Imagination Technologies Group plc (Kings Langley, England) has confirmed that its PowerVR SGX544MP graphics processing core has the design win inside Samsung's Exynos 5410 "Octa" chip. The Octa is the latest 4.4 big-little application processor from Samsung, announced at the Consumer Electronics Show with some fanfare but little detail. It is also expected to be the application processor at the heart of the Galaxy S IV smartphone, expected to be announced by Samsung on Thursday (March 14).
As Mali cores licensed from ARM Holdings plc (Cambridge, England) had been used in previous Exynos processors this could be construed as a rebuff for ARM and its aspirations in the GPU core licensing market.
An ARM executive had more or less confirmed the loss to me a couple of weeks back when she said that Samsung has always used a variety of graphics IP in its products and the continued use of PowerVR is to be expected.
It could be that Samsung is using a two-supplier, treat-em-mean to keep-em-keen strategy. But this is a second strike against ARM. ST-Ericsson, and STMicroelectronics before it, had extensive experience of working with Imagination. ST-Ericsson turned to Mali GPUs for a few processor designs before switching back to PowerVR SGX544 for the NovaThor series of modem-cum-application processors, which ST-Ericsson calls ModAps.
ST-Ericsson has also flagged it will use the next-generation PowerVR Series 6, known as 'Rogue' in the Nova A9600 processor, which would likely form the basis of a 96XX series of NovaThor ModAps.
So while ARM emphasizes the large number of its GPU design wins in Southeast Asia, it looks like Imagination is successfully regaining ground at the high-end of the market.
The key to medium- to long-term success here will be whether Apple – which licenses PowerVR graphics as well as ARM processor architecture – decides to design its own GPUs as part of a soup-to-nuts strategy, or whether Qualcomm – which deploys its own GPU cores – opts to save on development costs and begin licensing graphics in.
Right now both of those strategies seem to revolve around Imagination in graphics, rather than ARM.
ARM's argument and hope is that, over time, the close compatibility and interworking that can be created between ARM processors and ARM graphics cores will manifest as a performance advantage and bring the market towards it. Meanwhile Imagination will start to try and tune the combination of its MIPS-PowerVR offering.
Right now system-chip developers seem to prefer not having all their eggs in one basket.
The left-field alternative is whether it would be possible for a startup or an established player to create a unified multicore instruction set architecture that can cope efficiently with both graphics rendering, GPU-computer and running C programs.
"It could be that Samsung is using a two-supplier, treat-em-mean to keep-em-keen strategy. "
It also could be whatever-apple-do-we-must-also-do strategy. Last week Samsung invested in sharp to get access to IGZO panels, in-spite of having own display development.
'ARM's argument and hope is that, over time, the close compatibility and interworking that can be created between ARM processors and ARM graphics cores will manifest as a performance advantage and bring the market towards it. '
That is complete marketing bull, I'm surprised anyone let's ARM get away with that.
I don't think it is bull.
It makes sense to me that, in an era of GPU-compute, if you design and optimize the CPU and the GPU and all the pathways between them you can produce some advantages over a system with a CPU from one source and a GPU from another.
In time I expect Imagination to be making the same argument for MIPS-PowerVR combinations.
However, I also think that some startups that are working on unified CPU-GPU instruction set architectures might be able to produce something of interest.
I think it might.
I reckon it would depend on how effectively ARM could seperate the engineering arguments from the economic ones.
IF the engineering arguments are provably valid AND there is little or no discounting for companies taking both Cortex and Mali AND ARM only achieves moderate success against imgtec....then ARM will avoid problems.
If the engineering benefits appear slight, the discounting on royalty rates is heavy and ARM starts to take major market share ... well, if I was imgtec's lawyers I would be looking at anti-competitive behavior law around the world.
But that would appear to be a ways off yet.