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
NVIDIA@28nm@Kepler@GK110_A1@GeForce_GTX_Titan@1251A1_NFF528.MOW_GK110-400-A1___Stack-DSC04727-DSC04758_-_ZS-retouched-1

Kepler

Nvidia GPU microarchitecture focused on energy efficiency and programmability.

Photo: Fritzchens Fritz, CC0, via Wikimedia Commons

Kepler is a graphics processing unit design by Nvidia, launched in April 2012. It succeeded Fermi and prioritised power efficiency over raw clock speed increases. Manufactured by TSMC on a 28 nm process, it powered GeForce 600 and 700 series cards. The architecture introduced features like GPU Boost and Dynamic Parallelism to improve performance per watt.

Share on WhatsApp

  • First Nvidia architecture to prioritise energy efficiency
  • Introduced GPU Boost for dynamic clock scaling
  • Manufactured on TSMC 28 nm process node
  • Supports DirectX 12 feature level 11_0
  • Features Hyper-Q with 32 hardware work queues

What it is

Kepler is a graphics processor microarchitecture created by Nvidia. It reached retail markets in April 2012 as the successor to the Fermi design. This generation marked a strategic shift for the company, moving away from pure performance gains to focus on energy efficiency and programmability. Built on a 28 nanometre manufacturing process by TSMC, Kepler chips powered the GeForce 600 and 700 series, along with professional Quadro and Tesla computing modules. The architecture is named after the German mathematician Johannes Kepler.

How it came about

Nvidia developed Kepler to address the power consumption issues of its predecessor, Fermi. While Fermi prioritised compute and tessellation performance, Kepler aimed to balance performance with efficiency. The design team simplified instruction scheduling and removed complex hardware blocks that prevented data hazards. This reduction in die space helped lower power usage. The architecture was first introduced in April 2012 and remained in production alongside its successor, Maxwell, in later GeForce series.

Inside the chip

Kepler uses a new streaming multiprocessor called SMX. Each SMX contains 192 CUDA cores, a significant increase from the 32 cores per unit in Fermi. The design employs a unified GPU clock rather than separate shader clocks, which reduces power consumption by half in that area. Key features include Hyper-Q, which expands work queues to 32 to improve utilisation, and Dynamic Parallelism, allowing kernels to dispatch other kernels without CPU involvement. The GK110 variant supports double-precision calculations at one-third the speed of single-precision, while consumer models are limited to one-twenty-fourth.

How it was made

TSMC manufactured all Kepler chips using its 28 nm process node. The architecture includes various dies such as GK10x, GK110, and GK210. The GK210 increased register limits to 512 to handle high register pressure situations. Memory support includes GDDR5 with speeds up to 7 GHz. The design incorporates single-bit error correction code for all memory levels, including the register file. Video encoding is handled by the NVENC block, which supports H.264 content up to 4K resolution.

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

Where it went

Kepler GPUs were widely used in desktop gaming PCs through the GeForce 600 and 700 series. They also powered professional workstations via the Quadro K series and high-performance computing clusters through Tesla K modules. The GK20A variant served as the graphics component in the Tegra K1 system on a chip. Features like GPU Boost allowed these cards to dynamically adjust clock speeds based on thermal headroom, maximising performance within power limits. The architecture supports DirectX 12 feature level 11_0 and Vulkan 1.2.

Versions and family

The Kepler family includes several variants. Consumer cards mostly use GK10x dies with limited double-precision performance. High-end computing cards like the GK110 and GK210 offer improved double-precision ratios and more streaming multiprocessors. The GK208 and GK20A represent smaller or embedded implementations. Kepler was succeeded by the Maxwell microarchitecture, though some GeForce 800M series laptops used Kepler chips alongside newer Maxwell designs. The architecture is now unsupported by Nvidia.

Why it matters

Kepler established energy efficiency as a primary goal for Nvidia GPU design. Its introduction of GPU Boost changed how graphics cards manage power, allowing dynamic frequency scaling similar to central processing units. Features like Hyper-Q and Dynamic Parallelism improved software flexibility for high-performance computing tasks. The architecture supported advanced video decoding for H.265 and encoding for H.264, enabling efficient media processing. It laid the groundwork for future efficiency-focused designs like Maxwell and Pascal.

Questions people ask about Kepler

What is the main difference between Kepler and Fermi?

Kepler focuses on energy efficiency and programmability, whereas Fermi prioritised raw performance in compute and tessellation. Kepler uses a unified GPU clock and simplified scheduling to reduce power consumption, achieving better performance per watt than its predecessor.

Which products used the Kepler microarchitecture?

Kepler powered the Nvidia GeForce 600 and 700 series graphics cards. It was also used in the Quadro K series for workstations, Tesla K modules for servers, and the GK20A chip inside the Tegra K1 system on a chip for mobile devices.

What is GPU Boost in Kepler GPUs?

GPU Boost is a feature that dynamically adjusts the graphics card clock speed. It ensures the GPU runs at a base clock but increases speed when thermal and power limits allow. This maximises performance without exceeding the specified power target, typically 170 W for reference cards.

Does Kepler support modern graphics APIs?

Yes, Kepler GPUs support DirectX 12 up to feature level 11_0 and Vulkan 1.2. They also support OpenGL 4.6. However, they do not support all DirectX 11.1 features, such as target-independent rasterisation, which limits their compatibility with some newer software requirements.

Good reading and sources

Where this page comes from

The facts on this page come from the Wikipedia article "Kepler (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.