. They needed a high power density connector to make it so THEIR cooler design would fit/work. I wonder how many electrical engineers/mechnical engineers were overridden in these choices.

 

[rv8000]

4x8 pin meets the same power capacity while staying in spec.

3x8 CAN do it but you are technically out of spec and are into the safety factor in the design.

I would argue it more of the "Mac Effect", form over function. They needed a high power density connector to make it so THEIR cooler design would fit/work. I wonder how many electrical engineers/mechnical engineers were overridden in these choices.

You’re at a greater risk using 12vhpwr to deliver 600w than with 3x8pin, cable quality being the same (and good).

The only benefit there is to 12vhpwr is that its a single cable vs 3. Its proven to otherwise be a headache so far on a card that is only “spec’d” to draw 450w, same goes for the revision.

Ultimately it is what it is, but the thought process behind making this an actual spec and cable seems to not logic.

 

[lexluthermiester]

3x8 CAN do it but you are technically out of spec and are into the safety factor in the design.

That depends on the default clocks and boost power levels. If those were given a limit the connector would be compliant.

I would argue it more of the "Mac Effect", form over function. They needed a high power density connector to make it so THEIR cooler design would fit/work

I doubt that. This connector had been in the works before the RTX3000 series and the heatsink/fan design of the FE could be easily adapted to fit them.

However, if they had redesigned the exiting PCIe connector to a 10 or 12 pin version, specs would have been more flexible.

You’re at a greater risk using 12vhpwr to deliver 600w than with 3x8pin, cable quality being the same (and good).

This and it has to do with the amount of power flowing through a given space. 3x8pin connectors have more physical space to emit waste heat. The 12VHP connector does not have that design advantage.

 

[Zach_01]

No argue there...
The issue with this 12vhpwr connector is when there is mis-conductivity (even partially) between male-female pins because each pin pair has to sustain a certain amount of current and keep temp reasonable.
Common sense is to spread the same current across more pin pairs, but the nVidia PCB/cooler design does not allow this.
I guess nVidia wanted to make the difference in aesthetics and FE cooler performance and that circled with a hard bite

 

[Panther_Seraphin]

That depends on the default clocks and boost power levels. If those were give a limit the connector would be compliant.

3x8 pin is rated for 450 watts max so it will never meet what in theory the 12VHPWR is rated for.

I doubt that. This connector had been in the works before the RTX3000 series and the heatsink/fan design of the FE could be easily adapted to fit them.

So looking at the space required for at least 3x8 pin and 4x8 pin for 4090s it wouldnt be feasible without quite a decent rework of PCB design and then the indirect impact on airflow/fan size etc etc etc

However, if they had redesigned the exiting PCIe connector to a 10 or 12 pin version, specs with have been more flexible.

We end up at the same problems with the 12VHPWR. Do you increase the current capacity of the pins/cables to make up the power requirement? Do we say 12 Pins can do ~450 watts? (keep roughly in the same safety factor of existing mini fit connectors) What does Nvidia do then. 12 pin + 6 pin? They wanted a 600 watt capable card in case of AMDs chiplet gamble paying off and it coming up close to 4090 esque performance. Hell there is even supposedly 1000 watt BIOS's floating around for 4090s.

 

[rv8000]

3x8 pin is rated for 450 watts max so it will never meet what in theory the 12VHPWR is rated for.

Nothing to do with theory, you can literally pull more power through 3x8p than 12vhpwr.