Sony’s PlayStation Vita: Who’s Behind The CPU Incognita?

After taking some time off to wind down my EDN editorial responsibilities and sow some next-step employment seeds, it’s time to revive the blog(s). The past few weeks have been jam-packed with news-generating conferences; Computex, E3, Qualcomm’s Uplinq, and Apple’s WWDC (Worldwide Developers Conference), to name a few. As such, these first few next planned posts will have a common CPU-centric theme, acting (among other things) as addendums to my most recent EDN cover story.

As expected, Sony formally unveiled the PlayStation Vita mobile gaming console (formerly known by its PlayStation NGP project moniker) at E3. It’s the fourth iteration of the Nintendo-competitive handheld PlayStation line, following in the footsteps of the original PSP, the thinner and lighter PSP 2000 and 3000, along with the underwhelming PSP Go. And preliminary reviews based on prototype Vita hardware and software have been generally upbeat. Among other things, Sony seems to have atypically priced Vita at an aggressively low point ($249 for the Wi-Fi-only version, $50 more for a variant with added AT&T 3G cellular data support), increasing the probability of broad market acceptance…although profitability is yet to be determined.

However, after perusing the E3 coverage (I wasn’t personally at the show), I still have a couple of fundamental questions. As background, prior-generation PlayStation Portable-series products were based on MIPS CPU cores. The PS Vita, however, migrates to a quad-core ARM Cortex-A9 SoC. This shift is conceptually analogous to the ones experienced by the Xbox 360 (which migrated from the x86-based Xbox to a multi-core PowerPC approach) and the PlayStation 3 (which similarly moved to the PowerPC-derived Cell CPU, from the MIPS-based ‘Emotion Engine’).

As owners of either living room console are already aware, backwards-compatibility has been a hit-and-miss affair. Microsoft ended up substantially undershooting its initial Xbox title support promises, likely due to both emulation technical challenges and the desire to up-sell Xbox 360 owners on newer game title revisions. And whereas initial PS3 hardware versions contained both dedicated ‘Emotion Engine’ CPU and ‘Graphics Synthesizer’ GPU resources, cost-focused Sony first stripped away the CPU (switching to less-than-comprehensive software emulation on Cell) and then the GPU (completely eliminating PS2 backwards-compatibility in the process).

Sony’s PlayStation Vita seemingly takes a different tack, befitting the emerging era of downloadable game content versus traditional physical media distribution. The unit discards support for both the UMD optical discs and Media Stick flash memory modules used by prior PSP models, transitioning instead to yet another Sony-proprietary flash memory module, the NVG Card. However, as the product’s Wikipedia entry notes:

A main feature is full backwards-compatibility with all PlayStation Portable games that were digitally released on the PlayStation Network, via the PlayStation Store. The device’s dual analog sticks will be supported in only select games. The system will also upscale and smooth the graphics via a software emulator.

My suspicion is that, instead of leveraging MIPS-on-ARM emulation, Sony will create ARM-native versions of previously MIPS-only downloadable content. When a consumer requests a re-download of a previously purchased title, the Sony PlayStation Store server will detect that a Vita is the destination device and supply an ARM (versus MIPS) compiled-code variant.

Now, back to those earlier-mentioned fundamental questions of mine…First off, who designed the SoC containing (among other things) the CPU and GPU? Sony is listed on ARM’s site as a licensee of ‘classic ARM’ cores, specifically the ARM9…it’s conceivable that the license could have privately been extended to the ARM Cortex-A9, in which case Sony’s Semiconductor Division could have internally tackled the design. But when I heard the ‘quad-core ARM Cortex-A9’ specifications, my first thought was that this was the latest iteration of Sony’s partnership with Nvidia, specifically harnessing the latter company’s ‘Kal-El’ Tegra family processor design.

Sony had integrated Nvidia GeForce 7800-based technology in its RSX Reality Synthesizer IC for the PlayStation 3, relatively late in the system design after initially (or so say the rumours) initially believing that the Cell microprocessor’s SPEs (synergistic processing elements) were singlehandedly sufficient to handle the graphics task. And more recently, Sony and Nvidia had partnered to bring support for PlayStation game titles to Tegra 2 (dual-core ARM Cortex-A9)-based system designs. But my theory had one notable flaw; the system’s graphics core. Nvidia’s GPU in the Tegra line is proprietary, yet the PlayStation Vita supposedly harnesses a quad-core variant of Imagination Technologies’ SGX543MP PowerVR graphics engine.

The same graphics distinction, if true, disqualifies two other quad-core ARM candidates, Qualcomm and Marvell. As the “Graphics, the other key distinction for ARM-based SOCs” sidebar to my early-April EDN cover story noted, Qualcomm’s SoCs leverage proprietary Adreno graphics technology sourced from the company’s 2009 acquisition of ATI Technologies’ Xilleon handheld-graphics group, Meanwhile, Marvell partners with Vivante to obtain its graphics technology. Any of these three companies could have done a Sony-proprietary design leveraging a PowerVR graphics core, but it’s doubtful. And by virtue of the latter two suppliers’ ARM architecture licenses and consequent proprietary CPU designs, neither vendor’s SoC could strictly speaking be called a ‘Cortex-A9’ design, either.

The final feasible candidate is Texas Instruments, who as my article notes, has historically stuck to a predictable product path blending conventional ARM CPU cores and PowerVR graphics cores, in some cases augmenting them with TI-proprietary DSP technology. However, I’m not aware of any planned quad-core Cortex-A9 products on TI’s road map; to the best of my knowledge, the company will go straight from dual-core Cortex-A9 to dual-core (and beyond) Cortex-A15 aka ‘Eagle’. I guess we’ll have to wait for the teardown wizards at iFixit and Chipworks to dissect a Vita and the ICs inside it to learn more about the mysterious SoC’s source.

My other big-picture question concerns the PlayStation Vita’s likelihood (or not) of market success. Whereas Nintendo strives to differentiate the 3DS handheld from increasingly powerful tablets and cellular handsets via dubious-value auto-stereoscopic 3-D capabilities, Sony seems determined to apply substantial amounts of silicon processing potential to the competitive-differentiation problem. Yet, the company seems to be hedging its bets. Consider, for example, the earlier mentioned PlayStation content support across a diversity of Nvidia Tegra 2-based hardware.

Consider, too, the Qualcomm-based Sony Ericsson Xperia Play mobile phone, Android-based and capable of running PlayStation Suite content. Granted, the Xperia Play is not explicitly branded as a PlayStation device, yet a Sony Ericsson executive wasn’t reticent to tout its capabilities two weeks ago at the Uplinq conference, nor was he shy about making clear that it was the first in a series of such products from the company.