Leaked Specifications for NVIDIA GeForce RTX 5090 and RTX 5080

For several months, the tech world has been abuzz with leaks, rumors, and reports surrounding Nvidia’s highly anticipated flagship GPUs, the RTX 5090 and RTX 5080. Now, thanks to a significant and seemingly definitive leak, we’ve gotten a glimpse of what these powerful new graphics cards will offer.

A well-known and reliable Nvidia leaker, Kopite7kimi, recently took to their X (formerly Twitter) account to share some key specs for Nvidia’s upcoming flagship models. These insights provide a deeper look into the RTX 5090 and RTX 5080, setting expectations high for the next generation of GPUs.

RTX 5090: Powerhouse Specs Revealed

According to Kopite7kimi, the RTX 5090 will feature a 412-bit memory bus, a revision from earlier reports that had suggested a 448-bit configuration. The card will also boast 32GB of GDDR7 memory, making it a top-tier option for memory-intensive tasks such as high-resolution gaming and complex computational workloads.

At the heart of the new flagship is the GB202-300 GPU, which comes with 21,760 FP32 CUDA cores—a slight reduction from the 24,576 cores of the full chip. However, this reduction is offset by the card’s impressive memory configuration, featuring a 512-bit memory bus and the aforementioned 32GB of GDDR7 RAM. This architecture promises blistering speeds and unparalleled performance in demanding applications.

One of the more eye-catching aspects of the RTX 5090 is its 600W power consumption, a significant jump from the previous generation’s 450W. While this increase points to greater performance potential, Kopite7kimi did not specify whether this figure refers to the Total Board Power (TBP) or the Total Graphics Power (TGP), leaving some uncertainty about the exact power metrics.

Despite this substantial power increase, the card is expected to maintain a 2-slot cooler design. This hints at the possibility of a non-standard or highly advanced cooling solution from Nvidia, allowing the card to stay efficient and compact even with such high power demands.

RTX 5080: A More Moderate Yet Powerful Alternative

In addition to the RTX 5090, Kopite7kimi also shared insights into the specs for the RTX 5080. This card will feature a GB203-400 GPU with 10,752 CUDA cores, nearly half the number found in the GB202 of the 5090. However, the RTX 5080 will still offer solid performance, especially for users who don’t require the full power of the 5090.

The RTX 5080 is expected to come with 16GB of GDDR7 memory and a 256-bit memory bus, which should provide a smooth experience for gaming, video editing, and other GPU-intensive tasks. The card will reportedly draw 400W of power, up from the 320W of the previous RTX 4080 model, suggesting improved performance while maintaining efficiency.

Unfortunately, there is still no official word on the release date for Nvidia’s Blackwell graphics cards, including the RTX 5090 and RTX 5080. While the leaked specs provide plenty to be excited about, both Nvidia and leakers have remained tight-lipped about when we might see these GPUs hit the market

As with any leaks, it’s important to take the information with a grain of caution. However, Kopite7kimi has proven to be a reliable source in the past, so these specs are likely to be close to the final product.

Until Nvidia makes an official announcement, enthusiasts will have to keep their eyes peeled for more updates on what is shaping up to be a groundbreaking generation of graphics cards.