Will Samsung’s Exynos processors die?

On the one hand, Samsung will use Qualcomm Snapdragon processors for its next Galaxy smartphone. On the other hand, we hear rumors of a new chip from Samsung for 2025 with an AMD “super” GPU. Doubts about Samsung’s strategy in the middle. Who’s asking: Are Exynos chips at the end of the road?

During the Snapdragon Summit, smartphone chip giant Qualcomm confirmed a rumor that has been circulating for months: future Samsung Galaxy smartphones will be well equipped with Snapdragon chips in Europe. Micro earthquake – and the good news? – For the European market, which is only Exynos with Samsung’s in-house chips. In a strange twist, there are (apparently secret) technology sharing agreements between Qualcomm and Samsung that allow the Korean to partially rely on certain blocks of Qualcomm. The goal is for Samsung to continue offering powerful processors in smartphones. Samsung was still the first supplier of iPhone SoCs before Apple developed its own chips.

From the point of view of journalists and analysts who test the products, the switch to Qualcomm seems like good technical news. Because no matter what Samsung does, its partner’s chips are always superior in terms of both pure power and power consumption. So the Galaxy smartphones delivered to Europe were systematically a small step below those found in the US or South Korea (yes, Samsung offers terminals equipped with American chips in its national market!) the question has always been: why only Europe has to do it a little more Have a less efficient processor? Now the probable “death” of Samsung’s high-end Exynos chips is a developing question, even in part. Because the recent history and the tasks ahead are very big.

First drop in CPU cores

In the past, Samsung’s Exynos processors have incorporated “custom” ARM CPU cores.

Samsung has been developing SoCs since the dawn of smartphone history. And even custom-made hearts (custom in jargon) in a special R&D unit in Austin, Texas (USA). In November 2019, Samsung must face reality: the work presented by this team does not allow to make a difference compared to the initial plans presented by ARM. Samsung can’t offer more efficient chips than Qualcomm and often lags behind. Faced with the inability of these engineers to do better than ARM’s reference design, Samsung decides to close the division.

A few months ago, the Korean firm announced a partnership with AMD that was as surprising as it was attractive. It was then (and still!) on the back of the plans of the rising American GPU that Samsung announced that it was basing its next-generation GPUs integrated into its mobile processors. And as the months passed, AMD’s RDNA technology made people salivate more and more: we learned in the spring of 2020 that both the Xbox Series and Playstation 5 will use this RDNA architecture (consoles in the living room version “2”). ).

As 2020 and 2021 progress, Samsung’s two-year roadmap has been revealed. Until the January 18, 2022 announcement of the Exynos 2200 chip, which you’ll find in the European Galaxy S22 today. 100% ARM chip on the CPU side, but completely customized on the GPU side. For a gain we know…

AMD GPU didn’t notice (yet?).

If AMD begins to seriously threaten Nvidia in PC and controls all home consoles, its first cooperation in the field of mobile graphics processors did not give the expected “wow” effect. The Exynos 2200 powering the European Galaxy S22 really struggles to compare to the Snapdragon 8 Gen 1 and earlier Gen 1+.

According to an analyst we spoke to at the Snapdragon Summit who wished to remain anonymous, “ The fact that AMD’s GPU cores are based on high-powered cores makes the conversion to low-powered cores very difficult. “A review shared by Qualcomm engineers when asked if they could really scale the graphics power in future PC SoCs:” In addition to Oryon, our GPU will be scalable because our architecture is highly scalable*, so it’s easy for us to add cores and scale.” The x86/ARM battle proves them right: it’s easier to start with a low-power architecture and then scale up. While we now find ARM processors in PCs and supercomputers, x86 has never found a place in the mobile world.

The inability to match the Qualcomm chip, especially in 3D, is completely wrong of Samsung. For the division that designs the SoC, the division that manages the partnership with AMD, and is also responsible for manufacturing. Samsung’s 4nm was found to be less efficient than TSMC’s 4nm. So much so that Qualcomm has swapped dairy products for the “+” version and the new Snapdragon 8 Gen 2.

(*: scalable means “wide”, “scalable” and refers to a technology that very well supports increasing power by adding computing units)

Exynos fails to convince either Samsung or its customers

During a complete overhaul of its SoC design division in 2019, Samsung also announced that it wants to change its business model. In addition to designing chips for its own terminals, Samsung has said it wants to sell its Exynos chips to other manufacturers. A strategy that seems to have caught on. If the Koreans delivered entry-level and mid-range chips to Chinese manufacturers, no transplants were carried out. It’s even the other way around: Exynos’ market share continues to decline.

Here, not only Qualcomm hurts Samsung, but also MediaTek. The Taiwanese continue to expand their ranks and gain momentum. Using its status as an independent chip designer, MediaTek produces chips not only for smartphones, but also for internet boxes and TVs. Markets that drive R&D to a very high level. It allows you to sell chips not only for the entry level, as in the past, but also for high-end devices of certain brands. There are certainly those who prefer to get good chips from an independent player rather than a competitor!

Rotational management that hinders consistency

Samsung didn’t become the number 1 smartphone in the world by accident, and it has some great engineers in its ranks. With such a global weight in electronics—Samsung has just resumed PC sales in many European countries, including France—why is the Korean struggling so much to design a high-end chip? Samsung is paying for the management system here, according to several analysts who spoke on condition of anonymity. Many senior positions actually change every two years, and sometimes every year for other profiles. All this without much asking for their opinion. If preventing his troops from “rising” can be a good thing, this mobility also has negative effects.

Because in the semiconductor world, the development cycle of a processor is about three years and requires about 10,000 people. A huge salary that needs to be managed and directed in the right direction for thirty-six months. A huge challenge that only a few rare companies in the world can face… and it doesn’t match up with games of musical chairs that are too close together.

Samsung seems to be nailing the “return” with a “super AMD GPU” until 2025. Counting three years to design the SoC from scratch and integrating a more efficient (and optimized!) graphics part, this horizon seems possible. But Samsung is keen to stay the course and avoid changing its mind every four mornings. In the pain of ending up, one day, who knows, he has to close his department. And settle for Qualcomm and MediaTek chips.

Leave a Reply

Your email address will not be published. Required fields are marked *