Nvidia GeForce RTX 5080 guide: Everything we know about the new gaming GPU

Here are all the latest rumors about the RTX 5080 specs and price, as well as our estimate for the release date of this graphics card.

Leaks are hotting up about Nvidia’s new second-in-command, with the Nvidia GeForce RTX 5080 specs and performance rumored to offer several improvements over the current RTX 4080, but without rocking the boat too much. After the catastrophic launch of the GeForce RTX 4080, where one card ended up being “unlaunched” and the final card had an absurdly high price, we’re hoping that Nvidia has learned some lessons for this GPU’s successor.

While the RTX 5090 looks set to be the Nvidia flagship in its new Blackwell gaming GPU lineup, where it’s likely to take the best graphics card medal, the RTX 5080 should be much cheaper. Plus, while Nvidia misjudged the pricing of the original RTX 4080, it (just about) returned to form when it launched the cheaper and faster RTX 4080 Super for $999 at the start of 2024.

Nvidia GeForce RTX 5080 specs estimate

The latest Nvidia GeForce RTX 5080 specifications rumors point to the new GPU having around half the power of the RTX 5090, with just 10,752 CUDA cores and 16GB of GDDR7 VRAM. 

The Nvidia GeForce RTX 5080 will be built on a new architecture, using one of the company’s new Blackwell RTX 50 GPU designs. While the RTX 5090 is rumored to use a large new chip, codenamed GB202, the latest RTX 5080 leaks say that this new graphics card will use a smaller chip, codenamed GB203.

It’s likely that this chip will also be used in other GPUs further down the Nvidia stack, such as the RTX 5070 Ti, with some parts of the GPU disabled, but the RTX 5080 is rumored to use a variant codenamed GB203-400-A1.

According to the latest rumor from tech leaker kopite7kimi, the RTX 5080 CUDA cores will stand at 10,750, spread across 84 of Nvidia’s Streaming Multiprocessor (SM) units. This setup would also give you 84 RT cores, assuming Nvidia sticks to a similar block design to its current Ada Lovelace GPU, where each SM has an RT core.

Those are similar numbers to the 10,240 CUDA cores and 80 SMs found in the current RTX 4080 Super, but they’re dramatically lower than the rumored figures for the RTX 5090. The flagship GPU is rumored to have 21,760 CUDA cores, which is more than double the 10,750 rumored for the RTX 5080.

Similarly, according to some leaks, the RTX 5080 has 16GB of GDDR7 memory at launch, and while this will be faster than the GDDR6X VRAM used in the RTX 4080, the amount is staying the same. However, another recent leak claimed that a 24GB RTX 5080 would be launched later, giving the card a bit more headroom for running games at high resolutions with lots of detail.

What’s more, it’s rumored that this 24GB card would have the same core GPU spec as the 16GB card, a feat that’s now technically possible thanks to the new Samsung 24Gb GDDR7 chip that’s just been announced. This chip has a 3GB capacity in the same package size as a 2GB chip, meaning it could be a drop-in replacement for 16GB cards.

Sadly, though, it looks as though the memory interface is going to have the same 256-bit width as the RTX 4080, which is disappointing if this ends up being an expensive GPU. If the RTX 5090 does indeed have a 512-bit interface, then the RTX 5080 should arguably at least have a 320-bit interface.

On the plus side, the use of GDDR7 VRAM should provide a bit more bandwidth. What’s more, the latest leak points to the RTX 5080 VRAM running at a super-fast speed of 32Gbps, which would help negate the narrow bus. If this leak is true, that would mean the RTX 5080 memory bandwidth is 1,024GB/s, which is even quicker than the 1,008GB/s bandwidth of the RTX 4090.

Finally, the RTX 5080 power draw is rumored to stand at 400W, which means it will only need one 16-pin power connector. Some speculation has claimed the RTX 5090 needs two power connectors, and while this is already unlikely, particularly for Founders Edition cards running at stock speed, it’s out of the question for the RTX 5080 if it draws 400W.

Nvidia GeForce RTX 5080 release date estimate

Our best estimate for the RTX 5080 release date window is January 2025, based on rumor and speculation, as Nvidia hasn’t officially announced a date. 

Historically, the 80-series GPU has been the first to launch, with the RTX 1080, 2080, and 3080 all arriving before the more expensive cards. However, Nvidia threw us a curveball when it launched the RTX 4090 before the 4080 back in 2022, and there are rumors that Nvidia may do this again with its new Blackwell lineup.

Reports are mixed here, though. Early leaks pointed to the RTX 5090 launching in 2024, and being the only GPU from the new lineup to do so. However, tech leaker kopite7kimi later claimed that the RTX 5080 would launch first, with the RTX 5090 coming later.

The same leaker has now said that no new RTX 5000-series Blackwell GPUs are expected to launch in 2024, with the expectation that both the RTX 5090 and RTX 5080 will launch in January 2025, possibly with an announcement at the CES tradeshow.

Nvidia GeForce RTX 5080 price estimate

Our best estimate for the RTX 5090 price is $999, based on the current price of the RTX 4080 Super, which was significantly cheaper than the original RTX 4080 when it launched.

We’re really hoping that Nvidia learned a lesson from the price of the original RTX 4080. Launching at $1,199, it was only $400 cheaper than the RTX 4090, but significantly less powerful. What’s more, its predecessor the RTX 3080 launched at an MSRP of just $699, even if the following GPU shortage meant it was practically impossible to actually buy at that price.

Sadly, we don’t think we’ll see the days of $699 80-series GPUs again, at least not unless AMD or Intel steps up to offer some serious competition, but we don’t think Nvidia is going to shoot itself in the foot with another $1,199 80-series GPU launch. $999 looks like the price people are willing to pay for this level of power, and that’s the price we’re expecting for the RTX 5080.

Nvidia GeForce RTX 5080 benchmarks estimates

Our Nvidia GeForce RTX 5080 benchmarks estimate puts the card as slightly faster than the RTX 4080 Super in rasterization, but substantially quicker in ray tracing.

There haven’t been many leaks about RTX 5080 performance yet, but one rumor is that the RTX 5080 ray tracing speed will be even quicker than that of the RTX 4090. This comes from regular tech leaker XpeaGPU, who claimed that the performance of the GB203 GPU rumored to be used in the RTX 5080 is “close to AD102 in raster and faster in RT.”

Over the last few generations, Nvidia’s new GPU architectures have made big strides in ray tracing performance over previous generations, but non-ray traced game performance (rasterization) has improved by a smaller degree. It looks as though this will also be the case with the RTX 5080.

Even with its purportedly low specs, this leak suggests the RTX 5080 will still be faster than the RTX 4090, thanks to the improvements made to the Blackwell architecture. We’ll only know for sure when we’ve had the cards in our hands to test for ourselves, however.

In the meantime, while we wait for the RTX 5080 launch, you can read our full guide to Nvidia DLSS, where we explore what makes Nvidia’s upscaling and frame-generation tech tick.