This year kicked off with the launch of the GeForce RTX 40 "SUPER" series, which introduced three new GPUs based on the Ada Lovelace architecture. These GPUs are designed to deliver better performance and value than non-SUPER variants, and have received a positive response overall. And NVIDIA's next-generation GeForce RTX 5090 and RTX 5080 "Blackwell" graphics processors are rumored to be available in the fourth quarter of 2024.
The NVIDIA Blackwell GPUs released so far are entirely built for the AI and HPC segments, but what we know for sure is that there is already a Blackwell GPU IP that will power the next generation of RTX graphics cards. It is expected that these graphics cards will use the RTX 50 series brand.
It is reported that Nvidia will once again focus on high-end graphics cards, such as the GeForce RTX 5090, followed by the GeForce RTX 5080 in the fourth quarter of 2024. These graphics cards will allow NVIDIA's board partners to expand shipments, and the introduction of these GPUs will also help drive up the average selling price (ASP) of GPUs as more gamers upgrade to the top end.
As Nvidia has stormed the artificial intelligence business, graphics card-related products have also continued to innovate. Board makers expect Nvidia's latest RTX 50 series graphics cards to be unveiled in the fourth quarter. Initially, it will focus on the high-end and high-end 5090 and 5080 product lines. The firm is optimistic that the launch of Nvidia's biennial GPU update will help board makers ship this year and will drive up the average unit price (ASP) of graphics card products.
According to previous rumors, Nvidia is expected to launch two flagship chips, namely GB202 and GB203. The Blackwell GB202 gaming graphics processor will use TSMC 4NP nodes to significantly increase cache and SM throughput. The former is likely to be adopted by the GeForce RTX 5090, while the latter is expected to be adopted by the RTX 5080, and both GPUs are expected to use the latest GDDR7 memory interface, with bus interfaces up to 512 bits. Here are the memory configurations for the various Blackwell GPU SKUs:
GB202 - 512-bit/28 Gbps / 32 GB (maximum memory) / 1792 GB/s (maximum bandwidth)
GB202 - 384-bit / 28 Gbps / 24 GB (Max Memory) / 1344 GB/s (Max Bandwidth)
GB203 - 256-bit / 28 Gbps / 16 GB (Max Memory) / 896.0 GB/s (Max Bandwidth)
GB205 - 192-bit / 28 Gbps / 12 GB (Max Memory) / 672.0 GB/s (Max Bandwidth)
GB206 - 128-bit / 28 Gbps / 8 GB (Max Memory) / 448.0 GB/s (Max Bandwidth)
GB207 - 128-bit / 28 Gbps / 8 GB (Max Memory) / 448.0 GB/s (Max Bandwidth)
The last two major releases, Ampere (GeForce RTX 3090) and Ada (GeForce RTX 4090), were both at the end of Q3 or the beginning of Q4. As a result, the release dates of the next-gen flagships, the GeForce RTX 5090 and RTX 5080, in Q4 2024 sound credible.