• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

NVIDIA RTX 5090 "Blackwell" Founders Edition to Implement the "RTX 4090 Ti" Cinderblock Design

I'm not a huge fan of it but this GPU definitely needs one.
 
At first I thought the fan in the middle was a retractable handle...:kookoo:

1716546240096.jpeg
 
At this rate the biggest tech improvement for Nvidia generations of cards is how they managed to improve the cooling this time to keep pace with the elevated TDPs.

Interesting indeed, these self-imposed problems they fix.
 
Kinda reminds me of this :D
 

Attachments

  • Dad's GPU.jpg
    Dad's GPU.jpg
    87.7 KB · Views: 75
"RTX 5090, All of your slots are belong to us." It just markets itself.
 
LOL @ those people with those tiny little SFF builds.
eGPU cases will have to get bigger for them.

I can't wait to read a review and see how crazy the power consumption is.
 
Finally some innovation in this field.
 
Those poor PCI express slots....
 
With the size of these things getting larger each generation, they'll have to go external at this rate :D
 
The article mentions 3 separate PCBs - main PCB parallel with motherboard (goodbye to using any other slots!), output PCB with HDMI & DP sockets, Input PCB with PCIe slot. 12V Hpwr socket elsewhere on the heatsink.
Ya, that was the idea for the 4090Ti, i was perplexed because it's a clear deviation from the majority of PCIE addin cards in the market today. It's more complex to manufacture, pricier too. But we are speaking about nVidia..... so boh? Apparently the added cooling from having a complete passtrought airflow outweight the added complexity.
 
With the size of these things getting larger each generation, they'll have to go external at this rate :D
Since the introduction of a dual slot cards more than a decade ago, the standardized PCIe layout should've been changed, maybe the whole motherboard layout as well. But it didn't, so now we have this...
GPU was a lower power part blowing hot air to the higher power part, now it's vice versa.
 
True, ATX was never designed for anything like what we have now with so many big, hot and massive power draw components. It's about time it was redesigned from the ground up, it would be a nightmare but it really needs to be done. It will have to be done eventually.
 
A pass-through design with some X-shaped trim? Where have I seen this before? Oh yes! On every other Nvidia card in the last 5 years.

So where's the "leak"? :laugh:
See below:
The article mentions 3 separate PCBs - main PCB parallel with motherboard (goodbye to using any other slots!), output PCB with HDMI & DP sockets, Input PCB with PCIe slot. 12V Hpwr socket elsewhere on the heatsink.

Ya, that was the idea for the 4090Ti, i was perplexed because it's a clear deviation from the majority of PCIE addin cards in the market today. It's more complex to manufacture, pricier too. But we are speaking about nVidia..... so boh? Apparently the added cooling from having a complete passtrought airflow outweight the added complexity.

Yes, that was the previous leak. I agree this is very complex to manufacture compared to previous designs. If they solder the boards together, that's a whole set of fixturing and operations that add labor and time. If they're connectorized, they need to make sure all the tolerances on every mechanical part are really good or misalignment can disconnect things on people's ($2.5-3k?) cards during shipping or install. They're also relying on some heat-shrink on stiff copper bus-bars to run from the connector down to the board. There are just so many failure points that are carrying a pretty hefty amount of current (not just the bus bars). I'm also curious if the reference design is the same as the FE. Is this modular approach what they're going to push AIBs into, or will they be able to put everything on one board like usual?

The only thing that looks super robust here is that the 4-slot bracket is bolted thoroughly to the heat-sink frame, which will secure it in place pretty well. That said, I've seen a fair few cases out there with some pretty flimsy I/O panels after all the Swiss cheese ventilation they've added, so the weight might still be enough to pull on the whole panel lol.
 
There havn't been any fires, and no more further melting reports since people stopped using CableMod Adapters.

I think it was GamersNexus or HWUB that just did a report on it recently. A single repair shop that contracts with the board partners is still doing 200/month.

Mind you typically issues don't get much exposure beyond the first initial coverage. That's just how the media works in general. Some issues don't get picked up at all, like the fact that the 3080 and 3090 both feed noise back into the 12V Sense pin which causes issues on PSUs with sensitive protections.
 

Just because it's not a widely reported issue anymore doesn't mean it still isn't a problem, but then again, if you ask Nvidia they initially said it was user error.
I wonder if some PSU's just arn't up to the task, i consantly see people reporting 11.7 volts and lower.

My Seasonic PSU always sits around 12.1 - 12.2 volts for the 12VHPWR connector.

It could also be a shitty powergrid ofcourse.

That said the 12VHPWR cable is pretty much maxed out spec wise when running a 4090, so two connectors would fix it.
 
A GPU with separate boards for components is very interesting. This GPU seems like Nvidia is purposely limiting after market setups for cooling at least in the DIY space or at least making it much more difficult (I am sure there will be water blocks for it). I would be more curious if its possible to even save much space doing a water block on the design or if its purely for cooling power.

I feel like this card is going to make the 3090ti's power consumption look like child's play if this is how they are going about cooling it.
 
Since the introduction of a dual slot cards more than a decade ago, the standardized PCIe layout should've been changed, maybe the whole motherboard layout as well. But it didn't, so now we have this...
GPU was a lower power part blowing hot air to the higher power part, now it's vice versa.
I really do think they need to revise the specs if this is the trend. Perhaps redesign the board so that one x16 slot is above the CPU area, which could free up vertical space above the motherboard. Then it would just be a matter of the case supporting the dimensions of the card. They could even compartmentalize the GPU and CPU areas, allowing for independent airflow and cooling, better GPU and CPU cooler support, and better routing of power cables. They might even be able to do this without having to increase the overall case dimensions if they get creative with PSU placement.
 
Its just rated for 600W of (air)cooling. A big base with strong fans.
 
Do you still have the strength of a 8yr old virgin boy? That's always a nice addition to blaming nvidia for no fault of their own.
:cry:
Seems I hurt your personal feelings....

giphy.gif
 
Back
Top