This site may earn affiliate commissions from the links on this page. Terms of employ.

CPU-Wafer1

Back in April, Intel announced its upcoming Apollo Lake platform, billed as an architectural refresh for low-cost Pentium and Celeron systems. Apollo Lake is built on Intel's next-generation Goldmont CPU compages with Skylake-derived graphics, and volition debut on a 14nm procedure. Bodily performance figures, notwithstanding, weren't discussed.

A leaked slide spotted on the Anandtech forums sheds some more low-cal on things and offers a preview of what Intel has packed into Goldmont and its accompanying GPU. The translated Chinese appears to the correct of each information block on the slide.

If this slide is accurate, Goldmont will meliorate CPU performance by 30%, innovate a new Skylake-derived GPU architecture for a 30% uplift in graphics, support DDR3L, LPDDR3, and LPDDR4, improve bombardment life by 15%, introduce new I/O options with added USB Type-C support, and work under both Linux and Windows 10 64-fleck.

Equally updates go, these are adequately significant, though evidently they can't be verified until we see shipping hardware. The gap between Silvermont and Goldmont won't be nigh as large as the divergence between Clover Trail and Silvermont, but in that location's a simple explanation for that. The Cantlet core inside Intel's Clover Trail tablet platform back in 2012 was badly outdated, having originally debuted on 45nm in 2008. To put that in perspective, ARM had moved from the ARMv11 compages in 2008 to Cortex-A8, then to dual-core Cortex-A9. The get-go Cortex-A15 devices shipped in 2012, merely earlier Intel formally launched its 22nm Silvermont compages and kickoff meaningful CPU update in iv years. Silvermont was far more potent than Clover Trail had been, especially since it offered upwards to four cores (Clover Trail topped out at dual-cadre + HyperThreading) and significantly improved performance per clock.

Goldmont is sticking with four cores and eschewing Hyper-Threading, only higher efficiency and improved clock speed make a 30% improvement quite reasonable, especially since Intel can plainly evangelize information technology while simultaneously improving battery life for a cyberspace comeback in cost per watt. 30% GPU uplift isn't going to plow Atom or Celeron systems into gaming boxes, but it could hands brand the difference between a playable and unplayable frame rate. If nosotros fix 30 FPS as our minimum acceptable gaming target on a upkeep system of this nature, a thirty% GPU increase would plow a 25 FPS game into a 32.5 FPS title. It may not compare well to monster systems with a GTX 1080, but in a small laptop that kind of comeback is quite solid.

While Intel plans to pull out of the Android business and killed its SoCs, we should nevertheless see Goldmont debuting on lower-finish 2-in-1 devices and some low-cost portable laptops. Nosotros were quite impressed with devices similar the Asus T100 when Bay Trail debuted — hopefully Goldmont will give budget users more bang for their cadet, while simultaneously packing in enough GPU horsepower to brand the system useful for some very lite gaming.