Third-Generation Snapdragon: The Dual-Core Scorpion
As we’ve mentioned in the past, smartphones and tablets generally employ SoCs that integrate the processor, GPU, RAM, and several other subsystems in a single device. Since all of those components sit next to each other on the same chip, there is greater efficiency in data transfers, while reducing the amount of space consumed on the PCB.
Today’s SoC technology isn’t exclusive to each device vendor. The iPad 2 centers on Apple’s A5, while most Android-based tablets use Nvidia’s Tegra 2. And yet, both share a common RISC ISA by employing ARM’s dual-core Cortex-A9. Read Apple's iPad 2 Review: Tom's Goes Down The Tablet Rabbit Hole for a full discussion of Cortex-A8 and -A9 performance.
The APQ8060 powering HP's TouchPad is an almost-entirely different species. Despite its name, the APQ8060 is nearly the same as Qualcomm's MSM8260/MSM8660. The APQ prefix stands for Application Processor Qualcomm, and refers to a Snapdragon SoC that doesn’t include an integrated cellular modem. That critical differentiating component is part of the MSM (Mobile Station Modem)-designated Snapdragons, though. And that's why Snapdragons are a preferred choice amongst smartphone manufacturers looking for a single-chip solution.
The APQ8060 is a third-generation Snapdragon SoC. It shares a similar processor architecture with the A5 and the Tegra 2, but it's also slightly different. Instead of leveraging one of ARM’s reference designs, this processor is Qualcomm’s own creation, in use since the product's very first-generation incarnation. Dubbed Scorpion, Qualcomm claims its core is superior to the Cortex-A9 found in Apple's A5 and Nvidia's Tegra 2. Meanwhile, Nvidia claims Scorpion is closer to the Cortex-A8. The truth is somewhere in the middle.
Like the Cortex-A8, Scorpion has a dual-issue in-order architecture (some things can be done out-of-order), but that's where the similarities end. Scorpion is based on the same ARMv7 architecture found in the Cortex-A8 and -A9. However, according to the developer of Linpack for Android, Qualcomm’s design is supported by an array of 600 MHz DSP and accelerator cores for baseband and video processing that offer 128-bit single instruction, multiple data (SIMD) functionality. This means that it can process SIMD instruction in 128-bit wide chunks, whereas the Cortex-A8 and –A9 are restricted to 64-bit. The Scorpion also has a deeper pipeline (though how many stages is unknown), which includes VFPv3 (Vector Floating Point v3) commands not normally piped in Cortex-A8. Though, the new Cortex-A9 does implement a pipeline for VFPv3 commands. As Linpack is a measure of floating-point performance, Snapdragon-based devices will typically outclass their Cortex-A8 and -A9-based peers.
However, the real win for Qualcomm is actually in the power consumption department. The company claims 2.1 Dhrystone MIPS/MHz/core, whereas the Cortex-A8 is quoted at 2.0 DMIPS/MHz/core. Purportedly, the difference is that getting Cortex-A8 to its top speed depends on TSMC’s general-purpose fab process, whereas Scorpion uses low-power silicon. According to Berkley Design Technologies, the Scorpion consumes roughly 200 mW at 600 MHz (including leakage current, which is negligible in the low-power process). Compare that to the Cortex-A8, which consumes about 350 mW at 600 MHz manufactured at 65 nm.