According to the source’s tweet, the Exynos 2500 SoC may bring a G-NPU, an S-NPU, and a new component labeled TPU. For comparison, the Exynos 2200 in the picture has only one G-NPU, and the Exynos 2400 has one G-NPU and one S-NPU.

Samsung will launch the Galaxy S25 series in early 2025. The lineup is expected to have Exynos 2500 and Snapdragon 8 Gen 4 chipsets (from a market perspective). In order to win the trust of consumers, Samsung is constantly exploring the AI capabilities of the Exynos 2500 chip to enhance its ability to challenge Qualcomm Snapdragon 8 Gen 4.

The relevant introduction about NPU as follows:

G-NPU (General Neural Processing Unit): Designed for various machine learning tasks.

S-NPU (Specialized Neural Processing Unit): Designed for specific tasks such as image recognition.

DSP (Digital Signal Processor): A more general processor that can be used for machine learning tasks.

TPU (Tensor Processing Unit): An NPU designed specifically for machine learning tasks involving Tensors (multidimensional arrays).

The post Samsung Exynos 2500 chip will add NPU chip to build richer AI experience appeared first on TechGoing.

According to whistleblower Revegnus, Samsung is working with AMD to develop a GPU for the Exynos 2500 chip, which is based on AMD technology and specifically focused on optimization.

This new SoC is said to be designed for the Galaxy series of smartphones, suggesting that Samsung will eventually no longer rely on Qualcomm. For that to happen, though, chips must be developed that stand up to the competition. Moreover, Samsung currently lags behind TSMC in the field of foundry. However, there are rumors that mass-produced chips using the second-generation 3nm GAA process will appear in the second half of 2024. Samsung can use its advanced chip manufacturing technology to create the Exynos 2500 chip, but it is unclear which chip will be used. kind of node.

As for the optimization mentioned in the tweet, it may be related to improving power efficiency. The Exynos 2200 chip does not perform well in this regard. Although its ray tracing performance is better than that of the Qualcomm Snapdragon 8 Gen 2 chip, battery life is sometimes worse than the performance of a powerful GPU. important. However, it will be a few years before this new Samsung GPU takes shape, and before it, there is the Exynos 2400 chip to look forward to.

The post Samsung’s Exynos 2500 chip is under secret development, based on AMD Technology appeared first on TechGoing.

The report cites a reliable source, OreXda, that Samsung has essentially confirmed the cancellation of mass production of the Exynos 2300 chipset and subsequent development of prototypes, while the Exynos 2400 may suffer the same fate. Samsung will focus on working with Google and AMD to build the Exynos 2500 chipset.

The tipster also hinted that the Exynos 2500 is being developed around performance, with a 10-core CPU and a twice-redesigned Xclipse GPU with an NPU developed in collaboration with Google.

Mini-lesson.

Embedded neural network processors (NPUs) use a “data-driven parallel computing” architecture and are particularly good at processing massive amounts of multimedia data such as video and images.

The NPU is a specialized circuit in the processor that implements all the necessary control and arithmetic logic needed to execute machine learning algorithms, typically by operating on predictive models such as artificial neural networks (ANN) or random forests (RF).

Google’s Tensor and Tensor G2 chipsets are OEM’d by Samsung and offer powerful capabilities derived from AI (artificial intelligence) and machine learning. The company may work with Google to bring the best of AI technology to its processors.

The post Samsung canceled mass production of Exynos 2300/2400 chips, developing the Exynos 2500 appeared first on TechGoing.