GeForce 800M series

Lore & Background

The GeForce 800M series was originally planned to include both desktop and mobile chips based on the Maxwell microarchitecture, named after the Scottish theoretical physicist James Clerk Maxwell. However, because mobile GPUs under the GeForce 800M series had already been released using the Kepler architecture, Nvidia decided to rename its GeForce 800 series desktop GPUs as the GeForce 900 series. The Maxwell microarchitecture, the successor to Kepler, was the first Nvidia architecture to feature an integrated ARM CPU of its own, enabling Maxwell GPUs to be more independent from the main CPU.

First-generation Maxwell (GM10x) focused on power efficiency rather than consumer-facing features. Nvidia's video encoder NVENC became 1.5 to 2 times faster than on Kepler-based GPUs, encoding video at 6 to 8 times playback speed. PureVideo Feature Set E video decoding saw an 8 to 10 times performance increase due to the video decoder cache and memory efficiency improvements, though HEVC was not supported for full hardware decoding. A new low-power state called GC5 was introduced for video decoding. The L2 cache on GM107 was increased to 2 MB from 256 KB on GK107, allowing a reduction in memory bus width to 128 bits from 192 bits on GK106, saving power.

Nvidia changed the streaming multiprocessor design from Kepler's SMX to SMM, partitioning resources so that each of the four warp schedulers controls isolated FP32 CUDA cores, load/store units, and special function units, while texture units and FP64 CUDA cores remain shared. Nvidia claimed a 128 CUDA core SMM has 90% of the performance of a 192 CUDA core SMX. Maxwell provided native shared memory atomic operations for 32-bit integers and native shared memory 32-bit and 64-bit compare-and-swap (CAS).

Reader's Guide

The GeForce 800M series occupies a transitional position in Nvidia's mobile GPU lineup, bridging the Kepler and Maxwell architectures. Its significance lies in demonstrating Nvidia's shift toward power efficiency and integrated ARM processing in the Maxwell microarchitecture, which would later define the GeForce 900 series for desktops. The series also highlights Nvidia's naming confusion: the 800M series launched with Kepler-based chips while desktop Maxwell chips were originally slated for the GeForce 800 series but were rebranded as the 900 series. The inclusion of Fermi-based GF117 cores in the 820M shows the range of legacy architectures carried forward. The Maxwell-based models in the 800M series introduced features like improved NVENC performance, larger L2 cache, and the SMM streaming multiprocessor design, which allowed finer resource allocation and power savings. Support for DirectX 12 ensured compatibility with modern gaming APIs. The series' legacy includes its eventual transition to legacy driver support: Kepler-based notebook GPUs moved to legacy support in April 2019 and stopped receiving critical security updates in April 2020, while the GTX 830M, 840M, and 850M were unaffected by that change. In May 2025, Nvidia discontinued developer support for Maxwell, Pascal, and Volta architectures, and on July 1, 2025, announced that driver branch 580 would be the last to support these architectures, marking the end of active software support for the Maxwell-based models of the 800M series.

Did You Know?

• The GeForce 800M series includes chips based on Kepler, Maxwell, and Fermi architectures.
• The 820M uses GF117 cores based on the Fermi architecture, while higher-end models use Kepler or Maxwell.
• Maxwell-based GPUs in the 800M series were the first Nvidia architecture to feature an integrated ARM CPU.
• Nvidia skipped the GeForce 800 series for desktop graphics cards because the GTX 800M series already used the Kepler architecture.