[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: file.jpg (128 KB, 1360x752)
128 KB JPG
>solves your power connector issue
>>
>>108625441
I think it could work
>>
>>108625441
watercooled connector ?
>>
>>108625444
because you are a typical /g/ retard crossdresser. if the connector is going to melt, it is going to melt. throwing water on the wire crimps is not going to help.
>>
>>108625563
it's not like they instantly achieve 900c, they melt because they get too hot

if you had a copper jacket with good contact it legitimately would prevent it from getting to that runaway condition
>>
>>108625441
This doesn't even remotely address the issue
It just needs simple load balancing of the 6 wires
>>
>>108625441
issues?
>>
>>108625460
copper heatsink basically.
but how do you unplug it as it has a plastic latch on the connector?
>>
What has the world become?
>>
Why not simply use PoE to power the device?
>>
File: CM-CON-8PCI-R-1.jpg (49 KB, 1080x1080)
49 KB JPG
>>108625441
My power connector issue? where?
>>
>>108625685
better yet, transition to using 48V for graphics cards
>>
>>108625441
The technologically illiterate will buy this before undervolting their flagship cards.
>>
>>108625441
There are real solutions to these power ratings that exist in other electronics you know
>>
>>108625441
>make a new high-power connector
>it's not good and melts sometimes
>fix the issues
>the issues are fixed
>huehue retards will shit on your connector for 9000 more years as if the issues are still present
it's all so tiresome, I wish they would sue people for this blatant slander of a perfectly fine connector
>>
>>108625563
what does it have to do with crossdressers even
>>
isn't mixing electricity and water like a bad thing or something?
>>
>>108625907
>>the issues are not fixe
>> blame anything else
>> "You're plugging it wrong"
Why are you defending this?
>>
>>108626041
not if it's distiled water
>>
>>108625441
>powered connector with copper casing
I wonder why this doesn't exist anywhere. I'm sure there's a reason...
>>
>>108625907
>the issues are fixed
no, they're not
it's still shit
even when it's powering cards that only consume half of its rated wattage it still fucking melts
the cable is innately unsafe even when you give it a 300w overhead with the 5080
>>
>>108626176
>>108626208
Why are you slandering a perfectly fine connector? I run local ai on 2 rtx pro 6000 24/7 with this connector at 400-600W continuously for 5 months already and nothing ever melted. My gayming 5090 also never had any issues.
>>
>>108626235
>perfectly fine connector
it's objectively a downgrade to 8 pins and you have to be a clueless retard to pretend otherwise
its sole purpose is to look nicer than having 4 8 pins sticking out of the card, the most faggoty of reasons
>>
>>108626245
Its purpose is to deliver 600w of power over 12 wires in 1 connector. You can do it with 16 wires in 2 connectors, but there's no reason to do that when 1 connector and 12 wires are enough.
>>
>>108626270
no, its purpose is to make faggot founder's edition cards look better by letting them easily tuck the connectors away
it's not more robust, it doesn't do anything better, it's just more space efficient at the cost of everything else
>>
>>108625956
you need to go back
>>
>>108626270
The safety margin is so low that if there is even slightly bad connection due to manufacturing tolerances or some other issue it's going to catch fire. It would help if the individual wires had current sensing and balancing like the old 8 pin connectors do, but they don't even have that. Maybe they do have balancing and current sensing on something as expensive as a RTX PRO 6000, it would be literally criminal if they were counting nickles on something like that.
>>
>>108626292
at the cost of literally nothing because there's nothing wrong with 12V-2x6
>>108626318
>if there is even slightly bad connection due to manufacturing tolerances or some other issue it's going to catch fire
that was an issue with 12VHPWR, 12V-2x6 has longer pins
>current sensing and balancing like the old 8 pin connectors
current sensing is not the cable's job, ASUS cards have per-wire sensing as a premium feature
>Maybe they do have balancing and current sensing on something as expensive as a RTX PRO 6000
of course not, workstations are expected to be assembled correctly, if you cut one of the power wires and the rest of them melt - it's your problem
>>
>>108626346
>at the cost of literally nothing because there's nothing wrong with 12V-2x6
ok, you're just a fucking shill at this point, go fuck yourself
>>
>>108626346
>current sensing is not the cable's job, ASUS cards have per-wire sensing as a premium feature
That's what I'm talking about, retard. The cards don't have current sensing per wire when using the new connector. All cards used to sense current per wire on the 8 pin.
>>
>>108626288
Yes it does.
A greater differential in temperature leads to faster transfer of said heat.
It's all about heat build up and heat dissipation.
>>
>>108625460
yea so it doesn't get too hot :)
>>
File: file.png (194 KB, 615x268)
194 KB PNG
I have designed a much thicker beefier setup that will prevent this issue.
>>
How did nvidia spend so much money on AI slop that they couldn't hire an electrician to tell them those wires are too small for the power going through them?
>>
File: 1771280732154.jpg (11 KB, 210x295)
11 KB JPG
>>108625441
Is this what they mean by steam machine?
>>
>>108625907
>Fix involves overengineered workaround instead of addressing the core issue
Sometimes I'm starting to consider eugenics seriously...
>>
>>108625441
idgi what is this? how would you screw in the 2 circular connectors on the side if you're plugging the main connector straight in
>>
>>108625441
> enginerding around a shit design
no thank you
>>
>>108626883
they arent too small you just need to not be an idiot when you plug it in make sure its properly seated. and thats what nobody did.
>>
>>108629589
>Just work around the problem Nvidia created
Begone shill jeet
>>
>>108625441
It's as if a car manufacturer released a car that has a tendency to go up in flames because of a trivial design flaw that causes excess heat by a fuel line caused by the designers wanting it to look "neat", and instead of revamping the design to either make it beefier or to move it away from the source of heat, they slap on an auto-extinguishing unit by that fuel line and keep telling everyone that the design has no flaws.
>>
>>108625649
Unplug? whaddya mean? You just buy another gpu and psu
>>
>>108625907
Why is Nvidia's warranty usage going up 10x in the last year vs AMD's going up like 20% then?
>>
>>108629589
>need to take a scholarship to plug your gpu or you'll burn in a firehouse
Yeah, what a bunch of retards who can't plug this gpu.
>>
File: 1760159406090175.jpg (117 KB, 750x1000)
117 KB JPG
>product breaks because of product defect
>entitled to free repair or replacement
EU chads we can't stop winning
>>
>>108625563
active cooling the highest resistance point of the connection is... le wrong?
>>
File: 1745861911088567.jpg (119 KB, 576x1024)
119 KB JPG
>>108631219
yes, you retarded transvestite. The water jacket does not cover "le" point of highest resistance. The steel sleve (i.e. the part that melts) is outside of the water jacket. Only the backside of the pin sleeve (where the steel crimps to the copper wire) is covered by the waterblock.

