Graphics on ARM: Competition, And Eye Candy Charm

Before transitioning to other topics next week, I’ve got a few more thoughts to pass along on the ARM processor ecosystem, specifically with respect to graphics (which, you may have noticed, were a key aspect of the Nexus One’s charm). The words ‘ARM’ and ‘PowerVR‘, the latter the brand name of graphics and video processing cores developed and licensed by Imagination Technologies, are frequently linked in a synergistic fashion both in ARM licensees’ and implementers’ designs. As such, forgive yourself if you found yourself feeling surprised when you read the words ‘Graphics…ARM…Competition’ in the above headline.

It was no surprise to me, for example, to recently receive confirmation that Apple’s coming-soon iPad uses a PowerVR core, given the company’s desire for the platform to retain compatibility with iPhone and iPod touch apps, and in both native resolution and full-screen up-scaled fashions. And PowerVR’s reach even extends beyond ARM. Paradoxically, Intel’s chipsets for the ARM-competing Atom processor line implement Imagination Technologies-developed graphics cores, whereas Intel’s higher-end chipsets for notebooks and desktops rely on the company’s own tile-based architecture which dates back to a partnership with Real3D and the resultant 1998-era i740 GPU. But Imagination Technologies rapidly innovates and enhances its offerings, most recently reflected in January’s CES new-product unveilings. How Intel’s licensed-versus-in-house tug-of-war will play out in the era of Sandy Bridge and beyond is of great interest to me.

But PowerVR’s definitely not the only ARM graphics game in town. One other candidate, ironically, is ARM itself. Due in no small part to two key acquisitions, of graphics provider Falanx Microsystems in mid-2006 and of video processing core developer Logipard AB in early 2009, ARM now provides the Mali multimedia core line to its licensees. Widespread rumor that I’ve heard on numerous occasions suggests that ARM is offering Mali at fire sale prices (translation: free, at least sometimes) as a means of boosting market share. Then again, such discounts may be a particular necessity when ARM is raising the license fees for its CPU cores…

Most recently, Samsung announced its Mali adoption in mid-February. But although tile-based approaches make efficient use of memory and bandwidth capabilities (as I first wrote about way back in April of 1999), thereby particularly enhancing their appeal in scarce-resource non-PC designs, application and API compatibility is a challenge. It remains to be seen how closely the Mali upstart can feature set-pace, far from exceed, the PowerVR standards-bearer.

The next challenger is Qualcomm. The path here is a bit more twisted. ATI Technologies acquired two tile-based graphics startups, ArtX in April of 2000 and BitBoys Oy in May of 2006, as ATI assembled and expanded its Imageon handheld graphics suite. One month later, AMD announced its intention to buy ATI. And subsequent to the transaction close in October of 2006, AMD sold the Imageon product line and other resources to Qualcomm in January of last year. AMD had also sold the ATI-developed Xilleon video processor business to Broadcom in October of 2008; Broadcom previously obtained tile-based graphics expertise via its acquisition of Stellar Semiconductor in March of 2000.

Next is a small Silicon Valley startup named Vivante. I know little about them aside from the fact that their cores were used by Marvell in the Armada chips that the company unveiled last fall. And next is Nvidia. I’ve heard (and am not surprised to hear), but have not yet received definitive confirmation, that the company’s Tegra line employs internally developed graphics technology. What I am somewhat surprised to hear is that it’s supposedly of a traditional-architecture nature, not tile-based.

I’ll conclude the commentary with a few thoughts on another ARM licensee that was making some buzz around the CES 2009 timeframe (and a bit less bus around the CES 2010 timeframe), that being Creative Technology and its ZiiLabs subsidiary. I don’t know what graphics tech ZiiLabs uses, but that’s not my point in bringing them up. Here’s another twisted historical path; Creative bought 3Dlabs in June of 2002, back when the latter company’s primary business thrust was workstation graphics chips and boards based on them.

In February of 2006, the company decided to bail out of "shrinking professional workstation graphics market" and "refocus its 3D graphics business on the portable handheld device market". Fair enough. But in November of that same year, the 3DLabs subsidiary announced its intention to re-spin out from Creative into an independent company. The disconnection was never concluded, however, and the subsidiary was rebranded as ZiiLabs a bit over a year ago.

Back in April of 2008, when Apple first announced the purchase of P.A. Semiconductor, I wrote (in part):

I’m not surprised that Apple’s ‘cutting out the middle man’ and bringing IC design in-house. Microsoft made the same move in the Xbox-to-Xbox 360 generation transition, and it has both cost and control benefits in high-volume situations.

Creative Technology is, like Apple, a provider of (among other things) portable multimedia players. So conceptually, bringing a semiconductor supplier in-house might make sense. But the key words in the above quote are ‘high-volume situations". Creative was only a serious competitor to Apple in the iPod’s early days. It’s now a shadow of its former self, with scant market share. As such, ZiiLabs is now striving to sell its chips on the open market, including to Creative’s competitors.

So I ask you, dear readers, if you were such a potential purchaser, would you do the deal and why? If I were in your shoes, not only would I be worried that Creative might stiff my supply as a competitive move, I might be motivated to seek my silicon fortunes elsewhere (thereby fiscally hurting Creative) as a competitive move. ZiiLabs seems to have decently featured chips:

although I don’t know how robust (or not) their development tools suite is. But the whole business strategy seems a bit wacky (albeit understandable given the corporate situation) to me.

I welcome your thoughts, and wish you a happy weekend.