The codenames for NVIDIA Blackwell Graphics Architecture GPUs have been disclosed, with no successor for AD104.

The upcoming GeForce RTX 50-series graphics cards will be powered by the Blackwell graphics architecture, which is named after American mathematician David Blackwell. According to a reliable source known for NVIDIA leaks, the lineup of GPUs in this series is expected to include the GB202, GB203, GB205, GB206, and GB207. Interestingly, there is no mention of a "GB204" following the previous AD104, GA104, TU104 GPUs that have been successful in the performance-segment.

The GeForce Blackwell ASIC series starts with "GB" (GeForce Blackwell) followed by a 200-series number. The last time NVIDIA used a 200-series ASIC number for GeForce GPUs was with the "Maxwell" architecture, where the GPUs ended up being built on a more advanced node and with additional advanced features compared to the original conception of the architecture. In the case of "Blackwell," the GB202 logically succeeds the AD102, GA102, TU102, and other "big chips" that have powered NVIDIA's flagship client graphics cards. The GB103 follows the AD103, serving as a high SIMD count GPU with a narrower memory bus compared to the GB202, and it powers the second and third SKUs in the series. Interestingly, there is no "GB104" mentioned, even though NVIDIA's xx04 ASICs have been successful in the performance-thru-high end SKUs, such as the TU104 for the RTX 2080 and the GP104 for the popular GTX 1080 and GTX 1070 series. This has been a missing component in the past two generations.

The GA104 based on "Ampere" architecture powers the RTX 3070, which has sold well, but the maxed out RTX 3070 Ti has not performed as expected and has fallen short against the Radeon RX 6800 with a similar price. Similarly, the AD104 powering the RTX 4070 has seen good sales, but the maxed out chip for the RTX 4070 Ti has missed the mark against the RX 7900 XT with a similar price. As a result, NVIDIA has introduced the AD103 in the desktop segment, which is a high CUDA core-count silicon with a mainstream memory bus width of 256-bit, in order to justify the high-end pricing. This trend will continue in the GeForce Blackwell generation with the GB203.

Similar to the AD103, NVIDIA will utilize the high SIMD power of the GB203 to power high-end mobile SKUs. The introduction of the GB205 ASIC suggests that NVIDIA's performance-segment GPU will come with a feature-set that avoids the controversy faced when attempting to create the original "RTX 4080 12 GB" using the AD104 and its narrow 192-bit memory interface.