But even IF the entire pin assembly were doused in water, it would only SLIGHTLY delay a thermal runaway that would have otherwise resulted in total failure
>>
>>108625606
its honestly baffling they fucked it up that badly, and 2 generations as well. the 3090 worked correctly with the new connector
>>
>>108625907

My 5080 is being fixed as we speak because of a connector issue. Fuck Njudea for cheaping out on this shit. I know Ayymd is inferior in the vast majority of cases, but this soured my take on Nvigger cards.
>>
>>108625441
Use thicker wires
>>
https://youtu.be/mL-CxvhJuRo
>>
>>108632965
I was just joking I don't think 600W gpus should even be a thing, calm your pcbg tranny intrusive thoughts please
>it would only SLIGHTLY delay a thermal runaway
Really? Running a 1/8 inch tube of coolant through a copper block can dissipate maybe 200 watts, the losses in those connectors probably aren't even approaching numbers like that
>>
>>108625441
>Instead of balancing the load throughout all the cables, let's just add liquid cooling to a power connector
The shit around this connector is just mindboggling.
>>
>>108633400
>balancing the load
Physics does that for you.
>>
>>108633431
Yeah, it makes the electricity take the shortest route so you have 500W running though a single wire while the others barely have any current running through them, truly mindboggling this passed any QA checks whatsoever, mega corpos can't beat the allegations of testing in production.
>>
>>108625441
stop being anti-sematic! >:(
>>
>>108633471
The current is split based on resistance. All the wires will have the same current if they're connected properly.
>>
>>108633471
>electricity take the shortest route
a whole decade of compulsory education... for this?
>>
>>108633500
>>
>>108633514
>>108633531
It's been proven already that the issue with the connector is that all current passes by a single wire in this garbage, because all wires land into a single trace with no balacing as per spec, so the electricity just takes the shortest route through a single cable until it gets so hot it catches on fire.
>>
File: 1760983431862097m.jpg (246 KB, 1024x992)
246 KB JPG
>>108633550
holy kek what a fucking retard
>>108633531
California spends $26k/kid/yr for this
>>108633374
>losses
ok you can just STFU now
>>
Bigger, thicker wires.
We already manufacture those.
They are available, there are proper connectors.
>>
File: P7824_1[1].jpg (28 KB, 740x350)
28 KB JPG
>>108625441
high current DC delivery is a solved problem. this connector will handle 60A which is significantly more than even a 5090 needs, not to mention it's easier to plug and unplug. these stupid fucktards have to make everything more difficult than it needs to be.
>>
>>108633825
This look cute
>>
>>108633825
okay but what about the other pins
>>
>>108625685
>48W
are you just pretending?

