RTX 4xxx / RX 7xxx speculation

900W heatsink.

https://videocardz.com/newz/massive...sed-900w-geforce-rtx-40-gpu-has-been-pictured

NVIDIA-RTX-40-900W-COOLER-2-768x557.jpg
 
Axman said:
That must be for the rumored cancelled Ti.
Click to expand...
Probably more the rumored cancelled Titan from my understanding no talk of cancelled TI, there so much room on that 4090 chips (only 16384 of the 18176 core are used) and so much room on the power-cooling on the already shipped model that they just have to unlock them to have something they call a 4090TI without changing anything else, the upcoming 24 gddr6x being a nice bonus over that.

The cancelled Titan would have been a 48 GB version like previous Titan way out of touch with gaming requirement.

And 11-1% more core and ram bandwidth 4090 TI released before the end of 2024 depending on what RDNA 3 top model do seem written on the wall.
 
LukeTbk said:
Probably more the rumored cancelled Titan from my understanding no talk of cancelled TI, there so much room on that 4090 chips (only 16384 of the 18176 core are used) and so much room on the power-cooling on the already shipped model that they just have to unlock them to have something they call a 4090TI without changing anything else, the upcoming 24 gddr6x being a nice bonus over that.

The cancelled Titan would have been a 48 GB version like previous Titan way out of touch with gaming requirement.

And 11-1% more core and ram bandwidth 4090 TI released before the end of 2024 depending on what RDNA 3 top model do seem written on the wall.
Click to expand...

I never thought the 10-12% increase in performance was worth waiting for Ti. Rather enjoy the boost now then wait a year. Time is everything.
 
To be fair, there shouldn't be any issues with the 12vhpwr connector itself. It's just that stupid cable Nvidia supplies with their cards. Although that means everyone would have to buy custom PSU cables just to avoid a fire hazard which is an awful look for Nvidia.
 
cvinh said:
To be fair, there shouldn't be any issues with the 12vhpwr connector itself. It's just that stupid cable Nvidia supplies with their cards. Although that means everyone would have to buy custom PSU cables just to avoid a fire hazard which is an awful look for Nvidia.
Click to expand...
We don't yet know if this is a widespread problem or a few people on the internet. If it's anything like the Turing memory problems it will be well reported I'm sure.
 
cvinh said:
To be fair, there shouldn't be any issues with the 12vhpwr connector itself. It's just that stupid cable Nvidia supplies with their cards. Although that means everyone would have to buy custom PSU cables just to avoid a fire hazard which is an awful look for Nvidia.
Click to expand...
The connector is actually the problem. The wires pull on the connector terminals causing those to not seat "perfectly." I guess if we lived in a world where no one that builds a PC has to do cable routing...

1666707790855.png
 
FrgMstr said:
The connector is actually the problem. The wires pull on the connector terminals causing those to not seat "perfectly." I guess if we lived in a world where no one that builds a PC has to do cable routing...

View attachment 521243
Click to expand...

So the issue would still exist even when using an ATX 3.0 PSU? Would it be fixed using the 90 degree adapter cablemod is designing?
 
Last edited:
Gideon said:
Those pins are just way to small for the amperage they are being asked to carry. A engineering computer might say they are big enough, but the real world disagrees. I have a feeling Nvidia did not rigorously test these connecters in real world conditions.
Click to expand...
I haven't seen one yet, but are they actual pins smaller? The pins appear to be the same size, but they're closer together than the old style, which means that there's less plastic structure to keep them properly aligned.

You're clearly right, though, that they didn't test this well enough under real world conditions.
 
RazorWind said:
I haven't seen one yet, but are they actual pins smaller? The pins appear to be the same size, but they're closer together than the old style, which means that there's less plastic structure to keep them properly aligned.

You're clearly right, though, that they didn't test this well enough under real world conditions.
Click to expand...

I have to go by pictures I have seen, have not seen one in person, but they look similar to what we used in the automotive trade and we don't push more then 15 amps on those size pins. They tend to expand with heat and thus contact starts to degrade which causes overheating due to increasing resistance.
 
Gideon said:
I have to go by pictures I have seen, have not seen one in person, but they look similar to what we used in the automotive trade and we don't push more then 15 amps on those size pins. They tend to expand with heat and thus contact starts to degrade which causes overheating due to increasing resistance.
Click to expand...
55.2A spread across 6 power pins means each one is carrying around 9.2A. Nowhere close to 15A.
 
Guess I'm waiting for RDNA 3.0 Although there's only a few cases of melted connectors, that's a few too many for my liking. Hopefully AMD can offer matching performance at competitive pricing.
 
Gideon said:
Your assuming even draw across the pins, they likely are not setup that way.
Click to expand...
The only way to get the exact specifications is to be a member of PCI-SIG. I would say it's safe to assume it is considering it was a consistent 4.5A across the 3 power pins on the previous cable spec.
 
So the question is who is on the hook if it leads to your house burning down? Nvidia, AiB, PSU maker or PCI-SIG. Who is responsible to replace the entire PC at least? It is going to be a huge legal battle if this blows up.
 
FrgMstr said:
Well, there you go.
Click to expand...

