Das selbstentwickelte Flagschiff-Chip von Samsung Exynos steht kurz vor einem Aufbruch, aber besteht noch Hoffnung?

Although the Samsung Exynos flagship mobile SoCs in the past few years have received almost universally poor reviews in the market, the only concern this time no longer seems to be the performance of the chip.

Recently, more and more leaks about the Exynos 2600 have emerged on the Internet. Through relevant data, it's not hard to find that Samsung may have solved many of the technical problems that previously plagued its "self-developed flagship chips".

CPU, GPU, ISP... Samsung's self-developed flagship chips have made great progress

The earliest detailed leak came from Samsung's "fan site" sammyguru. According to relevant information, the Exynos 2600 will be built on Samsung's 2nm GAA process, featuring a 10-core (1+3+6) "all-large-core" CPU design (using Arm's latest C1 series architecture), and will be equipped with an Xclipse-960 GPU based on AMD's RDNA4 architecture with an 8CU specification.

Sammyguru shared five alleged CPU theoretical performance test results from Samsung's internal sources. Arranging these test scores chronologically, it's clear that the closer to the mass-production version, the higher the peak CPU clock frequency of the Exynos 2600, indicating that Samsung may have solved the leakage problem of the 2nm process.

Moreover, these benchmark data show that the CPU performance of the latest test version of the Exynos 2600 has exceeded that of Qualcomm's Snapdragon 8 Gen 5 Extreme and MediaTek's Dimensity 9500. With a single-core score of 3959, it should only be second to Apple's A19 Pro in the current industry.

GeekBench information of the early test version of the Exynos 2600, showing support for the latest SME2 architecture

Meanwhile, leaks from the same source also revealed some GPU benchmark scores of the Exynos 2600. Allegedly, thanks to the underlying architecture of RDNA 4, its GPU FP32 computing power can reach up to 6 TFLOPS, while the peak frequency is only around 1350 - 1400 MHz.

This means that on the one hand, Samsung's "Xclipse 960" GPU surpasses all current mobile competitors in terms of theoretical peak performance. On the other hand, because of its lower peak frequency, compared with the high-frequency designs of MediaTek and Apple, in theory, Samsung's GPU this time is less likely to be affected by frequency reduction during long gaming sessions.

In addition, another leak indicates that the Exynos 2600 has also made significant improvements in the design of its ISP (Image Signal Processor) compared to the past. Allegedly, its ISP can support a single camera with up to 320 million pixels or drive three 108-million-pixel sensors simultaneously. Moreover, its HDR engine can process 14-bit high-dynamic RAW format images through a 5-frame synthesis method.

In terms of video shooting capabilities, the Exynos 2600 is said to support 8K 60FPS or 4K 120FPS in HDR10+ format, which means its SDR shooting capability can reach up to 4K 240FPS. In terms of design concept, Samsung's ISP is weaker than Qualcomm's latest (18-bit) solution in handling ultra-high dynamic range (high color depth), but it may have higher internal computing power and image throughput, so it may have its own advantages in anti-shake and video shooting.

The theoretical performance is perfect, and the actual device design also seems to have a breakthrough

It should be noted that according to the description of relevant leaks, not all of the above theoretical performance test results are from actual engineering prototypes. When the chip is in the early development stage, manufacturers sometimes use larger test platforms and "simulate" the mobile phone usage scenario by limiting power supply and heat dissipation.

This inevitably makes us worried. Although the theoretical performance of the Exynos 2600 looks excellent, can it maintain such performance when installed in a mobile phone?

On November 4, 2025, Kim Dae-Woo, a senior vice president of Samsung Electronics, voluntarily disclosed the cooling design details of the Exynos 2600 in mobile phones during a relevant event.

According to him, the Exynos 2600 uses HPB (Heat Path Block) packaging technology, which directly combines the metal heat conduction layer with the main SoC and memory chips in a packaging form, improving the overall heat dissipation performance by 30% compared to the previous generation.

Allegedly, an engineering prototype of the cancelled Galaxy S26+ appeared on a second-hand platform

In addition, leaks from other sources show that Samsung may have developed an AI imaging technology called "neural composition" for its own mobile phones equipped with the Exynos 2600. The key to this technology is to achieve true "predictive imaging" of the camera through a high-speed neural network (especially the ultra-high bandwidth communication between the ISP and NPU). That is, before pressing the shutter, the AI has already predicted that "the current frame" may be the best composition and has performed a complete imaging calculation process such as multi-frame synthesis and noise reduction on the previous and subsequent frames in advance.

In this way, on the one hand, users will feel that they can get a photo quickly without waiting after pressing the shutter. On the other hand, it is said that this algorithm design that "front-loads" the complete photo calculation process will also subvert Samsung's existing color science, enabling them to narrow the gap in competition with other manufacturers or even gain a unique advantage.

It seems promising, but the real problem lies in "confidence"

If we assume that all the current leaks and official preheating information are "true", then both Samsung's 2nm GAA process itself and the mass-produced Exynos 2600 based on it seem to have made great progress in terms of function and performance this time, giving people hope that Samsung can return to the competition in the flagship mobile SoC market.

Then the only remaining problem is whether Samsung has enough confidence in its new flagship chip this time and whether it dares to break free from the shackles of the past?

Friends familiar with the history of Samsung Exynos may remember that its most "glorious" moment was the Exynos 7420 used in the Galaxy S6 series back then. At that time, it "happened" that its competitor, the Snapdragon 810, made a mistake in the process selection, so it won a lot of praise from consumers in the global market with its significantly better performance and energy efficiency.

According to our hands-on tests at IT-3eLife, Samsung also pays great attention to the subsequent optimization of its old chips

But even before the Exynos 7420 became popular, Samsung may have lost some confidence in its Exynos flagship chips. The evidence is that many of Samsung's flagship mobile phones sold in China at that time (such as the Galaxy Note3 and Galaxy S5) already used the Qualcomm Snapdragon platform instead of its own SoC solution. After the Exynos 7420, Samsung has always insisted on using the Qualcomm Snapdragon platform to build most of its Galaxy flagship models in the two major markets of China and the United States, rarely launching products using its own chips.

Interestingly, on November 7, 2025, when an overseas media reported on Qualcomm's business in the fourth fiscal quarter of 2025, it specifically mentioned that Cristiano Amon, the president and CEO of Qualcomm, praised the "growing partnership between Qualcomm and Samsung". According to the report, about 75% of Samsung's Galaxy devices now use Qualcomm chips, "showing a significant increase compared to about half in the early days".

It has to be said that it's almost impossible that Qualcomm's reiteration of its "partnership" with Samsung at this moment when Samsung has just disclosed its new self-developed flagship chip has no deeper meaning. And as we all know, the next-generation Galaxy S26 series will most likely still use a "mixed-use by region" mechanism of Qualcomm Snapdragon and Samsung Exynos.

In this way, a very real problem arises. If the Exynos 2600 is truly as powerful as the current leaks claim, significantly outperforming the contemporary Snapdragon flagship chips in performance, will Samsung really make the "Exynos version" of the S26 series models faster or have better imaging performance than the "Snapdragon version" in the final terminal products? Based on our long-term hands-on experience at IT-3eLife, we are deeply skeptical.

This article is from the WeChat official account "IT-3eLife" (ID: IT-3eLife), author: IT-3eLife. Republished by 36Kr with permission.