News MSI Launches Power Supply With Yellow Connectors to Prevent 16-Pin Connector Meltdowns

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
the atx molex connectors themselves need to be redesigned. they were never meant for 250W+ CPUs or 500W+ GPUs running on DC power. DC power is a different beast at high amperage and needs to be given far more respect and safety margin than is currently given. if a conductor is underrated or badly seated because of an aged connector design, it will melt the insulation leading to damaged parts. the solution is a connector that is easier to align, takes less torque to insert and is made for repeated plugging and removal, and a latching clip that won't click unless the coupling is properly seated, like what is used on modern automotive connectors. there is a reason why the automotive and aerospace industry has moved on from molex and amphenol style connectors from the 1980s. they are not weather proof, don't have click latching, and are difficult to align, and easy to achieve intermittent connections. they are also not capable of handling vibration and resonances very well. its time to move on to newer designs and look at what the automotive and aerospace industry uses to solve these problems.
 
purpleduggy said:
its time to move on to newer designs and look at what the automotive and aerospace industry uses to solve these problems.
Click to expand...
Ah, yes, lets make connectors 10X more expensive for no benefits at all as far as typical PC environments are concerned. PCs don't need to be weatherproof, don't really go in high-vibration environments, usually are in temperature-controlled places, etc.
 
  • Like
Reactions: TJ Hooker and RandomWan
newtechldtech said:
it is time to move to USB-C power and cables .. each USB-C PD can deliver 240 watts per port. and it is safe and friendly ! and even smaller !
Click to expand...
Only problem with it: where are you getting the 48V from?

Intel's 12VO was a missed opportunity to also bump the main system voltage to at least 24V for even better bulk power distribution efficiency and have a 20V source to run USB3.x/4/TB monitors, external HDDs, USB/TB hubs and other 20-100W external devices.
 
  • Like
Reactions: purpleduggy
Amdlova said:
why not put a fuse per pin :) if goes beyond of spec the fuse die and everyone is happy
Click to expand...
Depending on the rating of the fuse being used, it may not help since fuses don't act instantly. See:

Also the problem isn't current going through the cable. It's the connection isn't good enough (due to improper installation) to support the power going through it. The only thing that'll protect it here is a temperature monitor to make sure the connector isn't getting excessively hot.

EDIT: Actually the component the video mentions may help with that, since its resistance shoots up with temperature.

newtechldtech said:
it is time to move to USB-C power and cables .. each USB-C PD can deliver 240 watts per port. and it is safe and friendly ! and even smaller !
Click to expand...
USB-C is only rated for 5A maximum per pin. As mentioned, you're going to need to generate 48V somewhere for the 240W you want. And even then, only E-marked cables will do that since the USB-PD standard doesn't allow any ol' cable to transfer power past 65W.

Also since you need 3 of them to satisfy the up to 600W requirement and you'd probably want to space out the ports to fit most connectors, you're not saving any space.
 
Last edited:
hotaru.hino said:
USB-C is only rated for 5A maximum per pin. As mentioned, you're going to need to generate 48V somewhere for the 240W you want. And even then, only E-marked cables will do that since the USB-PD standard doesn't allow any ol' cable to transfer power past 65W.

Also since you need 3 of them to satisfy the up to 600W requirement and you'd probably want to space out the ports to fit most connectors, you're not saving any space.
Click to expand...

Does not need to be the exact USB-PD specs , and you are wrong about the cables , USB-PD specs allows for upto 240 watts ...

as for saving space , well it does if you stack them above each others.
 
InvalidError said:
Only problem with it: where are you getting the 48V from?

Intel's 12VO was a missed opportunity to also bump the main system voltage to at least 24V for even better bulk power distribution efficiency and have a 20V source to run USB3.x/4/TB monitors, external HDDs, USB/TB hubs and other 20-100W external devices.
Click to expand...

maybe then "borrow" from USBC PD design and increase the pins count in the connector.
 
newtechldtech said:
Does not need to be the exact USB-PD specs , and you are wrong about the cables , USB-PD specs allows for upto 240 watts ...
Click to expand...
USB-PD requires that 240W cables have a cable ID chip telling the power supply that the cable is rated for 48V 5A. If you use a cable with only a 5A ID chip, it will be voltage-limited to 20V. A cable with no chip at all is limited to 20V 3A.
 
Amdlova said:
why not put a fuse per pin :) if goes beyond of spec the fuse die and everyone is happy
Click to expand...
Maybe a circuit breaker so you don't have to spend money fixing fuses might work, but you would be better off with a good set of latches on both ends of the connector to keep it from happening in the first place.
 
sitehostplus said:
Maybe a circuit breaker so you don't have to spend money fixing fuses might work, but you would be better off with a good set of latches on both ends of the connector to keep it from happening in the first place.
Click to expand...
Having two loose-fitting latches wouldn't make things meaningfully better. Jayz just put out a video showing how the 90-degrees adapters have more latch slack than his direct cables. If slack bothers you, you should be asking for tighter tolerances to eliminate most latch slack.
 
newtechldtech said:
Does not need to be the exact USB-PD specs , and you are wrong about the cables , USB-PD specs allows for upto 240 watts ...
Click to expand...
https://usb.org/document-library/usb-power-delivery Here's the spec, filename USB_PD_R3_1 V1.8 2023-04.pdf. If you don't want to read the entire thing, look at page 76 where there's a table describing what power delivery is available between the standard power mode and "EPR" which allows for higher power modes (which to correct myself earlier, standard power mode allows up to 100W

However, 240W requires EPR, which if you look at page 186, there's a flow chart in how EPR negotiation works. And if you look on the bottom right, item #6a-d, it checks to see if the cable is an EPR one or not. Also, this negotiation is important. This makes sure the device isn't going to draw more power than the charger can provide.

