A few months ago Qualcomm’s outspoken veep Anand Chandrasekher made a few questionable remarks about 64-bit and octa-core chips. It didn’t take long for Qualcomm to react and the company promptly moved Chandrasekher to a new post and kindly asked him to stop talking about chips.
He bashed 64-bit chips after Apple’s successful introduction of the Cyclone core and then he also blasted octa-core chips, saying they are “dumb” and pointing out Qualcomm engineers “aren’t dumb.”
Turns out they are, at least by Chandrasekher’s standards.
Fast forward to 2014 and Qualcomm now has an octa-core chip of its own. The Snapdragon 615 features eight 64-bit cores, LTE and 802.11ac. There’s a quad-core version, too – the Snapdragon 610. Both chips are based on the Cortex A53, which is interesting as the old Snapdragon 600 featured a custom Krait 300 core.
The chips use Qualcomm’s RF 360 LTE solution, which supports five modes.
Both parts come with H.265 HEVC hardware support these 28nm processors are pin compatible with the Snapdragon 410 chip was announceda few weeks ago. The 610 has slightly more advanced multimedia feature compared to 410 that supports 1080p, while the 610 can support 4K screens and multimedia including H.265 content. It also support multi SIM as well as Quad HDR 2650x1600.
The A53 is the 64-bit successor to the frugal Cortex A7. The latter is used in MediaTek octa-core parts. Snapdragon 610/615 parts feature Adreno 405 graphics and on the whole they appear to be aimed at the mid-range market.
So why the sudden about-face?
Well, Qualcomm’s octa-core bashing was clearly the official line – these parts were in development when Chandrasekher made his controversial remarks. At the time we took a closer look at MediaTek’s concept and we talked to a few techies and analysts. We concluded that octa-cores based on small and efficient cores made a lot more sense than it seemed at first glance. You can check out why here.
Furthermore, some cores in the Snapdragon 615 are “more equal” than others. Anand reports that the cores are placed in two separate clusters, allowing the chip to make the most of them. One cluster is tweaked for efficiency, while the second one is all about performance. Both can work in unison, but when the chip isn’t too busy, it can rely on the power optimized cluster. Sounds like an interesting approach – I guess we could call it little.LITTLE.
