NVIDIA GeForce RTX 50 “Blackwell” GPUs may only appear at CES 2025

If it comes to fruition, we shouldn’t have any news about the NVIDIA GeForce RTX 50 “Blackwell” GPUs at events like SIGGRAPH or gamescom – although their commercial partners may release some details about what to expect from the hardware.

See more:

Possible specifications of GeForce RTX 50 GPUs

A pair of leaks that occurred in June showed more details of what to expect from the NVIDIA GeForce RTX 50 “Blackwell” GPUs – and further raise expectations for what we might see at CES 2025.

Based on the information released, we will have:

• GB202 – 24,576 cores (512-bit G7) / 33% more cores compared to AD102;
• GB203 – 10,752 cores (256-bit G7) / 5% more cores compared to AD103;
• GB205 – 6,400 cores (192-bit G7) / 16.6% fewer cores compared to AD104;
• GB206 = 4,608 cores (128-bit G7) / same amount of cores compared to AD106;
• GB207 = 2,560 Cores (128-bit G6) / 16.6% fewer cores compared to AD107

With this, it is possible that NVIDIA GeForce RTX 50 “Blackwell” GPUs will follow the following pattern:

GB202: will be used for the manufacturer’s high-end graphics cards, such as the RTX 5090. This GPU will have 12 GPCs, each with 8 TPCs and a total of 96 TPCs. Each of these TPCs will have 2 SMs, for a total of 192 SMs. Although it has a 512-bit interface, we can see cuts depending on the configuration – with 448-bit, 384-bit, 320-bit versions and others.

GB203: This will also be a GPU for high-end graphics cards. The chip has 7 GPCs, each containing 6 TPCs for a total of 42 TPCs. It maintains 2 SMs for each TPC, for a total of 84 SMs. The GPU will have a 256-bit interface and could be the configuration of choice for the GeForce RTX 5080.

GB205: It will be aimed at the mid-tier segment, probably targeting the GeForce RTX 5070 and RTX 5070 Ti. The GPU has 5 GPCs, with 5 TPCs each and bringing a total of 25 TPCs. Just like the versions above, there are 2 SMs for each TPC and a total of 50 SMs. The chip will arrive with a 192-bit memory bus interface.

GB206: NVIDIA intends to use it for mainstream video cards, with the GPU bringing 3 GPCs with 6 TPCs each, reaching a total of 18 TPCs. As seen, with 2 SMs for each TPC, we will have 36 SMs on the 128-bit memory bus interface. It is expected to be used in the GeForce RTX 5060 and RTX 5060 Ti – with a performance jump.

GB207: Aimed at entry-level SKUs, we will likely see the GPU being used in the GeForce RTX 5060 or RTX 5050. With 2 GPCs and 5 TPCs for each, there are 10 TPCs in total and 20 SMs. It will use the 128-bit interface and may arrive with support for GDDR7 and GDDR6 – while the others will use the GDDR7 standard.

A strategic positioning

Initially, rumors about the NVIDIA GeForce RTX 50 “Blackwell” GPUs didn’t point to CES 2025. Reports claimed that we would see the RTX 5080 on the market first, with the expected RTX 5090 arriving soon after.

It was not revealed whether only the two will be announced at the event or whether the entire line will be present.

Meanwhile, AMD is also preparing its lineup of new GPUs for CES 2025, with the Radeon RX 8000 being launched with RDNA 4 architecture closer to the event.

Fonte: WCCFTECH