Apple’s iPad 3: An Embedded Vision-Focused Critique

camera_extrusion_105

 

Last October, shortly after Apple's unveiling of the latest-generation iPhone 4S cellular handset, I published an analysis of the system's capabilities (both absolute and relative to predecessors and competitors) as an embedded vision application platform. Five days ago, for those of you who haven't already heard, Apple launched the "new iPad" (colloquially the iPad 3), with notable upgrades to both its "eyes" and "brain", so I thought I'd do the analysis all over again, this time from a tablet-centric viewpoint:

  • The iPad 2, unveiled in March of last year, included a SoC that combined a 1 Ghz dual-core, ARM Cortex-A9-based A5 CPU and dual-core Imagination Technologies PowerVR SGX543MP2 GPU. Given that Apple is loath to reveal detailed product specifications, little is known for sure about the A5X SoC in the "new iPad" until the in-progress shipments start appearing in customers' hands. But it's believed that the CPU portion of the new chip is architecturally unchanged from the A5 precursor (although it may run at a faster clock rate). Apple is, however, happy to tout the IC's "quad core" graphics capabilities, leading industry observers to interpret that the A5X embeds the "MP4" variant of the PowerVR SGX543 GPU.
  • A graphics upgrade makes sense because the "new iPad" includes a "Retina" display with 4x the pixels of its predecessor: a 2048×1536 pixel resolution, versus 1024×768 in the prior two tablet generations. The consequent improvement in displayed image quality perhaps goes without saying… note, too, that the beefier GPU will be equally beneficial in embedded vision applications that (via OpenCL, for example) leverage the graphics core as a CPU co-processor.
  • Beefier graphics require a beefier graphics frame buffer; since the A5X employs a unified-memory approach, it would therefore not be a surprise if the "new iPad" contained increased system memory as compared to the past. And indeed, leaked developer screenshots suggest that the "new iPad" embeds 1 GByte of DRAM, versus 256 MBytes in the original iPad and 512 MBytes in the iPad 2.
  • The front-facing camera in the "new iPad" is seemingly unchanged from that in the iPad 2 predecessor, capable of 0.3 Mpixel (VGA) still and video image capture. However, the rear-facing camera gets a substantial upgrade, with 5 Mpixel still and 1080p video capture capabilities (versus 0.7 Mpixel still and 720p video capture on the iPad 2). Both the second and third generation iPads' rear cameras capture 16:9 aspect ratio video, with the front cameras capturing 4:3 aspect ratio video. As with the rear-view camera in the iPhone 4S, Apple touts the third-generation iPad "iSight" camera's five-element lens, hybrid infrared filter, and f2 maximum aperture for low-light applicability. However, note that unlike its iPhone 4S sibling, the "new iPad" does not embed an LED flash unit for supplemental illumination…and the iPhone 4S rear camera resolution is also 8 Mpixel.
  • Wireless connectivity marks the other notable enhancement in the "new iPad" versus the iPad 2. LTE support (at least in the United States, with both AT&T and Verizon) substantially boosts both the theoretical maximum upstream and downstream cellular network speeds of the new tablet versus the 14.4 Mbps HSDPA/2.2 Mbps HSUPA peak capabilities of the iPad 2. Higher bandwidth will give a boost to augmented reality applications, for example, which often obtain their supplemental data from remote servers. And it'll also be of notable assistance to "cloud"-based image processing applications, by enabling the tablet to upload larger amounts of image data more quickly.

Followup: Yep, same CPU, same clock speed, twice the RAM.

Here you’ll find a wealth of practical technical insights and expert advice to help you bring AI and visual intelligence into your products without flying blind.

Contact

Address

Berkeley Design Technology, Inc.
PO Box #4446
Walnut Creek, CA 94596

Phone
Phone: +1 (925) 954-1411
Scroll to Top