Anticipation for the Nvidia GeForce RTX 5090 is high as consumers expect another record-breaking GPU. Here’s everything we know about this upcoming card, including its speculated price, specifications, and rumored performance power.
The Nvidia GeForce RTX 4090 is the fastest consumer graphics card on the market, and as such its successor has big shoes to fill. The RTX 5090 is the hotly anticipated follow-up to the RTX 4090, and consumers are expecting big things from this upcoming card.
Nvidia has not yet released any solid details on this GPU, and much of the information we have on it has been obtained via industry rumors and leaks. As always, this means that until Nvidia issues official statements, this information needs to be taken with a generous pinch of salt.
Release window speculation
According to Taiwanese news publication United Daily News, the RTX 5090 is planned for release in Q4 of 2024. This date would mark two years since the reveal of the RTX 40 series. According to this leak, this decision is to: “help board card manufacturers shipments this year and drive up the average unit price of graphics card products.”
Regular leaker Moore’s Law is Dead says that the RTX 5090 is set to launch first, and will be the only card in the RTX 50 series to launch within 2024. Fellow leaker Kopite7Kimi has challenged this, however. Kopite expressed the belief that the RTX 5080 will be released first, with the RTX 5090 to follow at the tail end of 2024.
Leaked specifications
It is expected that the RTX 4090 will use the GB202 chip. This chip is the flagship of the new Blackwell architecture line, which was officially announced by Nvidia at GTC 2024.
Analysts predict that the GB202 gaming chip will have significant performance gains compared to the current Ada Lovelace AD102 chips. The L1 cache in particular is expected to see large gains in performance.
Reliable leaker Kopite7Kimi indicated that the GB202 will use the same process node as the AI-focussed GB101 chip. In addition, Kopite7Kimi suggested that the chip would have 12 GPCs, with 8 TPCs each, for a total of 96TPCs.
Another reliable hardware leaker called Panzerlied shared some possible specifications for the RTX 5090, which if accurate could show some very impressive improvements.
Spec | RTX 4090 | RTX 5090 |
Streaming Multiprocessors | 128 | 192 |
CUDA Cores | 16,384 | 24,576 |
Ray Tracing Cores | 128 | 192 |
Tensor Cores | 512 | 768 |
Boost Clock | 2.52 GHz | 2.9 GHz |
L2 Cache | 72MB | 128MB |
Memory Bandwidth | 1,008 GB/s | 1,532 GB/s |
When it comes to the memory bus, there is some debate. Early information indicated that Nvidia was planning to reuse the existing 384-bit memory bus from the Ada chips. However, more up-to-date information suggests that Nvidia will be using a 512-bit memory bus, but only on flagship chips such as the GB202 that is expected to be used in the RTX 5090.
A May 2024 rumor also from Panzerlied suggests that the RTX 5090 will feature a very ‘dense’ memory layout. The leaker specifies that the GPU will use four memory modules on top, with a further five on each side and two at the bottom.
Panzerlied supplied further information in July 2024 that the Base Clock speed for the RTX 5090 could be as high as 2.9GHz. This approaches the 3GHz barrier and is an astonishing 30% uplift in Base Clock speed compared to the RTX 4090.
The Boost Clock speed has not yet been specified, but with such a high Base Clock speed, the RTX 5090 should be capable of speeds well above 3GHz with the Boost clock engaged.
As a comparison, here are the currently rumored specifications of the RTX 5090 compared with the RTX 4090:
GPU | Bus interface | Base Clock | Memory | VRAM |
GeForce RTX 5090 (Rumor) | 448-bit (512-bit full chip) | ~2900 MHz Base Clock | 14 GDDR7 Modules (16 Full) | 28 GB VRAM |
GeForce RTX 4090 (Official) | 384-bit (384-bit full chip) | 2235 MHz Base Clock | 12 GDDR6X Modules (12 Full) | 24 GB VRAM |
A Twitter/X user called Everest found that Seasonic had included data from the RTX 50 series of graphics cards on the PSU wattage calculator. Though this information may not be completely accurate, it does fall in line with expected figures.
