NVIDIA has unveiled the RTX 4080, which will soon be launching in two configurations. As expected, the RTX 4080 series are some of the first graphics cards based on NVIDIA's Ada Lovelace architecture. Leveraging TSMC 4N nodes, NVIDIA will sell the pair as the RTX 4080 12 GB and RTX 4080 16 GB. However, there are more differences than their names suggest, with NVIDIA equipping the RTX 4080 12 GB with a less GPU than the RTX 4080 16 GB.
Specifically, the RTX 4080 12 GB features 7,680 CUDA cores, putting it between the CUDA core volumes RTX 3070 Ti and RTX 3080. Additionally, the RTX 4080 12 GB utilises 16 GB of GDDR6X VRAM, which NVIDIA claims helps the card to outperform the RTX 3090 Ti, but only with DLSS 3 enabled. By contrast, the RTX 4080 16 GB has 9,728 CUDA cores, along with GDDR6X VRAM like the RTX 4080 12 GB. Purportedly, the RTX 4080 16 GB offers 2x the performance of the RTX 3080 Ti when using DLSS 3 and matches the RTX 3090 Ti while consuming less power.
So far, NVIDIA has confirmed that the RTX 4080 series will also feature third-generation RT cores, fourth-generation Tensor cores and an Ada Optical Flow Accelerator. Moreover, the RTX 40 series offers Shader Execution Reordering and Dual NVIDIA Encoders (NVENC). Reputedly, NVENC will cut export times by up to half while featuring AV1 support too.
Currently, NVIDIA is only committing to launching the RTX 4080 12 GB and RTX 4080 16 GB in November at US$899 and US$1,119 (MSRP). Following EVGA's public departure from NVIDIA, ASUS, Colorful, Gainward, Galaxy, GIGABYTE, Innovision 3D, MSI, Palit, PNY and Zotac will be on hand to deliver custom SKUs. Likewise, Acer, Alienware, ASUS, Dell, HP, Lenovo and MSI will soon release pre-built systems with the RTX 4080 series as well.
Source(s)
NVIDIA