Maxwell is Nvidia’s most power-efficient GPU

Nvidia’s (NVDA) Maxwell, like its predecessor Kepler, is a massively parallel architecture consisting of hundreds of CUDA cores. Maxwell architecture is based on the same 28nm manufacturing process as Kepler. However, Maxwell is more power efficient.

In Kepler’s design, many CUDA cores were sometimes idle during any given operation. These CUDA cores continued to consume power even when microprocessors were not working, because the GPU’s (graphic processing unit) control logic could not turn them off individually. AMD’s (AMD) Graphics Core Next (or GCN) and Intel’s Knights Corner architecture are peers of Nvidia’s Kepler chips in the high-performance computing market.

You can consider investing in the VanEck Vectors Semiconductor ETF (SMH) and the PowerShares QQQ Trust (QQQ) to gain exposure to Nvidia. Nvidia makes up about 1.94% and 0.22% of these ETFs, respectively.

Realizing the issues in Kepler’s architecture, Maxwell architecture broke Kepler’s monolithic core logic into a number of independent components, so that each of those control a small number of CUDA cores. Thus, whenever an individual processor is idle, the control logic managing it can turn off the power. Nvidia claims that this more precise power control improves performance per core by 135% percent, as the above chart shows. The more precise power control also doubles the GPU’s performance per watt.

Maxwell succeeded the Kepler architecture that serves as the foundation for Nvidia’s most powerful video cards. Nvidia made Maxwell architecture available in economically priced GPUs, GeForce GTX 750 and GeForce 750 Ti, at its launch.

In comparison to other video cards, Nvidia said a generic desktop PC powered by an Intel (INTC) Core i5 with an integrated Intel HD Graphics 4600 processor is capable of delivering Batman: Arkham Origins at 8 frames per second (or fps). On the other hand, this rate jumps to 53 fps with the GeForce GTX 750 Ti.