Nvidia introduced its new graphics cards with the GeForce RTX 50 at CES 2025 in Las Vegas. Featuring Blackwell AI accelerator architecture, this generation has the powerful RTX 5090 as its flagship but also includes the amazing RTX 5080 and RTX 5070.
While there is a lot of eccentric news in the hardware world, TecMundo has prepared a complete guide about the new desktop RTX 50 and its technical features. GPUs full of innovations promise improved performance and high prices.
RTX 5090: unlimited power
Creating a trend in the graphics card industry once again, Nvidia introduced the naming ending with 90 to its high-end product. GeForce RTX 5090 comes as the world’s most powerful new home graphics card, and this is not an exaggeration because there is no model with features that can rival this GPU.
This is the model developed specifically for what we call enthusiasts, that is, gamers who want to get the most out of every game, content creators, and even professionals working with heavy rendering and 3D modelling. Simpler users will also have access to the model, but they will have to pay a lot of money, and there are more cost-beneficial options.
As for its technical features, the GeForce RTX 5090 brings a significant generation leap compared to the RTX 4090, with 32% more CUDA cores with 21,760 units. The new GPU offers faster GDDR7 type VRAM memories in a 32 GB configuration with very high bandwidth of 1,792 GB/s.
- NVIDIA CUDA Cores: 21,760
- Tensor Cores (IA): 3,352 5th Generation AI TOPS
- Ray Tracing Cores: 318 4th Gen TFLOPS
- Clock Acceleration: 2.41GHz
- Base Time: 2.01GHz
- Default Memory Configuration: 32GB GDDR7
- Memory Bus Width: 512 bits
- NVIDIA Architecture: blackwell
- Supported technologies: DLSS 4, Super Resolution, DLAA, Beam Reconstruction, Frame Rendering, Multi-Frame Rendering, Reflex 2, PCI Express Gen 5, Resizable BAR, RTX Remix and others.
- Maximum Digital Resolution: 4K at 480 Hz or 8K at 120 Hz (with DSC – Video Stream Compression)
- Standard Video Connectors: 3x DisplayPort and 1x HDMI
- Multiple Monitor Support: Up to 4 simultaneous monitors
- Length: 304mm
- Width: 137mm
- Thickness: 2 slots
According to the graphics published by Nvidia itself, all this firepower translates into performance. In games like Cyberpunk 2077, Black Myth: Wukong and Alan Wake 2, the new GPU can deliver twice the performance compared to the RTX 4090 with Ray Tracing enabled and the new DLSS 4.
Speaking of DLSS, not only the RTX 5090, but the entire RTX 50 generation focuses on AI to the maximum. For comparison, the RTX 5090 has an AI TOPS (unit that measures a chip’s performance on relevant tasks) of 3,352. This is a significant increase compared to its predecessor’s 1,321.
With many innovations, the 5090 is a power-hungry GPU. The manufacturer reports that the graphics card has a consumption of 575W and recommends users to use sources with a power of at least 1000W.
The GeForce RTX 5090 will be available for purchase on the international market starting January 30 for US$ 1,999 (about R$ 12,000 in direct conversion, excluding taxes).
RTX 5080: improved performance
If we move the power scale down a bit, the GeForce RTX 5080 is the second most powerful model announced by Nvidia. This GPU is a model that appears to be in the high-end category and is considerably inferior to the RTX 5090, as it costs half the price and has a significant cut-off in the chip.
If you still can’t understand the extent of the cut the company has made, the RTX 5080 only has 10,752 CUDA cores compared to its bigger sister’s 21,000 cores. This number isn’t that surprising when we throw in the standard RTX 4080 and its 9,728 units; this represents only a 10% increase. Compared to the RTX 4080 SUPER (10,240 cores), the technical difference is around 5%.
Just like the RTX 4080, the new RTX 5080 continues with 16 GB VRAM, but also uses the new GDDR7 type with 960 GB/s bandwidth. Energy consumption does not change much and the chip draws increases from 320W to 360W and, like its predecessor, requires an 850W power supply.
Speaking of what really matters, the performance figures again show an interesting generation leap. Cyberpunk 2077, Alan Wake 2 and Black Myth run at twice the performance compared to the regular RTX 4080 with Ray Tracing turned on and with the help of DLSS 4.
So how can the RTX 5080 achieve 2 times more performance if the core count is only 10% higher than the previous model? The answer is that the RTX 50 generation introduces the Blackwell microarchitecture; This means that cores and other chip structures are optimized; In addition, DLSS 4 and Multi-Frame Generation are exclusive to this new series, frame rate.
The GeForce RTX 5080 will be priced at US$999 (about R$6,000 in direct conversion) and will be released on January 30.
