Qualcomm’s Next-Gen Snapdragon Chip to Feature New Adreno GPU With Matrix Cores, Neural Fusion

Qualcomm is gearing up to introduce its next-generation flagship mobile processor at the Snapdragon Summit in Hawaii later this month. Ahead of the announcement, the US-based chipmaker has detailed the GPU upgrades that its chipset will have. The chip, widely referred to as the Snapdragon 8 Elite Gen 6, is confirmed to feature a new Adreno GPU featuring Adreno Matrix Cores for AI workloads, alongside a new rendering technology called Adreno Neural Fusion.

Qualcomm Details New Adreno GPU Upgrades

In a blog post, Qualcomm said that the most notable upgrade on its next-generation flagship chip will be the addition of Adreno Matrix Cores to the GPU. These are said to be dedicated to running AI workloads directly within the graphics pipeline. As per the company, keeping these workloads on the GPU will result in reduced latency and power consumption compared with moving AI processing between different parts of the platform.

The Adreno Matrix Cores are designed to power Adreno Neural Fusion, which is Qualcomm's new AI-driven rendering technology. It is claimed to combine neural processing, AI super resolution and frame generation into a unified graphics pipeline. Qualcomm says the technology is designed to address common issues associated with upscaling, including shimmering, ghosting and loss of fine detail.

This indicates Adreno Neural Fusion could be Qualcomm's take on Nvidia's Deep Learning Super Sampling (DLSS) technology, which also renders images at a lower resolution and then upscales them to a higher resolution using AI to deliver improved frame rates.

The new Adreno GPU is clocked at 1.45GHz, compared with 1.2GHz on the Snapdragon 8 Elite Gen 5. Qualcomm claims a 12 percent improvement in efficiency for the new GPU architecture.

The chipmaker also confirmed that its next-generation flagship Adreno GPU uses a three-slice architecture that divides shader cores and other graphics hardware into separate sections. Qualcomm claims this allows the platform to manage power and performance more granularly depending on the workload. This architecture will be paired with 18MB of Adreno High Performance Memory (HPM), which is integrated into the GPU subsystem.

The company says the dedicated HPM is designed to keep tile-based rendering data, frame buffers and compute workloads closer to the GPU, effectively resulting in lower latency and improve efficiency during graphics-intensive workloads. For developers, Adreno Neural Fusion will be supported across major game engines, including Unity and Unreal Engine. The technology is said to be commercially available and can be integrated into existing development workflows.

While Qualcomm did not explicitly name the upcoming SoC, previous leaks suggest that it will launch two flagship SoCs this year. The chipset with model number SM8950 could be branded as the Snapdragon 8 Elite Gen 6, while the one bearing the model number SM8975 may be marketed as Snapdragon 8 Elite Extreme Gen 6.

We can expect more details about the next-generation Snapdragon processors to be revealed at the 2026 Snapdragon Summit, which is scheduled to take place from September 22 to September 24 in Maui, Hawaii.