Intel Core i7-3687U vs Intel Core i3-7020U vs Intel Celeron 3965Y

Intel Core i7-3687U

The Intel Core i7-3687U is the fastest ULV-processor for laptops based on the Ivy Bridge architecture at the time of introduction in Q1 2013. Due to Hyperthreading, the two cores can handle up to four threads in parallel leading to better utilization of the CPU. Each core offers a base speed of 2.1 GHz but can dynamically increase clock rates with Turbo Boost up to 3.1 GHz (for 2 active cores) and 3.3 GHz (for 1 active core). As its Sandy-Bridge-based predecessors, it offers a 4 MB L3-Cache.

Ivy Bridge is an enhanced die shrink of the Sandy Bridge architecture with improvements on both GPU and CPU performance. The CPUs are produced in 22nm (versus 32nm Sandy Bridge CPUs) and are the first to introduce 3D transistors for increased energy efficiency when compared to similarly clocked Sandy Bridge processors. Other new features are integrated PCI Express 3.0 and  DDR3(L)-1600 support.

The performance of the Core i7-3687U is slightly above a similarly clocked Sandy Bridge processor due to the architectural improvements. Therefore, and because of the higher clock speed, the old Core i7-2677M is easily outperformed.

The integrated Intel HD Graphics 4000 offers 16 Execution Units (EUs) clocked at 350 MHz up to 1200 MHz with Turbo Boost, delivering a performance comparable with an AMD Radeon HD 6620G.

The ULV-CPUs are rated at a TDP of 17 W including graphics card and memory controller.  Therefore, the CPU is suited for small laptops and ultrabooks.

Intel Core i3-7020U

The Intel Core i3-7020U is a dual-core processor of the Kaby Lake architecture. It offers two CPU cores clocked at 2.3 GHz (without Turbo Boost) and integrates HyperThreading to work with up to 4 threads at once. The architectural differences are rather small compared to the Skylake generation, therefore the performance per MHz should be very similar. The SoC includes a dual channel DDR4 memory controller and Intel HD Graphics 620 graphics card (clocked at 300 - 1000 MHz). It is manufactured in an improved 14nm FinFET process at Intel. Compare to the old i3-7100U, the newer i3-7020U is clocked 100 MHz lower (CPU cores) and therefore the current entry level model for the Core i3 line. The similar Pentium Gold 4415U offers less Cache (2 versus 3 MB) and a slower GPU.

Architecture

Intel basically uses the same micro architecture compared to Skylake, so the per-MHz performance does not differ. The manufacturer only reworked the Speed Shift technology for faster dynamic adjustments of voltages and clocks, and the improved 14nm process allows much higher frequencies combined with better efficiency than before.

Intel Celeron 3965Y

The Intel Celeron 3965Y is a low power entry level processor for small laptops or tablets and based on the Kaby Lake architecture. It was announced in Q2 2017 and offers two processor cores clocked at 1.5 GHz (no Turbo Boost). Compared to the Pentium 4410Y, the Celeron 3965Y offers no HyperThreading and therefore can execute only up to two threads simultaneously. The chips also includes the Intel HD Graphics 615 GPU, a dual-channel memory controller (DDR3L/LPDDR3) as well as VP9 and H.265 video de- and encoder. It is still produced in a 14 nm process with FinFET transistors. Currently (2017) it is the slowest Kaby-Lake based processor.

Architecture

Intel basically used the familiar micro architecture from the Skylake generation, so the per-MHz performance is identical. Only the Speed-Shift technology for faster dynamic adjustments of the voltages and clocks was improved, and the matured 14 nm process now also enables much higher frequencies and better efficiency than before. 

Performance

Due to the missing HyperThreading, the Celeron 3965Y is slightly slower than the similar clocked Pentium 4410Y in some multi threaded benchmarks. The single thread performance is due to the missing Turbo Boost significantly worse than the more expensive Core m3-7Y30. This should be noticeable in daily tasks. 

Graphics

The integrated Intel HD Graphics 615 GPU has 24 Execution Units (EUs) like the old HD Graphics 515 and runs with clocks between 300 and 850 MHz in combination with this processor. The performance heavily depends on the TDP limit as well as the memory configuration; with fast LPDDR3-1866 RAM in dual-channel mode, the GPU should sometimes be able to compete with the HD Graphics 520, but can also be much slower in other scenarios. Modern games from 2016 will, if at all, only run smoothly in the lowest settings.

Contrary to Skylake, Kaby Lake now also supports hardware decoding for H.265/HEVC Main10 with a 10-bit color depth as well as Google's VP9 codec.

Power Consumption

The chip is manufactured in an improved 14 nm process with FinFET transistors, so the power efficiency was once again improved significantly. The TDP is rated at 6 Watt and can be reduced to 4,5 Watt (cTDP down).