- Architectural: blackwell
- NVIDIA CUDA Cores: 10752
- Tensor Cores (IA): 1801 AI TOPS (5th generation)
- Ray Tracing Cores: 171 TFLOPS (4th generation)
- Speed increase: 2.62GHz
- Base clock: 2.30GHz
- Graphics memory: 16GB GDDR7
- Memory Band: 960GB/s
- Supported technologies: DLSS 4, Super Resolution, DLAA, Beam Reconstruction, Frame Rendering, Multi-Frame Rendering, Reflex 2, PCI Express Gen 5, Resizable BAR, RTX Remix and others
- Maximum Digital Resolution: 4K at 480 Hz or 8K at 120 Hz with DSC (Display Stream Compression)
- Standard Video Connectors: 3x DisplayPort and 1x HDMI
- Multiple Monitor Support: Up to 4 simultaneous monitors
- Length: 304mm
- Width: 137mm
- Thickness: 2 slots
- TGP: 360W
- Required system resource power: 850W.
RTX 5070 Ti vs RTX 5070: superior performance?
As with other generations, Nvidia announced the top-of-the-line RTX 5090 and RTX 5080 series, but also revealed cheaper models. In this sense, GeForce RTX 5070 is the most basic graphics card of the new generation and promises performance beyond expectations.
This board currently has 6,144 CUDA cores; this is 5% less than found on the RTX 4070. The new graphics chip comes with 12 GB GDDR7 VRAM with 670 GB/s bandwidth in addition to 250 W consumption. With a 650W source recommendation for use in computers.
The biggest attraction of this announcement, which surprised many consumers, was Nvidia’s announcement that the RTX 5070 will have the performance of the RTX 4090 at an extremely low price. Promotional graphics confirm that it offers twice the performance in games than the standard RTX 4070, but there is a secret to the RTX 4090’s claim.
Nvidia says that the RTX 5070 will have the highest level of power of the last generation, yes, but this will be done with the help of artificial intelligence. This performance increase will only be possible thanks to DLSS 4 and frame rendering capabilities, but raw performance (without the help of any technology) will keep the two cards separate.
The second announcement in this family was the RTX 5070 Ti, which was also an unexpected surprise. This GPU will have 8,960 cores, a significant increase over the standard 5070, and will also have 16GB of GDDR7 VRAM; This will make it much more suitable for running 4K games than its younger sibling.
Compared to the RTX 4070 Ti, the new Nvidia card also has the same operating mode as the others, doubling the performance in certain games; It also offers upscaling via Ray Tracing and DLSS 4.
The price of the GeForce RTX 5070 will be 549 US dollars (R$ 3,300), while the price of the GeForce RTX 5070 Ti will be 749 US dollars (R$ 4,500). Both GPUs are hitting the international market in February.
RTX 5070 Features
- Architectural: blackwell
- NVIDIA CUDA Cores: 6,144
- Tensor Cores (IA): 988 AI TOPS (5th generation)
- Ray Tracing Cores: 94 TFLOPS (4th generation)
- Speed increase: 2.51GHz
- Base clock: 2.16GHz
- Graphics memory: 12GB GDDR7
- Memory Band: 672 GB/s
- Supported technologies: DLSS 4, Super Resolution, DLAA, Beam Reconstruction, Frame Rendering, Multi-Frame Rendering, Reflex 2, PCI Express Gen 5, Resizable BAR, RTX Remix and others
- Maximum Digital Resolution: 4K at 480 Hz or 8K at 120 Hz with DSC (Display Stream Compression)
- Standard Video Connectors: 3x DisplayPort and 1x HDMI
- Multiple Monitor Support: Up to 4 simultaneous monitors
- Length: 242mm
- Width: 112mm
- Thickness: 2 slots
- TGP: 250W
- Required system resource power: 650W.
RTX 5070 Ti specifications
- Architectural: blackwell
- NVIDIA CUDA Cores: 8,960
- Tensor Cores (IA): 1406 AI TOPS (5th generation)
- Ray Tracing Cores: 133 TFLOPS (4th generation)
- Speed increase: 2.45GHz
- Base clock: 2.3GHz
- Graphics memory: 16GB GDDR7
- Memory Band: 896GB/s
- Supported technologies: DLSS 4, Super Resolution, DLAA, Beam Reconstruction, Frame Rendering, Multi-Frame Rendering, Reflex 2, PCI Express Gen 5, Resizable BAR, RTX Remix and others
- Maximum Digital Resolution: 4K at 480 Hz or 8K at 120 Hz with DSC (Display Stream Compression)
- Standard Video Connectors: 3x DisplayPort and 1x HDMI
- Multiple Monitor Support: Up to 4 simultaneous monitors
- Length: varies depending on manufacturer
- Width: varies depending on manufacturer
- Thickness: varies depending on manufacturer
- TGP: 300W
- Required system resource power: 750W.
Source: Tec Mundo
I am a passionate and hardworking journalist with an eye for detail. I specialize in the field of news reporting, and have been writing for Gadget Onus, a renowned online news site, since 2019. As the author of their Hot News section, I’m proud to be at the forefront of today’s headlines and current affairs.
