Update: A previous version of this article included benchmarks from a 1.9GHz Snapdragon SoC that were labeled as being from the Exynos 5 Octa SoC. The text and charts have been corrected to reflect this.
Samsung officially unveiled its new flagship Galaxy S 4 smartphone on Thursday after weeks ofspeculation, leaks, and strange ad campaigns. The company’s presentation was focused mostly on the software side of the equation, with all of the hardware information rattled off in just a few minutes at the beginning of the presentation.
Despite the fact that the S 4 looks a lot like its predecessor, there’s quite a bit of new hardware under the hood. Today, we wanted to take a quick look at the chip that powers the international versions of the phone, Samsung’s new Exynos 5 Octa system-on-a-chip (SoC). We should note: the US versions of the S 4 likely won’t include this chip, but if precedent tells us anything, we will eventually get our hands on it, possibly in the form of a future Samsung tablet.
It’s a given that this chip will be faster than the Exynos 4 Quad that powered the international Galaxy S III, but the new chip’s architecture also brings a few interesting things to the table. Let’s take a look.
Eight cores (technically)
Most mentions of the Exynos 5 Octa simply say that it has eight CPU cores. This isn’t untrue—the chip actually does have eight distinct CPU cores—but not all of these cores are created equal.
The biggest issue in designing a chip for a smartphone or tablet is balancing performance and power consumption, and most modern chips attempt to do both—the chips can use multiple cores and higher clock speeds when higher performance is called for, but will typically disable cores and lower clock speeds during light or idle use. The Octa attempts to solve this problem using a CPU configuration that ARM calls “big.LITTLE.”
Big.LITTLE pairs two distinct CPU cores, one larger and faster (in this case, a Cortex-A15 running at 1.2GHz) and one that is smaller and more power-efficient (a Cortex-A7 running at 1.6GHz). These two cores support the same instruction sets and can execute all of the same code, so speed and power consumption are the main differences between them. Lighter tasks like Web browsing and e-mail checking will be executed on the power-saving Cortex-A7 cores, while more computationally intensive tasks like gaming will be sent over to the Cortex-A15s.
The core switching is controlled by a firmware layer that sits in between the software and the chip itself. Operating systems can be tweaked to better support big.LITTLE’s particular arrangement of cores, but any OS that supports power state switching for CPUs (any mainstream operating system from the last decade or so) can take advantage of big.LITTLE without any additional changes.
Different versions of this idea have existed in shipping products for some time now—the most prominent example is probably Nvidia’s Tegra chips, which include a single “companion” or “shadow” core that kicks in for light use so that the more power-hungry main CPU cores can switch off. Big.LITTLE simply takes it further, pairing each high-end CPU core with a slower one. Since the Exynos 5 Octa is the first big.LITTLE chip to ship, we don’t have much real-world evidence that one approach is superior to the other, but in both cases the concept is similar.
There are two different implementations of big.LITTLE that hardware makers can use: one in which the Cortex-A7 and A15 cores can be active at the same time (called “big.LITTLE MP” in ARM’s documentation) and one in which a Cortex-A7 core powers down when its corresponding A15 core powers up and vice versa. By all appearances, the Octa uses the latter implementation.
Samsung’s demo video for the chip has some CPU usage examples toward the end, and as long as the examples used here are representative of how the chip actually works, the A7 and A15 cores can’t both be used at the same time—the chip has eight cores, but only four of them can be active at any one time. The upshot of this is that the Exynos 5 Octa’s maximum performance will be consistent with a quad-core Cortex-A15 chip like Nvidia’s Tegra 4.
Samsung’s Exynos 5 Quad demo video. The CPU usage examples in question start at about the one minute mark.
Benchmarks aside, the relatively low clock speed of the Octa in the S 4 is a byproduct of using a chip like the A15 in a smartphone. In a tablet with more room to dissipate heat and a larger battery to compensate for the resultant bump in power usage, there should be plenty of room to ramp the clock speed up a bit. Given that the Exynos 4 Quad from the international Galaxy S III later made its way into both the Galaxy Note 10.1 and the Galaxy Note 8.0, it’s a safe bet that we’ll see some version of the Octa make its way into future Samsung tablets.
This article is from Ars Technica
