Artificial intelligence has been a hot topic through 2017 and 2018 with multiple vendors throwing their hats into the proverbial ring, targeting everything from mobile devices to specialized accelerators for HPC applications. Now, MediaTek is hopping on the bandwagon itself, with its new P60 SoC design.
MediaTek typically uses an “X” for its highest-end flagship SoCs and a “P” for its midrange SoCs. (X stands for eXtreme, “P” for Premium). The new P60 supplants the previous P30 at the top of the Premium stack and the dramatic model number jump is, in this case, completely justified.
First, on the CPU side, the new P60 combines four Cortex-A72 cores and four Cortex-A53 cores. The P30, in contrast, had just eight Cortex-A53 cores, and while clock speeds haven’t been disclosed, there’s no doubt which configuration would win any benchmark test you care to name. A quad-core cluster based on Cortex-A72 may not literally be the latest and greatest, but it offers much better performance than an eight-core A53 cluster.
MediaTek has paired its new big.Little configuration with a Mali-G72MP3 graphics core. Again, this is a significant increase compared with previous SoCs — the P30 used a Mali-G71MP2 GPU, which means we’re looking at 1.5x more processing power and then improved clock speeds and efficiencies on top of that.
As this slide shows, the Mali-G72 is substantially faster at FP16 workloads, which are commonly used these days for — you guessed it — AI calculations. The G72 also featured a larger L1 cache and a larger tile memory buffer, both of which help fuel its improved performance overall.
With its new NeuroPilot, MediaTek is offering a framework that can translate code written for other frameworks like TensorFlow, TF Lite, and Caffe 2, theoretically helping development by boosting cross-usage between the various platforms. There’s also a hardware AI processor baked into the platform with support for inferencing workloads, though MediaTek hasn’t shared much in the way of specifics on its overall design.
Between the improved CPU / GPU performance and the AI capabilities, we’ll be curious to see what kind of uptake the SoC wins in upper-midrange smartphones this year. It makes more sense for MediaTek to focus on the midrange SKUs where robust competition still exists, as opposed to top-end sockets that major chip vendors increasingly build their own silicon to corner. This is particularly true in the US, where high-end smartphone sales are almost entirely owned by two companies: Samsung and Apple.
Google’s AI-Focused Tensor Processing Units Now Available in Beta
Google is ready to open up its Cloud TPU platform to developers and researchers looking to test machine learning workloads — and it's got a new, more powerful Cloud TPU design than the chips we've previously discussed.
ARM Announces Project Trillium, a New Dedicated AI Processing Family
ARM has launched a new series of AI-focused processors, with the goal of improving machine learning in mobile devices.
MIT Neural Network Accelerates MRI Image Processing by 1,000 Times
It can take hours for a computer to match all the locations in a 3D map, but researchers from MIT have developed an algorithm that could cut that time to less than a second.