>>108633825
There are even bigger Anderson connectors for that. Don't forget the lack of proper cooling of the power circuitry and the ambient temperature in excess of the perfect 23'C, even this piece will melt under long 600W power draw runs.
100Amps is no fucking joke, either up the voltage or lower the power cuntsumption, you chink jacket.
>>
>>108633825
fuck RC crap
I want 40A d-sub instead
>>
File: rmdl9j4smnpg1.png (844 KB, 1160x726)
844 KB PNG
>>108626346
> there's nothing wrong with 12V-2x6
hey faggot, you smell your PC every day to see if the plastic is melting? lmao
>>
>>108635252
you dont need other pins, just + and - is enough. as if you don't have enough regulators on that big ass card.
>>
>>108635782
You need multiple power rails so they are galvanically separated, otherwise you'd get stray inductance everywhere and graphics cards would be impossible.
Did you really think they made the connector a bunch of smaller pins instead of two big ones for shits and giggles? Jesus you goymer niggers are completely oblivious
>>
>>108636096
>galvanically separated
They're all shorted together inside the cable
>>
>>108636112
I don't even need to test it to know that this is horseshit. Even in the op pic you can very clearly see that individual strands run in separate insulation
>>
>>108636205
>>
>>108636205
They seem to wire them in pairs.
>>
At this point, the gpu should just be an external device (independently powered, better cooled due to not needing to fit the traditional "pcie board" design") that connects to a dedicated port on your motherboard.
>>
>>108633029
>the 3090 worked correctly with the new connector
The 3090 didn't draw enough power that it could overload due to faulty wiring. It went beyond spec, but there's still safety margin beyond spec.
Also- iirc the 30xx series still had card-side balancing of draw. Nvidia cut that out of the board on the 40xx series and up, to save space and not have to redo their entire board topology. (Aka basic bone-headed cost-cutting.)
>>
File: images(6).jpg (25 KB, 452x678)
25 KB JPG
>>108636096
1.21 jiggerwatts?!
>>
>>108635508
>Chip mommy
would
>>
>>108625441
Or like... Idk... use 24V instead of 12? Only half of the Amps needed to deliver the same power.
>>
>>108636096
>You need multiple power rails so they are galvanically separated
So how does it go from two wires in my wall to 8 in my GPU? Oh, power converters?
The multiple wires are there to distribute load so you don't need huge gauge wires. Not for galvanic separation.
>>
>>108636269
This sounds like an interesting idea. How about a dedicated graphics server. It just does graphics calculations and sends the pixel information derectly to your monitor.
>>
We can call it "computer." So you'll basically have a compuer connected to your PC just for graphics.
>>
>>108636497
What does "PC" stand for then in your case?
>>
>>108625858
It wouldn't surprise me if they arent doing this cause of the "style"
>>
>>108636507
Yeah, no I get it. I mean I guess you could call it GCompuer (graphics computer) but thats kind of lame. It seems kinda wrong to come up with an alternate name for the "graphics computer" when in fact it's the one doing the heavy lifting, IT's the one doing the real cumputing. So i'm thinking let's call it computeR. We capitalize the the final letter to distinguish it from the main PC without coming up with any lame names.
>>
To verbally distinguish between the computer and the computeR we can pronounce computeR "compute-ahr."
>>
>>108636515
These look rad as fuck.
>>
>>108636452
There's a transformer between your gpu and a wall outlet, you can wind multiple secondaries onto a single core you know.
Why is /v/ allowed to relentlessly shit up this board and I get IP wipes for light shitposting?
>>
>>108636112
>>108636205
They are not supposed to be joined at the cable but the cables are all joined together on the pcb side anyway on all but the most ultra premium cards.
So the problem of uneven load distribution remains. A single contact point that either works or not works at all would be safer.
>>
>>108637096
All the good psus are single rail.
>>
File: IMG_8838.jpg (189 KB, 568x772)
189 KB JPG
Why not just make them out of a material that doesn't melt?
>>
>>108633029
12V-2x6 fixes the problem. You have to intentionally sabotage it for not to work correctly. The move to 12VHPWR was completely unnecessary and was done purely for marketing purposes. Nvidia coping Apple's Lightning: "one small sexy cable to do it all!"
>>
>>108625441
How does this solve anything? Connector melts right at the end when it enters GPU, there's no point cooling irrelevant plastic part.
And the reason it melts is that it was poorly engineered, they put wrong tolerances, so connectors might melt even if it is manufactured within tolerances.
>>
>>108637734
the pins have heat pipes running to them, so it cools the entire pin structure
>>
>>108625441
I really love how by trying to make the connector smaller, its still winding up bigger with extra sense pins and now thermal monitoring

it really is a humiliation ritual because its not even a money savings....
>>
>>108639529
>because its not even a money saving
are you retarded? the 8 pin plug had two unused full size wires, the sense pin on the 12 pin are tiny



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.