Have you ever wondered what the differences are between the much talked about Tegra 3 and Qualcomm S4 processors? Are you weighing up a smartphone purchase and wondering whether the extra cores of a quad-core smartphone make it more powerful than a dual-core handset? Keep reading as Sean Fowler explains the differences in detail in this guest post for the Clove Blog and also considers how the aforementioned line up next to the Exynos 4 Quad of the Samsung Galaxy S III.
In the PC market we had the MHz race and then the cores race. Now things seem to have settled down a bit and we can get on with twittering, facebooking, browsing the internet and all the other things we really really needed all that power for.
Given the immaturity of the mainstream smartphone market it’s not surprising that mobile CPUs are still stuck in that mindset. There’s a lot of talk about MHz and cores without taking account of the fundamentally different architectures, and also of how much power we actually need.
This has come up repeatedly in the comparison between Tegra 3 and the Qualcomm S4 (Krait). Tegra 3 has 4 cores and the S4 only has 2, so Tegra 3 must be better, right?
Well if it was right I’d stop typing right here…
Nope, still going, and that’s because in general the S4 is significantly faster than Tegra 3, largely because the S4 has a far more advanced architecture. Tegra 3 uses 4 pretty standard Cortex A9 cores which have been kicking around for a while now, they’re not that much different to those in Tegra 2 and in the vast majority of workloads the extra two cores won’t be used.
In contrast Qualcomm has designed its own architecture based on the ARM instruction set and has made it far more powerful. The S4 is actually more comparable to the Cortex A15 cores that everyone else will be switching to later this year or in 2013.
The S4 features an older GPU, which was necessary to get it released soon enough to steal a march on its competitors. Consequently the Tegra 3 does have an advantage in heavily-threaded games, but even then it’s typically only around 10-15%. When the S4 is paired with faster GPUs later this year Tegra 3 will be left choking on its beautifully-rendered silicon dust.
The S4 has another major advantage over the Tegra 3; it’s based on a 28nm process rather than the Tegra’s 40nm process. This means it’s far more power efficient and gives a battery life that beats pretty much anything else out there. Tegra 3 isn’t bad, largely thanks to its fifth companion core (although that only helps while the phone isn’t doing much), but S4 is better.
You couldn’t ask for a better juxtaposition of the S4 and Tegra 3 than HTC’s One-X phone. The US version uses the S4 because it has the US version of LTE (4G) built-in; the non-US version uses Tegra 3. Here’s an excellent comparison of their performance and battery life.
If I was going to choose between the two models I’d go for the S4 version; only if I was heavily into 3D gaming would I opt for Tegra 3.
Of course I’m not going to choose between them; I’m going to buy a Samsung Galaxy S3. So where does this fit in with the S4 and Tegra 3? Well it uses Samsung’s Exynos 4 Quad CPU, which is similar to the Exynos 4 Dual in the Galaxy S2 but has twice the number of cores and is built on a 32nm HKMG process rather than 45nm. The GPU appears to be the same as in the Galaxy S2 but clocked far higher. So it’s not really a new architecture but it is built on a new, more efficient process.
Samsung claims that its 32nm HKMG process is 40% more power efficient that the 45nm process used in the Galaxy S2; we’ve yet to see how that compares to the S4. As for speed, the two versions of the One-X were briefly the fastest smartphones on the planet but early benchmarks are showing the Galaxy S3 doing a Flo-Jo on them.
So do I need all of this power in a phone? Of course not, but I want it; it’s a geek thing. And if you’re reading this then you probably want it too. May 30th will be a good day; just make sure you’re not ahead of me in the queue…