Can 12VHPWR issues be fixed?

[Panther_Seraphin]

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

Out of offical spec yes. At that point what is the point of safety factors/stated specs?

Lets just run everything right up to the point before spontaneous combustion.......just like the 12VHPWR...Oh Wait thats the whole problem we have run into currently.

 

[rv8000]

Out of offical spec yes. At that point what is the point of safety factors/stated specs?

Lets just run everything right up to the point before spontaneous combustion.......just like the 12VHPWR...Oh Wait thats the whole problem we have run into currently.

The point being you have 0 safety margin left with a 12vhpwr cable, yet 150w more with 3x8pin with more surface area and overall dissipation of heat spread across more wires, more connectors, and larger more secure contact pins.

TLDR

Only advantage is theres a singular cable, and otherwise inferior in terms of power delivery and safety.

 

[lexluthermiester]

Out of offical spec yes.

The defined specs account for a +/- 15% variance.

 

[Klemc]

Corsair, bc i buyed Shift 1000 when out, sends me a newer 12VHPWR cable VESA updated !

 

[oxrufiioxo]

I wouldn't link the tracking information on a public site @Klemc

 

[the54thvoid]

Corsair, bc i buyed Shift 1000 when out, sends me a newer 12VHPWR cable VESA updated !

Removed link for safety.

 

[delshay]

Out of curiosity, who voted for "what is electricity" ???

 

[LabRat 891]

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.

Not sure of cycle life but,
something like the connections for modular redundant PSUs, seems like it would've been an appropriate direction (for compact-ness)

HP Common Slot is good for something like 1.2-2kw. I'd imagine much smaller pads would suffice.

Basically, something like the old 5.25 FDD connector. Maybe even smaller?

Like the AUX Power input on Apple AGP Pro cards but, bigger and on the top/front of the card, integrated into the PCB.

 

[Panther_Seraphin]

The defined specs account for a +/- 15% variance.

3x8 pin at 15% over spec is still only ~510 watts vs the 600 of the HPWR. But yeah if you run it right up to the safety factor HWPR has it would be just over 750 watts in theory.

 

[lexluthermiester]

3x8 pin at 15% over spec is still only ~510 watts vs the 600 of the HPWR.

Yeah, you're kinda missing the point. We're not talking about absolute wattage. We're talking about safely. The 12VHP connector is a safely hazard. It doesn't matter if the standard PCIe connectors can handle the same wattage or not, they are SAFE. Let's be real, the 12VHP isn't either or we wouldn't be having these problems or this conversation. That's the key point and the most important point.

 

[R-T-B]

Yeah, you're kinda missing the point. We're not talking about absolute wattage. We're talking about safely. The 12VHP connector is a safely hazard. It doesn't matter if the standard PCIe connectors can handle the same wattage or not, they are SAFE. Let's be real, the 12VHP isn't either or we wouldn't be having these problems or this conversation. That's the key point and the most important point.

I'm not sure I agree that a properly built one is a safety hazard, but knowing how manufacturers fudge things, I am increasingly feeling it should probably have had a larger margin built in ala the older pcie power plugs (well maybe not THAT much but certainly more than it has).

 

[lexluthermiester]

I'm not sure I agree that a properly built one is a safety hazard

That's the problem. "Properly" built jacks have been melting too. The problem is triple fold.
A. The design is trying to push more power through a small space than is safe.
B. The design of the connector leaves no room for error.
C. If not properly connected the connector becomes a safety hazard.

There are a number of reviewers and industry testers who have definitively shown that the connector can LOOK properly connected even when it isn't. This problem does not exist with the standard PCIe connectors.

 

[R-T-B]

There are a number of reviewers and industry testers who have definitively shown that the connector can LOOK properly connected even when it isn't.

Yeah the new revision is supposed to address that little part but I was trying to speak to the broader, base issues.

 

[lexluthermiester]

Yeah the new revision is supposed to address that little part but I was trying to speak to the broader, base issues.

Even the revision is lipstick on a pig. They need to take a step back and improve the existing, well tested and well proven connectors. The 12VHP needs to go away.

