Nvidia RTX 5090: Leaked specifications, performance & more

Looking for the latest on the RTX 5090? We’ve got you covered with the latest leaks and rumors, including rumored specifications, performance, and more.

Nvidia’s RTX 50-series graphics cards are already being worked on, using what many presume to be Nvidia’s Blackwell architecture. the top tier GPU is posed to be the RTX 5090. It will serve as a successor to the RTX 4090, which is the fastest gaming GPU you can buy today.

With scarce information out there on what the GPU actually might be, we’ve scoured the interest for the latest information on the RTX 5090, and what you should come to expect from the GPU ahead of its official announcement.

Just remember, since these are all leaks and rumors, these specs are subject to change, and you should take them with a pinch of salt until Team Green announces something official.

RTX 5090 leaked specifications

Nvidia

• CUDA Cores: 24,576
• Memory bandwidth: 1.53 TB/s
• Memory bus width: 384-bit / 512-bit
• Cache: 128MB
• Boost clock: 2.9 Ghz

According to the first-ever reports on what we can expect from the RTX 5090, it will sport blazing-fast speeds. It was initially posted by hardware leaker Panzerlied on Chiphell forums but was later deleted. The post detailed various performance uplifts in certain parts of the card, which we can extrapolate into hard specifications.

While the current CUDA core count raises some eyebrows, the GPU might be able to realistically achieve it if Blackwell is not monolithic. This means that the chip could be spread across multiple chiplets, which could be denser than one, big monolithic design.

Secondly, the leaked memory bandwidth of the RTX 5090 showcases that the GPU might have at least a 384-bit or 512-bit bus width. Additionally, the RTX 5090 might also make use of GDDR7 memory, which will allow for transfer speeds of up to 32 GT/s. This is more than enough to meet the memory bandwidth claims. Kopite7Kimi further corroborated that the 5090 might feature a massive 512-bit bus.

Combined with multiple sources, I confirm the gaming flagship of Ada-next will have a 512-bit memory interface.

— kopite7kimi (@kopite7kimi) July 27, 2023

You should also expect a 128MB cache, though it is not explained whether or not this is a shared cache, or limited to L1, L2, or L3. We suspect that more details surrounding the cache will be brought to light in the future.

Lastly, you can expect the GPU to reach a boost clock of up to 2.9 GHz, this is not too surprising, since we have seen overclocked RTX 4090’s break the 3GHz barrier in the past. One thing that we do not know, is how much power the RTX 5090 might take up while maintaining those clocks. But, if Blackwell ends up being anything like Ada Lovelace, then we might have an incredibly efficient chip on our hands.

GMB202

According to a report by Videocardz, the RTX 50-series of graphics cards will be using the GMB2XX moniker in order to discern between chips. The RTX 5090 is likely to use one of the largest chips from this lineup and might be granted the GMB202 moniker.

Again, take this initial leak with a grain of salt, and we’ll keep our ears to the ground when we hear more.

RTX 5090 leaked performance speculation

Dexerto

According to the Panzerlied leak, the RTX 5090 will end up being around 70% faster than the RTX 4090. Though, it’s unclear how, or where this figure might be replicated. It’s likely that the leaker is talking about pure, rasterized performance, with no DLSS bells and whistles to boost the performance up.

This ends up being around 1.7x faster than the current fastest gaming GPU, meaning that the Blackwell generation might end up as one of the larger generational leaps in GPUs that we have seen for quite some time.

RTX 5090 release date speculation

Nvidia’s RTX 5090 might not debut for some time, with many anticipating it to be released in 2025. It’s likely that the earliest we might see it debut would be in Fall 2024. However, Nvidia will want to ensure that the RTX 40-series of graphics cards is given as much of a lifespan as possible, as rumors report that Blackwell will be more expensive to manufacture than previous nodes.

RTX 5090 price speculation

It’s likely that the RTX 5090 will retail at around $1699 at launch. This is $100 higher than the RTX 4090, simply down to market factors like the chip being more expensive to be made, inflation, and other global factors. We hope it’ll be less than that, but with TSMC’s chips more in demand than ever, it’s likely that Nvidia will stick to its guns when it comes to pricing. The company might also be entirely unchallenged in the high-end, essentially allowing them to charge whatever they want.

Should you wait for the RTX 5090?

If you’re looking to drop a load of cash on a GPU, it’s likely that in the immediate term, you might be better off picking up an RTX 4090 for now, especially since the RTX 5090 might be a little while away from releasing right now.