News
IndustryTSMC’s grip on advanced chipmaking will face competitionIndustryCadence joins European Chip Design Platform to support semiconductor startups and SMEsMaking chipsMarc Hijink on ASML’s Culture, Supply Chain, and GeopoliticsProcessorsIntel's post-Nova Lake CPUs take shape as Razor Lake desktop and laptop variants surfaceAI chipsHow Google’s New AI Chip Software Helps It Battle NvidiaIndiaMore IC Packaging Plants Emerge In IndiaIndustryUS Tightens China AI Chip Access as Nvidia, TSMC and Semiconductor Stocks SlideDesignAxelera AI moves into HPC with TitaniaMaking chipsChinese envoy says chip equipment export curbs threaten EU-China trade thawIndiaSEMICON India 2026 Day 2 Sees 25 MoUs, Strategic Collaborations Across Semiconductor Value ChainIndustryTSMC’s grip on advanced chipmaking will face competitionIndustryCadence joins European Chip Design Platform to support semiconductor startups and SMEsMaking chipsMarc Hijink on ASML’s Culture, Supply Chain, and GeopoliticsProcessorsIntel's post-Nova Lake CPUs take shape as Razor Lake desktop and laptop variants surfaceAI chipsHow Google’s New AI Chip Software Helps It Battle NvidiaIndiaMore IC Packaging Plants Emerge In IndiaIndustryUS Tightens China AI Chip Access as Nvidia, TSMC and Semiconductor Stocks SlideDesignAxelera AI moves into HPC with TitaniaMaking chipsChinese envoy says chip equipment export curbs threaten EU-China trade thawIndiaSEMICON India 2026 Day 2 Sees 25 MoUs, Strategic Collaborations Across Semiconductor Value Chain
EveryChip
A Palit-brand Nvidia GeForce GTX 750. This particular card came with 2GB of GDDR5 RAM, and outputs for legacy VGA, DVI and HDMI.

Maxwell

Nvidia GPU microarchitecture focused on power efficiency

Photo: Blakegripling ph, CC BY-SA 4.0, via Wikimedia Commons

Maxwell is a graphics processing unit design by Nvidia, launched in February 2014. It succeeded Kepler and prioritised energy efficiency through a new streaming multiprocessor layout. The architecture powered the GeForce 900 series and Tegra X1 chips.

Share on WhatsApp

  • Launched in February 2014 with GeForce GTX 750
  • Increased L2 cache to 2 MiB in first generation
  • Used in Tegra X1 system on a chip
  • Manufactured by TSMC on 28 nm, 20 nm, and 16 nm nodes
  • Succeeded by Pascal architecture

What it is

Maxwell is a graphics processor microarchitecture created by Nvidia. It arrived in early 2014 as the follow-up to the Kepler design. The primary goal was to boost performance per watt rather than just raw speed. This approach allowed for smaller dies and lower power consumption in laptops and desktops. It supports CUDA Compute Capability 5.0 and 5.2 depending on the specific chip generation. The name honours James Clerk Maxwell, a physicist known for his work in electromagnetism.

How it came about

Nvidia developed Maxwell to address the power limits of previous designs. The first products appeared in February 2014 with the GeForce GTX 750 and 750 Ti. These used the GM107 chip code. Later that year, second-generation models like the GTX 970 and 980 launched with GM20x codes. The architecture also found its way into mobile devices via the Tegra X1 system on a chip. Support for security updates continued until October 2028.

Inside the chip

The core innovation was the Streaming Multiprocessor, renamed SMM. Unlike Kepler, which shared resources across a large pool, Maxwell partitioned execution units. Each scheduler controls its own set of integer cores and load units. This modular layout reduces the need for power-hungry crossbars. The L2 cache grew significantly to two megabytes in the first generation. This larger cache reduced pressure on memory bandwidth, allowing narrower memory buses. Video encoding and decoding engines were also upgraded for better media handling.

How it was made

TSMC manufactured the chips using several process nodes. The first generation used 28 nanometre technology. Second-generation parts utilized 20 nm and 16 nm processes. The design changes allowed for smaller die areas despite increased cache sizes. The physical layout supported tiled caching, which differs from tile-based rendering. This distinction helps manage memory access patterns more efficiently without changing the fundamental rendering pipeline.

Maxwell
A drawing made by EveryChip from what this page says. It is a diagram, not a picture of the real chip.

Where it went

These processors powered the GeForce 700, 800M, and 900 series graphics cards. They were common in gaming PCs and some OEM systems. The Quadro Mxxx series used them for professional workstation tasks. Mobile variants appeared in the GeForce 800M laptop GPUs. The Tegra X1 automotive and embedded platform also relied on this architecture. It handled games, video playback, and general computing tasks across these devices.

Versions and family

First-generation chips carried GM107 and GM108 codes. They offered CUDA capability 5.0. Second-generation variants included GM200, GM204, and GM206. These supported CUDA 5.2 and added features like HDMI 2.0. The GTX 970 had a unique memory configuration due to disabled units. Pascal succeeded Maxwell, bringing higher bandwidth memory interfaces. Maxwell remains distinct for its focus on efficiency over pure throughput.

Why it matters

Maxwell shifted industry focus toward power efficiency in graphics. It demonstrated that architectural changes could yield gains without shrinking transistors alone. The design influenced subsequent Nvidia architectures. Its support lifecycle extended well beyond its initial release. The Tegra X1 integration showed its versatility beyond discrete graphics cards. It set a baseline for future mobile and desktop GPU designs.

Problems and reception

Users noted the improved power efficiency in laptops. The GTX 970 memory segmentation caused some controversy among consumers. Nvidia clarified the technical reasons for the split memory bus. Despite this, the card remained popular for its price-to-performance ratio. The lack of full hardware H.265 decoding in first-gen parts was a minor limitation. Overall, it was seen as a mature and stable design.

Questions people ask about Maxwell

What is the main improvement of Maxwell over Kepler?

Maxwell focuses on power efficiency. It uses a partitioned streaming multiprocessor design that reduces energy waste. The larger L2 cache also lowers memory bandwidth requirements, saving power and die area.

Which graphics cards use the Maxwell architecture?

It powers the GeForce GTX 750, 750 Ti, 960, 970, 980, and Titan X. It is also found in the GeForce 800M series and Quadro Mxxx professional cards.

Does Maxwell support HEVC video decoding?

First-generation Maxwell chips do not support full hardware HEVC decoding. Second-generation GM20x chips added HEVC encoding support and improved H.264 capabilities up to 4K resolution.

Who manufactured the Maxwell chips?

TSMC manufactured the Maxwell GPUs. They used various process technologies including 28 nm for the first generation and 20 nm or 16 nm for the second generation chips.

Good reading and sources

Where this page comes from

The facts on this page come from the Wikipedia article "Maxwell (microarchitecture)", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryChip rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.

It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

Photo credits on this page

The small pictures above come from Wikimedia Commons under free licences. Some are cropped to fit.

The weekly digest

Every Monday: the chip of the day, new pages on EveryChip, and the conferences and anniversaries of the week ahead.