 

[R-T-B]

Even the revision is lipstick on a pig. They need to take a step back and improve the existing, well tested and well proven connectors. The 12VHP needs to go away.

I'd agree, but a.) I know it won't and b.) the numbers still suggest the issue is small.

I would've liked them to think it out more yes and it is indeed pretty damn criminal they didn't but sadly I feel the industry ship has sailed. I'm not fearful of cards with it atm, but certainly don't hate the fact that my current radeon did without it, either.

Given a choice I'd take pcie over it for peace of mind, I think.

 

[Panther_Seraphin]

Yeah, you're kinda missing the point. We're not talking about absolute wattage. We're talking about safely. The 12VHP connector is a safely hazard. It doesn't matter if the standard PCIe connectors can handle the same wattage or not, they are SAFE. Let's be real, the 12VHP isn't either or we wouldn't be having these problems or this conversation. That's the key point and the most important point.

Oh I agree that the 12VHWPR was a complete dumpster fire from inception.

I am just pointing out that when people are saying "but the old 8 PIN is so much better" etc etc etc is because of that health safety margin installed. Its the same with higher end Motherboards/GPUs etc. You look at their power stages and they are rated for some INSANE power outputs so for example the Asus 4090 Strix has something like 3000 watts of Core power delivery at 1.8 VCore so ~1800 at 1.1 Volts. Yet you barely see 1kw of draw even under LN2 etc.

Even the revision is lipstick on a pig. They need to take a step back and improve the existing, well tested and well proven connectors. The 12VHP needs to go away.

I would say with the amount of power draw we are starting to expect we just need to move away from 12v being the primary input voltage. We have 400 watt CPUs and 400 watt GPUs becoming MAINSTREAM. Moving to 24 or God forbid 48 volts just removes the current handling completely and we go back to early 2000s current draw even with such high power usage. We can go back to extreme PCs having 20/24 PIN mobo with MAYBE a 4 Pin CPU connector and an old HDD Molex connector for GPUs/other cards and be safer than current.

 

[lexluthermiester]

Oh I agree that the 12VHWPR was a complete dumpster fire from inception.

I am just pointing out that when people are saying "but the old 8 PIN is so much better" etc etc etc is because of that health safety margin installed. Its the same with higher end Motherboards/GPUs etc. You look at their power stages and they are rated for some INSANE power outputs so for example the Asus 4090 Strix has something like 3000 watts of Core power delivery at 1.8 VCore so ~1800 at 1.1 Volts. Yet you barely see 1kw of draw even under LN2 etc.

People and GPU makers need to learn how to do more with less or build a MUCH better connector.

The engineering needed to been done better.

Given a choice I'd take pcie over it for peace of mind, I think.

This is why I'm ditching my 3080 in favor of a 4070 that has standard connectors. I'm tired of worrying that my home might burn to the ground because one of those damned pins goes wonky while in the jack..

 

[Arkz]

This is why I'm ditching my 3080 in favor of a 4070 that has standard connectors. I'm tired of worrying that my home might burn to the ground because one of those damned pins goes wonky while in the jack..

lol. It's a 3080.
A 4090 pulling 600W under full load is one thing to worry about. A 3080 pulling about 320W is perfectly fine, not even gonna warm up the pins.

 

[lexluthermiester]

lol. It's a 3080.
A 4090 pulling 600W under full load is one thing to worry about. A 3080 pulling about 320W is perfectly fine, not even gonna warm up the pins.

Don't care. I don't trust it. I've gotten close to 460W on this card BTW and the connector was very warm. Granted, I'm going up to a 4070 for other reasons as well, but I'm not sorry to see that damn thing go.

 

[Chrispy_]

What we had with two PCIe power connectors was 12 large pins delivering 300W. That's 50W per circuit and 4.17A per wire.

12VHPWR also has 12 pins, so for the same safety margin as PCIe it would be a 300W connector with no more than 50W per circuit and no more than 4.17A per pin.

