The story is a short one, the current Qualcomm GPU is fine for Android and related mobile OSes, but the bloat of Windows on ARM (WARM) requires a higher DX level than Adreno to make things tolerable. Tolerable to the MS that is, the rest of the world doesn’t actually need all the leakage from the otherwise useless transistors. Since MS has never, and shows no signs of ever making clean code, Qualcomm had two options for WARM, a ‘modern’ GPU or never get MS’s seal of approval.
In case you hadn’t realized it yet, Qualcomm didn’t want to give up on the tens of unit sales, after returns, that WARM will bring in, so they didn’t take the ‘no thanks’ option. Fools. That left them with two other options, develop their own DXmoremodern GPU or license one. As we mentioned earlier, Microsoft’s WARM requirements lead to massive die bloat, leakage, and many other things that chips used for mobile OSes that actually sell in volume want to actively avoid.
Qualcomm could take Krait and hobble it on both cost and power fronts to please Microsoft like Nvidia did with Tegra, or they could wisely skip out on it and have a competitive chip for Android and every other OS out there. Instead of picking one, Qualcomm decided to split the difference, the dual core uses the much saner Adreno GPU, and quad-core WARM oriented part will have a MS pleasing monster.
That decided, there was one other decision to make, buy or build. There wasn’t enough time to build their own before WARM disappoints many this holiday season, so that leaves buy. With that in mind, SemiAccurate’s sources tell us that they have licensed Imagination for their Windows aimed parts, but it is likely just a stopgap until their internal GPU efforts ramp up.
The only remaining question is what Imagination GPU Qualcomm is using, 5- or 6-series. No one is directly telling us which way it will go, but we doubt that it will be a 6-series/Rogue GPU for time and cost reasons. Pairing Krait/S4 with Rogue would be a seriously powerful match, but in the end, a big 5-series GPU will probably surface in the quad-core parts later this year.S|A
Latest posts by Charlie Demerjian (see all)
- Thing go bump(gate) in the night for Nvidia’s GP100 Pascal GPU - May 3, 2016
- Cavium’s Octeon TX blends compute and packet moving - May 2, 2016
- Qualcomm releases Zeroth API to developers - May 2, 2016
- Another detail about Qualcomm server SoCs revealed - Apr 27, 2016
- AMD finally really honestly launches the dual Fury - Apr 26, 2016