During its keynote at CES 2025, NVIDIA officially introduced the highly anticipated next-generation GeForce RTX 50 series GPUs, built on the new Blackwell architecture. The lineup includes four models, from the RTX 5070 to the RTX 5090, with prices ranging up to $1,999. According to NVIDIA, these GPUs are set to launch between January and February, offering up to 2x the performance of their predecessors.
The RTX 50 series targets a wide range of users, with offerings for enthusiasts, high-end gamers, and intermediate-level users. As expected, the new cards bring significant gaming and AI capabilities advancements—a key focus for NVIDIA.
Jensen Huang, NVIDIA’s CEO, highlighted the incredible power of each GPU, measured in TOPS (Trillions of Operations Per Second), to emphasize their AI processing capabilities. With this announcement, NVIDIA has firmly positioned the RTX 50 series as a major leap forward in gaming and AI performance.
GeForce RTX 5090: the new high-end GPU
Priced at $1,999, the RTX 5090 stands out as NVIDIA’s most powerful GPU, costing $400 more than its predecessor. It delivers up to 3,352 TOPS for AI applications, a significant leap in processing power.
The RTX 5090 is equipped with 21,760 CUDA cores and introduces 32 GB of GDDR7 memory, marking the debut of this cutting-edge memory technology. Running at 28 Gbps on a 512-bit bus, the GPU achieves an astonishing memory bandwidth of 1,792 GB/s, ensuring exceptional gaming and AI workload performance.
| GeForce RTX 5090 Specifications | ||
| GPU | RTX 5090 | RTX 4090 |
| CUDA Cores | 21,760 | 16,384 |
| Base and boost clocks | 2.01/2.41 GHz | 2.23/2.52 GHz |
| Memory | 32GB GDDR7/512-bit | 24GB GDDR6X/384-bit |
| Memory speed | 28Gbps/1,792GB/s | 21 Gbps/1.01 TB/s |
| TDP | 575W | 450W |
| Launch price | $1,999 | $1,599 |
According to NVIDIA, the RTX 5090 delivers up to twice the gaming performance of its predecessor, the RTX 4090, in demanding titles such as Cyberpunk 2077, Alan Wake 2, and Black Myth: Wukong. These games were tested at 4K resolution, with maximum graphics settings, ray tracing, and DLSS in performance mode combined with frame generation.
However, this incredible power comes with a hefty energy requirement—the GPU has a base TDP of 575W, emphasizing its high-performance nature. An interesting detail about the RTX 5090 is its more compact reference design. Unlike the RTX 4090 Founders Edition, known for its bulk, the new card fits into just two slots, making it significantly slimmer and more practical. The RTX 5090 is set to launch globally on January 30th, promising to set a new benchmark for gaming and AI performance.
GeForce RTX 5080
Following in the footsteps of the flagship RTX 5090, the RTX 5080 also claims to deliver up to 2x the performance of its predecessor, the RTX 4080, in the same demanding games and conditions. Equipped with 1,801 TOPS, this high-end GPU features 10,752 CUDA cores operating at clock speeds between 2.30 GHz and 2.62 GHz.
| GeForce RTX 5080 Specifications | ||
| GPU | RTX 5080 | RTX 4080 |
| CUDA Cores | 10,752 | 9,728 |
| Base and boost clocks | 2.30/2.62 GHz | 2.20/2.50 GHz |
| Memory | 16 GB GDDR 7/256-bit | 16GB GDDR6X/256-bit |
| Memory speed | 30Gbps/960GB/s | 22.4Gbps/716.8GB/s |
| TDP | 360W | 320W |
| Launch price | $999 | $1,199 |
The RTX 5080 includes 16 GB of GDDR7 memory—half the VRAM of the RTX 5090—running at an impressive 30 Gbps on a 256-bit bus, resulting in 960 GB/s of bandwidth. Its power consumption has increased slightly to a TDP of 360W, which is 40W higher than the RTX 4080.
The RTX 5080 will launch alongside the RTX 5090 on January 30th, priced at $999, making it a compelling option for gamers seeking top-tier performance without reaching for the flagship model.
GeForce RTX 5070 and 5070 Ti
In the mid-range segment, NVIDIA has introduced the RTX 5070 Ti and RTX 5070, offering powerful performance at more accessible price points.
The RTX 5070 Ti features 8,960 CUDA cores and shares the same 16 GB GDDR7 memory configuration as the RTX 5080 but at a reduced speed of 28 Gbps, providing 896 GB/s bandwidth. Its power consumption is rated at a 300W TDP.
| GeForce RTX 5070 Ti and 5070 specifications | ||||
| GPU | RTX 5070 Ti | RTX 4070 Ti | RTX 5070 | RTX 4070 |
| CUDA Cores | 8,960 | 7,680 | 6,144 | 5,888 |
| Base and boost clocks | 2.30/2.45 GHz | 2.31/2.61 GHz | 2.16/2.51 GHz | 1.92/2.48 GHz |
| Memory | 16 GB GDDR 7/256-bit | 12GB GDDR6X/192-bit | 12GB GDDR7/192-bit | 12GB GDDR6X/192-bit |
| Memory speed | 28Gbps/896GB/s | 21Gbps/504GB/s | 28Gbps/672GB/s | 21Gbps/504GB/s |
| TDP | 300W | 285W | 250W | 200W |
| Price | $749 | $799 | $549 | $599 |
The RTX 5070, capable of delivering up to 2x the performance of its predecessor, comes equipped with 6,144 CUDA cores and operates at clock speeds between 2.16 GHz and 2.51 GHz. It includes 12 GB of GDDR7 memory with a 192-bit bus, achieving 672 GB/s bandwidth and consuming 250W.
Both models are set to launch globally in February 2025 (exact date TBD). Pricing is expected to start at $549 for the RTX 5070 and $749 for the RTX 5070 Ti. NVIDIA CEO Jensen Huang stated that the RTX 5070 will deliver performance equivalent to the RTX 4090 but with significantly lower costs and power consumption.
NVIDIA has not revealed details about the entry-level RTX 5060 series GPUs.