I just saw this tweet on a youtube channel and guessed it had to be you. I'm leaning heavily towards the 7000 series now, the 4090 availability doesn't look like it will improve with Nvidia changing production over.
 
cvinh said:
So the issue would still exist even when using an ATX 3.0 PSU?
Click to expand...

On an ATX 3.0 PSU you would probably end up with the connector on the end of the modular 12VHPWR cable that plugs into your PSU burning up also.
 
  • Like
Reactions: cvinh
like this
Did Nvidia "invent" this connector and the PCIe standards group adopt it? Or the other way around?

But to me, this seems like an issue of sloppy tolerances and manufacturing. The smaller pins and plastic warrant a plastic that is more dimensionality stable at higher temperatures and a beefier and robust female pin recptiacle
 
Lol i got shit on for posting a video about the power connector a while ago around launch time, sorry it's happening but I do feel vindicated. Did they just not want to have 4 8 pins on the board? I guess that would get kind of absurd, this on the other hand is far worse.
 
GotNoRice said:
On an ATX 3.0 PSU you would probably end up with the connector on the end of the modular 12VHPWR cable that plugs into your PSU burning up also.
Click to expand...
Pure speculation or informed opinion ?

If I understand this correctly:
https://wccftech.com/atx-3-0-12vhpw...r-safety-risk-using-adapter-confirms-pci-sig/

They saw unbalanced load on adapter being used that maybe you do see otherwise.

the cable of Nvidia is supposed to help regulate the draw by those 4 cable but not sure we can assume it work has well has direct GPU-PSU no ?
 
LukeTbk said:
Pure speculation or informed opinion ?

If I understand this correctly:
https://wccftech.com/atx-3-0-12vhpw...r-safety-risk-using-adapter-confirms-pci-sig/

They saw unbalanced load on adapter being used that maybe you do see otherwise.

the cable of Nvidia is supposed to help regulate the draw by those 4 cable but not sure we can assume it work has well has direct GPU-PSU no ?
Click to expand...
The PCI-SIG report is based upon Nvidias own findings on the PSU side. Problem exist on both sides of the connector, since its the risk of sideloading thats reported as the major problem as shown with FrgMstrs post and earlier post in this thread. There has been a lot of discussions about the connection cycles, but the main issue have always been bending. If a pin is not straight inside the connector, it increases resistance on that pin which again produces heat.

Perhaps if they reinforced the connector a bit stronger beyond the tape and shrink tube approach, it might avoid the issue?

On the PSU side, perhaps they should make it non-modular, since space is often limited for PSU connectors in many cabinets and bending is a must for many?
 
Tamlin_WSGF said:
The PCI-SIG report is based upon Nvidias own findings on the PSU side. Problem exist on both sides of the connector, since its the risk of sideloading thats reported as the major problem as shown with FrgMstrs post and earlier post in this thread. There has been a lot of discussions about the connection cycles, but the main issue have always been bending. If a pin is not straight inside the connector, it increases resistance on that pin which again produces heat.

Perhaps if they reinforced the connector a bit stronger beyond the tape and shrink tube approach, it might avoid the issue?

On the PSU side, perhaps they should make it non-modular, since space is often limited for PSU connectors in many cabinets and bending is a must for many?
Click to expand...

Easy solution: 90 degree connectors (or 180 degree -- cable can't torque the connector if it's lying on the back of the card).

Less-easy solution: maybe the time for ATX24vo (or ATX48vo) has come. Like, 20% of motherboards and video cards (by area) is power regulation as it is. You can build a system with the bare-minimum component part count and have it pushing nearly a kilowatt at full-tilt with consumer-grade parts. Pushing ~10 amps per pin through glorified Molex connectors ain't gonna cut it. Bite the bullet and make things less stupid.
 
Tamlin_WSGF said:
. If a pin is not straight inside the connector, it increases resistance on that pin which again produces heat.
Click to expand...
From my understanding, when it increase resistance it reduce current instead on that pin and make more current goes on the lower resistance pin which would be the one overheating and thus my question if the ATX 3.0 rating PSU would accept that load imbalance, deliver high level unbalanced amperage between them or prefer to refuse to deliver the power instead detecting the issue, the answer to this can be no obviously and if people know that it is the case they know and can say it.

Having a 90 degree elbow adapter to avoid bending, could have been a good solution cost wise ?
 
Which way do you make the 90 degree bend? If you route your wires from the opposite side, then you have an even larger bend on the cables.
 
The problem really is just that everyone has cheaped out and combined the 3 8pin connector leads terminal with the 12pin connector. There should be the 3 8pin leads, a terminal block, then a chunky lead, then the final 12pin connector.
 
Wat said:
Did Nvidia "invent" this connector and the PCIe standards group adopt it? Or the other way around?

But to me, this seems like an issue of sloppy tolerances and manufacturing. The smaller pins and plastic warrant a plastic that is more dimensionality stable at higher temperatures and a beefier and robust female pin recptiacle
Click to expand...
The other way around. The 12-pin that NVIDIA introduced with the Ampere generation was a forward-compatible interim solution until ATX 3.0 spec products were available. The PCI-E 5.0 16-pin power specification was available back in 2019.
 
  • Like
Reactions: Wat
like this
You must log in or register to reply here.
Back
Top