Assuming the data is accurate, customers may wish to upgrade their PSU before purchasing an RTX 5090. The Seasonic data puts the TDP of the RTX 5090 at a whopping 500W.
Benchlife has contributed to rumors about the TDP of the RTX 5090 as of early September 2024. Benchlife has close ties with Nvidia board partners and is well-placed to have access to accurate figures. It says that the RTX 5090 is set to be a 550W card.
Subscribe to our newsletter for the latest updates on Esports, Gaming and more.
Kopite7Kimi posted that the RTX 5090 would draw a massive 600W of power. However, further data from Benchlife suggests this relates to a high-wattage cooling module for the card, and the 600W figure refers to the heat dissipation capacity, not the TDP.
Design speculation
If the memory configuration suggested above is accurate, it will require an entirely new PCB design to host this unusual layout. It also further strengthens rumors that the RTX 5090 will be using a 512-bit memory bus, as significant bit width would be required for such a dense memory configuration.
Related to this is a late May 2024 leak, which once again comes from Kopite7Kimi. This leak proposes that the RTX 5090 Founders Edition will use a radical new layout for the PCB. Specifically, the board will be divided into three sections: the main board, a rigid I/O board, and a separate PCIe slot section.
According to this leak, the main board and I/O board are needed to host the increased number of components required by Blackwell architecture. The Rigid I/O board could be used to house display controllers, the power delivery system, or other components that do not need active cooling. The PCIe slot, by contrast, will act as a riser, not containing any circuitry on its own.
Based on these leaked design specifications and looking back at previous multi-PCB designs from Nvidia, it is possible that this layout could potentially make the RTX 5090 thinner than its predecessors.
Once again, Kopite7Kimi is backing up this theory. A user on Twitter/X questioned if the RTX 5090 would use the bulky 4-slot cooler originally designed for the unreleased RTX 4090 Ti. Kopite replied that the RTX 5090 Founders Editon would instead use a two-slot cooler with a dual fan layout.
Kopite further stated that, unlike rivals AMD, Nvidia is not using chiplet construction. The GB202 chip is likely to power the RTX 5090, and Kopite’s leak suggests that this chip will continue to use monolithic GPU dies.
Performance speculation
According to the Panzerlied leak, the RTX 5090 will end up being around 70% faster than the RTX 4090. However, it’s unclear how, or where this figure might be replicated. The leaker is likely 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.
Features speculation
Current data suggests that the RTX 5090 will feature support for Display Port 2.1 as well as HDMI 2.1, and update on the Ada Lovelace series, which did not offer support for DisplayPort 2.1.
The power connector has turned out to be something of a sticking point. There have been numerous reports of reliability issues with the 12VHPWR connector used in the RTX 4090 and 40 Super cards, with some consumers reporting melted connectors.
Another persistent and reliable leaker called Moore’s Law is Dead reported that Nvidia is planning to switch to a new type of connector for its newest generation of cards. The leak goes on to say that Team Green is working on an entirely new type of 16-pin connector. So far, no other leaks have substantiated this rumor, however.
Price speculation
Though very little information is currently available on what can be expected regarding the price structure for the RTX 50 series, some analysts are predicting that the RTX 5090 will retail with an MSRP of $1699 at launch.
The RTX 4090 was unchallenged in its category. With the RTX 5090 set up to have similar dominance at the extremely high end of the market, this could give Nvidia carte blanche to charge whatever they choose.
Should you wait for the RTX 5090?
Customers looking to buy a top-end GPU might be advised to wait until the end of 2024. If the RTX 5090 launches during the expected window, those with deep pockets will have a shiny new toy to play with. Those looking for a bargain could also benefit by snatching up discounted RTX 4090 or RTX 4080 Super cards when retailers look to offload old stock.