Except they also made the pins smaller, so we know that its less safe then 2x 8-pin because laws of physics apply - specifically Joule's first law and Ohm's law. This is grade school stuff, nothing complicated.

Doing some napkin math on the mating area of the two connectors, I'm going to ballpark Amphenol's Minitek interface current-handling capability at about 75% of Molex's MiniFit Jr's. The the total contact surface area of each pin/barrel interface seems to be about half but Amphenol have done some flexing beam thing rather than dimples to provide contact force so let's give them some margin back. However, the change from dimples to flexing beam is unlikely to really account for the overly-generous gain I'm giving it since ECR (electrical contact resistance) is usually so negligible that it's difficult to measure and can be practically ignored as other resistivity factors scale faster with current and temperature. There's no getting around the obvious fact that smaller pins simply have less contact surface to carry current.

Using that rather generous 75% capability value, the Amphenol Minitek power connector at 75% of MiniFit Jr's 300W is a 225W connector. I don't care if Amphenol say it's rated for 9.5A per pin, it's a significant downgrade from MiniFit Jr. and looks closer to 2.5A per pin to me. At just 6.25A per pin (so 450W as per the connector rating) it's already being asked to do double what we've been calling "safe" for the last 20 years.

I'm not against a new connector. Using 4x 8-pin is pretty silly for these behemoth 600W GPUs, but at the same time, I'd like it to be better, more reliable, and safer (or at least as safe) than our existing PCIe standard. There are so many wasted pins in multiple PCIe connectors and quite honestly the 450W GPUs in question have no shortage of physical space to accommodate something larger than 12VHPWR.

 

[lexluthermiester]

Except they also made the pins smaller, so we know that its less safe then 2x 8-pin because laws of physics apply - specifically Joule's first law and Ohm's law. This is grade school stuff, nothing complicated.

Exactly! Thank You!

 

[xorbe]

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

This is the ENTIRE crux of the situation, 3x8 exceeds its rating by a healthy margin, whereas 12VHPWR apparently doesn't. Who cares about specs and ratings, when 3x8 works while 12VHPWR is melting.

 

[SPDIF]

As long they produce cards that swallow 600W, the problem will never go away. What's next? A 1KW video card and a standard 1.5KW power supply? The environment and your children thank you for that... It's completely ridiculous to play a game you swallow 600W from the power outlet... And most of that is being immediately transferred in heath. You easily can heat your house with that.

 

[Zach_01]

As long they produce cards that swallow 600W, the problem will never go away. What's next? A 1KW video card and a standard 1.5KW power supply? The environment and your children thank you for that... It's completely ridiculous to play a game you swallow 600W from the power outlet... And most of that is being immediately transferred in heath. You easily can heat your house with that.

Watts/power/current is not necessarily the problem for the melting connectors. Trying to invent and standardize your own creation without proper development (R&D) for the actual standards and safety features needed… is the problem.

A “normal” 300+mm long PCB card can easily fit 4x8-pin PCI-E connections that can power even a 650-700W GPU 24/7 if you account 50-60W from the board.
And I’m really confident that most of these 8-pins can handle 200W each.

…until we meet a better connector type or even increased voltage, like 12 >> 24V which will cut current (A) in half. Like some high powered EVs are doing, going from 400V to 800V

 

[Chrispy_]

Watts/power/current is not necessarily the problem for the melting connectors. Trying to invent and standardize your own creation without proper development (R&D) for the actual standards and safety features needed… is the problem.

A “normal” 300+mm long PCB card can easily feet 4x8-pin PCI-E connections that can power even a 650-700W GPU 24/7 if you account 50-60W from the board.
And I’m really confident that most of these 8-pins can handle 200W each.

…until we meet a better connector type or even increased voltage, like 12 >> 24V which will cut current (A) in half. Like some high powered EVs are doing, going from 400V to 800V

The 8-pins are, worst-case, rated to 288W based on Molex MiniFit Jr specifications. The fact PCIe CEM standard limits them to 150W is safety margin to account for damaged connectors and cheap Chinese manufacture.