It is at the beginning of 2025 that Nvidia should unveil its next GeForce RTX 50 graphics cards using the Blackwell architecture. Here is everything we know about these future GPUs which are already shaping up to be power monsters.
The next generation of Nvidia graphics cards is expected around the corner after the Ada Lovelave architecture, which is very popular with gamers, even if it is far too expensive in the high-end segment. Nvidia should launch its most powerful GPUs first before tackling the mid-range and entry-level.
With Nvidia now one of the most publicly traded companies, its products are the subject of numerous leaks and rumors, some of which are now very reliable. We have brought together the most serious of them in this file to get an overview of what the Blackwell generation will offer in 2025.
GeForce RTX 50: what technical sheets for what power?
The technical specifications of the next GeForce RTX 5000 have been widely leaked in recent months, specifically for the RTX 5090 and RTX 5080. The two high-end graphics cards should get the ball rolling in 2025 with several technical advances to note.
We can, without being too wrong, estimate that all cards of this generation will use ultra-fast GDDR7 memory, almost doubling the bandwidth on the same bus width.
The RTX 5090 should therefore offer 21,760 CUDA cores, coupled with 32 GB of GDDR7 memory clocked at 28 Gbps on a 512-bit bus. The GPU would reach a record bandwidth of 1792 GB/s, unheard of on a consumer GPU. It would then consume 600W, compared to 450W for the RTX 4090.
The “little sister” RTX 5080 would only offer half the number of CUDA cores (10,752) with only 16 GB of GDDR7 memory. Some leaks suggest a slightly higher speed (32 Gbps) on a 256-bit bus, bringing the bandwidth to 1 TB/s. Its announced TGP would increase to 400.80W more than the RTX 4080. A 24 GB version would also be planned, thanks to 3 GB GDDR7 memory modules.
These two graphics cards are geared towards 4K gaming and ray tracing, but also towards AI and creativity workloads. We do not yet know the details of their Tensor and RT cores in this regard, but we imagine that they will be significantly increasing.
Other graphics cards could see the light of day without us having much information on their technical sheet. An RTX 5070 could be announced at the same time as the first two, with 12 GB of video memory on a 192-bit bus and a 250W TGP.
It is also possible that we will see an RTX 5070 Ti, but also the highly anticipated 5060 and 5060 Ti at the entry/mid range. These are generally the best-selling GPUs each generation: the RTX 3060 and 4060/Ti are the four graphics cards that take 4 of the top 5 spots of the most used graphics cards on Steam.
| NVIDIA GeForce RTX 5090 | NVIDIA GeForce RTX 5080 | NVIDIA GeForce RTX 4090 | NVIDIA GeForce RTX 4080 | |
|---|---|---|---|---|
| Number of SMS | 170 | 84 | 128 | 76 |
| CUDA hearts | 21760 | 10752 | 16384 | 9728 |
| Clock frequency | TBA | TBA | 2520 MHz | 2505 MHz |
| Memory | 32GB GDDR7 | 16GB GDDR7 | 24GB GDDR6X | 16GB GDDR6X |
| Memory bus | 512-bit | 256-bit | 384-bit | 256-bit |
| Memory speed | 28Gbps | 32Gbps | 21Gbps | 23Gbps |
| Memory bandwidth | 1792 GB/s | 1024 GB/s | 1008 GB/s | 736 GB/s |
| TGP | 600W | 400W | 450W | 320W |
| Connections | 3x DP 2.1a, 1x HDMI 2.1 | 3x DP 2.1a, 1x HDMI 2.1 | 3x DP 1.4a, 1x HDMI 2.1 | 3x DP 1.4a, 1x HDMI 2.1 |
| Interface | PCIe 5.0 x16 | PCIe 5.0 x16 | PCIe 4.0 x16 | PCIe 4.0 x16 |
RTX 50: a latest generation connectivity offer
Nvidia should finally embrace the additional bandwidth of PCIe 5.0 on motherboards released in recent years. The GPUs will, however, remain compatible with the PCIe 4.0 interface and the difference in performance between the two should be minimal as was the case with the RTX 4000.
On the connection itself, Blackwell should be fully compatible with DisplayPort 2.1a in UHBR20 (Ultra-High Bit Rate), or 80Gbps. HDMI 2.1 will logically be supported.
Finally, the connection to the power supply should be made via a 16-pin 12V-2×6 connector, the evolution of the maligned 12VHPWR present on the RTX 40. A 12-pin version of this connector had been discreetly implemented on the mid-range models. and high-end to prevent any risk of overheating.
What technologies and features for the RTX 50?
According to information on the site WccftechNvidia is reportedly preparing several announcements of “next generation AI technologies for gamers”, even described as a “major surprise”.
We imagine that DLSS will move to version 4.0 with the arrival of new technologies to improve both the image quality of our games and their performance. The ray reconstruction of DLSS 3.5 allowed better ray tracing rendering, with performance gains as a result.
A new generation of ray tracing cores, but also Tensor cores, should be there to accelerate the rendering of ray and path tracing games, but also applications linked to AI.
Nvidia could go further by using AI to “generate textures and objects” that are even more faithful, as Jensen Huang mentioned at Computex 2024. We remain here in the realm of supposition while nothing The software technologies that will be revealed alongside graphics cards in January 2025 have been leaked.
Finally, it is almost certain that the Nvidia App application will come out of its beta status during this launch. This new control panel should definitively replace GeForce Experience by the end of 2024 for an official launch in early 2025.
What price for the GeForce RTX 50?
It is currently very complicated to estimate the prices of these future graphics cards. In this exercise of pure speculation and according to the latest rumors relayed by the YouTube channel Moore’s Law Is Deadthe RTX 5090 could exceed the 2000 dollar mark (minimum 2400 euros), the RTX 5080 at minimum 1200 dollars (1450 euros) and the RTX 5070 between 600 and 800 dollars (minimum 720 euros).
It is therefore the RTX 5090 which would take the large price level while we can expect an intermediate model, a more muscular evolution of the RTX 5080. However, we take these figures with a grain of salt, hoping that this inflation prices are not confirmed.
GeForce RTX 50 release date
Nvidia has confirmed its presence at CES 2025 in Las Vegas for a conference on Monday January 6, the day before the show opens. We expect a complete presentation of this new mainstream GPU architecture based on Blackwell, the different models, technologies and finally, prices.
According to the latest rumors from Board Channels forumsthe GeForce RTX 5090 and RTX 5080 are expected to launch in January 2025 while the RTX 5070 Ti and 5070 are scheduled for February. The future RTX 5060 Ti and RTX 5060 would finally be launched in March.
