Samsung Exynos has a Problem

For years, Samsung has been integrating two different types of SoCs into their phones now: Qualcomm’s Snapdragon and their own Exynos chips have both been used to power their devices. Snapdragon is used in the US and China (and South Korea sometimes), while the rest of the world gets the Exynos. So, chances are that if you are reading this on a Samsung device, it is powered by Exynos.

But why would Samsung use two different processors for the same device anyway? The easiest answer to this question would be, that different parts of the world have different needs, and it makes the supply chain much easier to handle if a custom chip that fits those needs is used instead of the same one for all regions. This is especially true since Samsung is the largest smartphone manufacturer in the world, and if Qualcomm can’t supply them with enough SoCs, this could be fatal for their market share and their image. Additionally, Samsung’s semiconductor business is very important for Samsung’s revenue (providing them with 52.21 billion dollars last year), and with Qualcomm recently moving away from Samsung as their main provider, they can make sure that at least their own Exynos chips are produced in-house. Lastly, it is likely that producing the Exynos chips is much cheaper than buying the Snapdragon ones, because Qualcomm obviously has to make some money with their products as well.

So, if there are so many valid economic reasons for Samsung to go with Qualcomm, why wouldn’t they just use Exynos everywhere? Well, Qualcomm has many important patents regarding the CMDA mobile network standard, which is used in the US and China, meaning that Samsung can’t build modems that support CMDA into their own chips. So, unless Samsung wants to have a large competitive disadvantage when it comes to connectivity, it is basically forced to use Snapdragon in those markets. Samsung also cannot sell Exynos to other OEMs, because of an agreement between Samsung and Qualcomm from 1993. In return, Samsung often gets priority access to Qualcomm’s processors, meaning that they can release a new flagship phone earlier in the year.

The real reason you are reading this article, though, is probably because you are concerned about the performance difference between the two. If two SoCs designed by two completely different teams of engineers are used in the same phone, it is highly unlikely that they will perform the same, meaning that one market gets a “better” phone than the other. In recent years, Samsung has earned a lot of criticism for their own Exynos chips, as they reportedly perform inferior to their Snapdragon counterparts. However, the very purpose of the Exynos was, to create true contender to Apple’s A-series of chips, which until today remain the most efficient Arm-based CPUs that exist.

Unsurprisingly, the company mostly responsible for this lack of competition is Qualcomm. Until mid-2013, the Snapdragon was used in almost every Android-based device, and their chips enjoyed a comfortable lead over Apple’s A-series. In late 2013, Apple surprised almost everyone by releasing the 64-Bit Apple A7 in the iPhone 5s, a product that was far superior to everything Qualcomm’s 32-Bit based chips could do at the time. Qualcomm was essentially two years behind in designing 64-Bit CPUs and have remained so ever since.

As Apple’s biggest competitor in the mobile phone segment, Samsung naturally tried to stop this trend of using inferior SoCs in their own products. Samsung’s R&D center in Austin which was found in 2010 immediately started to work on a custom CPU core that could compete with Apple’s, and the first custom Exynos chip debuted in the 2016 Galaxy S7. This solution gave Samsung a significant advantage over Qualcomm, because the Snapdragon lineup was stuck using Arms’ Cortex-A series of cores. Additionally, the Qualcomm chip in the S7 only had four cores compared to eight in the Exynos, which was therefore able to beat the Snapdragon by 12% in single- and 38% in multi-threaded workloads. While Exynos failed to match Apple in performance that year, these results are clear proof that Exynos did have a purpose.

Qualcomm still had a significant advantage over Samsung when talking about the GPU, though. Here, Snapdragon had a custom (and therefore superior) “Adreno” solution, while Samsung used off-the-shelf GPUs from Arm – a complete reversal of the CPU situation. Unfortunately, the Exynos 8995 in the Galaxy S8 used almost the exact same CPU cores as the chip in the S7, and because Qualcomm made huge improvements to their CPU and had a superior GPU, the Snapdragon 835 was a slightly better performer than its Exynos counterpart.

From there on, Exynos just continued to lose ground against Snapdragon. The gap widened with the Galaxy S9 and with the Galaxy S10, Samsung not only have a process node disadvantage by producing in-house, but only two of the eight cores were custom designed by Samsung. Shortly after, Samsung announced that their R&D center in Austin for their custom cores would be shut down, and it can be expected that this year’s Exynos 990 is the last SoC to feature them. It is important to note that even today, the custom cores outperform the Cortex-A cores by Arm. So why did Exynos fail in the end?

Power efficiency is one of the most important aspects of a mobile chip. In fact, we could just use our x86 desktop chips in our smartphones, if power wasn’t a limitation. And while the Exynos cores are able to marginally outperform the stock ARM cores, they take up significantly more die space and consume multiple times as much power. This is the reason, that for the last two generations, Samsung had to replace two of its custom cores with ARM cores from the last generation. So, while two of the cores are marginally faster, two other cores are made significantly slower to compensate for the extra energy usage of the faster cores. This gives a Snapdragon chip a multi-core advantage over an Exynos one, while the Exynos still consumes more power.

After all, it is understandable, though unfortunate that the custom CPU division in Austin was shut down. It made Exynos slower, gave devices worse battery life and sometimes even lead to overheating issues. If Samsung and Qualcomm had teamed up to build a chip with both a custom CPU and GPU, they might have been able to catch up with Apple in the end. Also, in a less power-constrained environment like Windows on Arm, Samsung’s custom cores could have really shown their full potential. As it stands now, though, from next year on every mobile chip maker will be using whatever Arm has to offer for their CPU cores, and nothing indicates they can catch up to Apple like that.

However, for the reasons I mentioned in the introduction, Samsung won’t stop making Exynos chips, they will just not include their custom cores anymore. It can be assumed that from next year, Exynos could pack a custom GPU solution from AMD, featuring the same graphics architecture that will be used in the next-gen consoles. Time will tell whether the (now very similar) Exynos or Snapdragon model of 2021 will perform better, but Samsung will have to re-gain the trust that customers have lost in their own CPUs over the last years.