GeForce RTX 5090 With Four 16-Pin Power Connectors Spotted In The Wild
NVIDIA's GeForce RTX 5090 remains a fascinating GPU, and continues to interest us as we learn more about it. Case in point, a prototype design that is packing four power connectors recently came into view, and it looks like something engineers would have a fun time with.
Shared by Yuuki_ans on X, this damaged GeForce RTX 5090 was never meant to be inside a gaming PC. Rather, it likely served as a test unit to perhaps try different approaches to what the GPU could theoretically do when it has much more power. The four 16-pin connectors are abundant and would allow such testing. For reference, the standard GeForce RTX 5090 has a single 16-pin connector, and a 575-watt rating. Each connector can support 600 watts, so in theory this prototype GPU could draw up to a staggering 2,400W, but it would likely stress out other components in the chain with the massive thermal output.
This bizarre GPU could also be meant as something very different than the RTX 5090, including a number of AI or data center GPUs that NVIDIA may have in mind. The Blackwell architecture for NVIDIA's RTX 50 series GPUs do have similarities between the Geforce gaming stack and more high-powered Pro models for workstations and data centers. While the RTX 5090 has a healthy 32GB of GDDR7 VRAM, the RTX Prp 6000 has 96GB of GDDR7.
Shared by Yuuki_ans on X, this damaged GeForce RTX 5090 was never meant to be inside a gaming PC. Rather, it likely served as a test unit to perhaps try different approaches to what the GPU could theoretically do when it has much more power. The four 16-pin connectors are abundant and would allow such testing. For reference, the standard GeForce RTX 5090 has a single 16-pin connector, and a 575-watt rating. Each connector can support 600 watts, so in theory this prototype GPU could draw up to a staggering 2,400W, but it would likely stress out other components in the chain with the massive thermal output.
This bizarre GPU could also be meant as something very different than the RTX 5090, including a number of AI or data center GPUs that NVIDIA may have in mind. The Blackwell architecture for NVIDIA's RTX 50 series GPUs do have similarities between the Geforce gaming stack and more high-powered Pro models for workstations and data centers. While the RTX 5090 has a healthy 32GB of GDDR7 VRAM, the RTX Prp 6000 has 96GB of GDDR7.
Not all RTX 5090s are the same, even amongst released models in the wild. The ASUS ROG Astral RTX 5090 LC for example foregoes traditional air cooling and uses an attached radiator to function at higher tolerances. While GPUs such as the ROG Astral can often set records, it is still far from what a potential prototype can do when the power is truly unleashed.
In the GeForce space, we also have not seen a Titan-level GPU from NVIDIA in a long time. While the xx90 series products have essentially replaced the Titan lineup, it is still possible this prototype GPU could be some form of test for a higher-tier graphics solution. NVIDIA is very profitable with its data center cards, so unless there is a good reason to mix both markets, it likely prefers to focus on data centers.
While the GPU in question here is neat to see with four 16-pin power connectors, it seems unlikely to be anything more than a novelty at this point.
In the GeForce space, we also have not seen a Titan-level GPU from NVIDIA in a long time. While the xx90 series products have essentially replaced the Titan lineup, it is still possible this prototype GPU could be some form of test for a higher-tier graphics solution. NVIDIA is very profitable with its data center cards, so unless there is a good reason to mix both markets, it likely prefers to focus on data centers.
While the GPU in question here is neat to see with four 16-pin power connectors, it seems unlikely to be anything more than a novelty at this point.
