Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
NVIDIA introduced the GeForce RTX 40 line, which so far includes only the RTX 4090 and two variants of the RTX 4080. The new products are based on the Ada Lovelace architecture and are made using TSMC’s 4nm process technology. According to NVIDIA, video cards provide a huge performance boost compared to the RTX 30. In particular, performance in shader processing has increased by 2 times, in tasks using RT cores – by 2.8 times, and the energy efficiency of new products has improved by almost 2 times .
At the start of the generation, NVIDIA is focusing on the high-end category, so for now only RTX 4090 and RTX 4080 graphics cards are presented. In the future, of course, RTX 4070, RTX 4060 and other options will appear, but now the focus is on demonstrating the capabilities of the Ada Lovelace architecture.
So, the main technical characteristics of the RTX 4090 and RTX 4080:
| RTX 4090 | RTX 4080 (16 GB) | RTX 4080 (12 GB) | |
| Number of CUDA Cores | 16384 | 9728 | 7680 |
| Number of shader cores | 128 | 76 | 60 |
| GPU clock (in Boost state) | 2520 MHz | 2505 MHz | 2640 MHz |
| Amount of memory (type) | 24 GB (GDDR6X) | 16 GB (GDDR6X) | 12 GB (GDDR6X) |
| Bus speed | 384 bit | 256 bit | 192 bit |
| Memory Bandwidth |
1008 GB/s | 736 GB/s | 504 GB/s |
| Power consumption | 450 W | 340 W | 285 W |
| Price (Founders Edition) | $1499 | $1199 | $899 |
NVIDIA notes that the RTX 4080 with 12 GB of memory should be the ideal option for most gamers, as it offers the best of the latest Ada Lovelace architecture. The effective throughput is just over half of what you can get with the RTX 4090, but for the vast majority of gamers, this will be more than enough.
Lovelace architecture: high performance and consumption
Ada Lovelace — is the name of the latest NVIDIA graphics architecture used in the GeForce RTX 40 series graphics cards. This architecture provides a significant performance boost over Ampere. The company says the new lineup of GPUs delivers up to 83 TFLOPS of processing power for PC gaming, double that of RTX 30 cards.
The performance of RT-cores, according to NVIDIA, is 191 TFLOPS, which is 2.8 times more than its predecessor. Also, third-generation RT cores are designed to provide realistic lighting effects.
DLSS 3 is a new version of NVIDIA’s supersampling technology that’s made even better by using a new “neural network to increase frame rates while maintaining image quality.” The engineers also managed to bypass CPU limitations, allowing the GPU to render frames created with DLSS Frame Generation on its own.
Finally, hardware support for the AV1 encoder has been added. This news will certainly please content creators, as the time to encode videos in AV1 will noticeably decrease.