You need a special cable to use 240W. Also note that even to use anything beyond 60W, the spec says you should use a 5A cable. Considering that it has to specifically call that out, that should tell you something about what kind of cables are available and that you can't plug in just any USB-C cable and expect high power delivery, either at all or safely.

newtechldtech said:
as for saving space , well it does if you stack them above each others.
Click to expand...
Even taking some of my E-marked cables and stacking them together such that they're basically touching, it still doesn't save that much space, if any at all:
Q38fjou.png


newtechldtech said:
maybe then "borrow" from USBC PD design and increase the pins count in the connector.
Click to expand...
That's what they did. 12VHPwr has 6 pins of power delivery in one connector. For a typical 8-pin PEG connector, there's only 3 power delivery lines. So why didn't they go with the same Mini-Fit connector? Who knows? As far as I know, PCI-SIG didn't come forth with an answer.
 
Last edited:
InvalidError said:
Having two loose-fitting latches wouldn't make things meaningfully better. Jayz just put out a video showing how the 90-degrees adapters have more latch slack than his direct cables. If slack bothers you, you should be asking for tighter tolerances to eliminate most latch slack.
Click to expand...
Trust me, you do not want them loose-fitting. You want it snug enough so the connector stays where it's at and does not unplug.

You can use either latches, or a screw together set up.
 
InvalidError said:
Ah, yes, lets make connectors 10X more expensive for no benefits at all as far as typical PC environments are concerned. PCs don't need to be weatherproof, don't really go in high-vibration environments, usually are in temperature-controlled places, etc.
Click to expand...
well you're welcome to stay with the firestarter molex ones. why not find even cheaper ones so you can increase the risk because you're worried about the cost of a connector while powering an RTX4090. if cost is an issue for you, you can stick to cheaper cards like an RTX4060 or older. they don't have the wattage where this is a problem. the 4090 does. why not forego the connector at all and solder the wires to the GPU since connectors are a luxury not needed on a GPU. can't have luxuries in frugal RTX4090 builds. I'd pay extra for every psu to have automotive connectors that can handle the higher amperage, is vibration, high temperature and moisture resistant and is low torque easily aligned and intant click while being rated for continual use.

the newer automotive connectors are just injection molded plastic with rubber seals and better engineered crimped terminals and can be had for $10 for a hundred of them. if you can make the ancient molex style ones, you can make the new ones. will probably be easier too since proper psu manufacturers use name branded molex and amphenol. All modern cars use them for a reason and they can be had in multiple numbers of pins. you can instantly see if they are properly seated as they click. the automotive and aerospace industries stopped using molex for a reason, they cause intermittent connection issues, allow water ingress causing oxidation and have very low amperage ratings, and cannot hold up to heat. the modern connectors can handle engine bay heat in excess of 200 degrees celsius without melting. try that with molex. we did in old cars. it sucked. molex sucks.
 
Last edited:
purpleduggy said:
the automotive and aerospace industries stopped using molex for a reason, they cause intermittent connection issues, allow water ingress causing oxidation and have very low amperage ratings, and cannot hold up to heat. the modern connectors can handle engine bay heat in excess of 200 degrees celsius without melting. try that with molex. we did in old cars. it sucked. molex sucks.
Click to expand...
Sounds more like they stopped using the wrong connector for the job. If your PC is subjected to the same environment as an ECU, then I have some questions to ask.
 
purpleduggy said:
well you're welcome to stay with the firestarter molex ones. why not find even cheaper ones so you can increase the risk because you're worried about the cost of a connector while powering an RTX4090. if cost is an issue for you, you can stick to cheaper cards like an RTX4060 or older.
Click to expand...
The HPWR connector is intended to eventually power everything from 80W to 600W, not only 4090s. The connector does need to be budget-minded.
 
InvalidError said:
The HPWR connector is intended to eventually power everything from 80W to 600W, not only 4090s. The connector does need to be budget-minded.
Click to expand...
the connectors are simply plastic and metal pins. the cost will be cheaper since millions of cars already use them. It will result in a massive increase in reliability and ease of use for the consumer. HPWR clearly isn't doing what its supposed to. there have been thousands of recalls. each PC with an RTX4090 in them is a failure mode waiting to happen. PSU manufacturers all know this. NVIDIA knows this. its ok to remain in denial, but when the recalls affect your brand name, you'll have to face the fact that HPWR is a bad design that melts and in reality cannot handle the wattage required. Automotive connectors that are high temperature would have solved this and would have been cheaper and an off the shelf item already proven and in use in millions of cars. Its time to abandon molex connector designs and leave them in the 1970s and in museums. Modern automotive connectors solve all these issues, they are cheap and proven. HPWR is a custom design used nowhere else, it is an engineering failure:

The Truth About NVIDIA’s RTX 4090 Adapters: Testing, X-Ray, & 12VHPWR Failures
View: https://www.youtube.com/watch?v=ig2px7ofKhQ
 
Last edited:
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom