The Geforce brand is the series of gamer-orientated GPUs. Unlike the Quadro workstation GPUs, the Geforce have limited double-precision float points performance (except the GTX Titan, Titan Black and Titan Z) and lacks of the ECC (Error Correction code) RAM.

Begin with GeForce 8 series, Nvidia has implemented CUDA, a parallel computing platform and programming model, into their GPUs.

Generations

GeForce 256

The very first GeForce GPU family. Developed under the code-name NV10, It is the successor to the RIVA GPUs by the same company. It is fabricated in 220 nm process, and has the 128-bit bus width for video memory.

It has the twice the pixel pipelines of that of RIVA TNT2 GPU, and supports hardware Transform and Lighting. It is also the very first DirectX 7.0-compliant GPU.

GeForce2

Introduced in April, 2000, it is the second-generation of GeForce GPU family. Those GPUs support improved Transform and Lighting, NVIDIA Shading Rasterizer and High-Definition Video Processor. All GPUs are fabricated in 180 nm process.

It is also the first GPU family to have mobile version of the GPUs for mobile PCs.

GeForce3

Introduced in March, 2001, it is the 3rd-generation GeForce GPU family. It is the first GeForce GPUs to have vertex shaders, and supports pixel shader 1.1 and DirectX 8.0. It introduced the supports for shadow buffer and 3d textures. All GPU were fabricated in 150 nm process. It doesn't has mobile GPUs.

The GeForce 3 GPUs has the Lightspeed Memory Architecture (LMA) memory manager to reduces overdraw and conserves memory by compressing the z-buffer.

While the NV20 chips were only support DirectX 7.0 and OpenGL 1.2, the NV2A chips, that is first seen on Microsoft's Xbox, supports DirectX 8.0a and OpenGL 1.4.

GeForce4

Introduced in February, 2002, it is the 4th-generation GeForce GPU family. It introduced supports for accuview anti-aliasing. All models were fabricated in 150 nm process. This line of GPUs has the better LMA II memory controller.

Oddly the NV1F, NV17 and NV18 chips lacks the vertex shader units and only support up to DirectX 7.0 and OpenGL 1.2, while the rest of this lineup supports up to DirectX 8.0a and OpenGL 1.3.

The GeForce4 PCX4300 is the first GPU to supports PCI-Express x16 interface.|

GeForce FX

Introduced in 2003, it is the 5th-generation GeForce GPU family. NV34 and NV34B chips were manufactured in 150 nm and 140 nm process respectively, while others were fabricated in 130 nm process. All models supports DirectX 9.0b and OpenGL 1.5.

Unfortunately this line of GPUs suffers from weaker parallelism and calculation throughput that results in weaker performance in the Shader Model 2 programs. Also the GeForce FX 5800 series graphics cards suffers from heavy noise because of its cooler design.

Some of the GPUs in this GPU family utilizes GDDR memory.

GeForce 6

Introduced in 2005, they are the 6th-generation GeForce GPU family. This GPU family introduced the Mutil-GPU SLI technology and PureVideo post-processing decoder and support Shader Model 3.0, DirectX 9.0c and OpenGL 2.0.

MCPx1 and C51M GPUs were manufactured in 90 nm process and NV40, NV41, NV45 and NV41M were made in 130 nm process, while other GPUs were made in 110 nm process.

The GeForce Go 61xx series were suffers from high failing rate, and Nvidia faced heavy charge against revenue because of this. Some of the GPUs in this GPU family are the only one that supports Nvidia's TurboCache.

Some of the entry-level GPUs were integrated into Nvidia's NForce motherboard chipsets.

GeForce 7

Introduced in June 2005, it is the 7th-generation GeForce GPU family. As with the GeForce 6 series, the GeForce 7 GPUs supports DirectX 9.0c, Shader Model 3.0 and OpenGL 2.1. Few GPUs were made in 110 nm and 80 nm process, while most are made in 90 nm process. Instead of the 'NV' initials in their code names, most GPUs has the 'G' initial.

They were the last Nvidia GPU family to supports AGP interface.

The GeForce 7900 GX2 and 7950 GX2 are the very first dual-GPU graphics cards for Nvidia.

The custom version of G70 GPU, known as RSX 'Reality Synthesizer', is the GPU uses on Sony PlayStation 3 gaming console.

GeForce 8

Introduce in November 2006, they were the 8th-generation GeForce GPU family. The GPU family is based on the Tesla architecture, replacing the old pixel shaders and vertex shaders with the unified shaders. The Tesla architecture is also the first to implements Compute Unified Device Architecture (or CUDA for short), making this and later GPU families capable to doing General Purpose GPU workload.

All models support DirectX 10.0, Shader Model 4.0 and OpenGL 2.1 (3.3 with later driver support). The GPU family also support Coverage Sample Anti-Aliasing, Angle-Independent Anisotropic Filtering and 128-bit OpenEXR HDR.

The GPUs were made in 90 nm, 80 nm, 65 nm and 40 nm process.

Some models in this GPU family support PCI-Express 2.0 x16 bus interface.

GeForce 9

Introduce in February 2008, the GeForce 9 series is the 9th generation GeForce GPU family. It is based on a slightly repolished Tesla architecture, which adding PCI-Express 2.0 support and improved color & z-compression.

All models were bulit on 65 nm and later 55 nm process.

GeForce 100

The GeForce 100 series, introduced in January 2009, is a series of GPU family consists for re-branded graphics card from previous generations based on Tesla architecture. This GPU familiy only available through OEM.

GeForce 200

Introduced in 2008, it is the 10th generation GeForce GPU family (excluding 100 and 300 series). It is based on the second generation Tesla architecture.

All models supports Coverage Sample Anti-Aliasing, Angle-Independent Anisotropic Filtering, and 240-bit OpenEXR HDR.

Most models were made in 65 nm and 55 nm process, and only the entry level GPUs were made in the newer 40 nm process.

Oddly, it is the only GPU family to have a GPU that has 512-bit memory bus width.

GeForce 300

Introduced in 2009, it is the second GPU family to be a consists of re-branded GPUs after GeForce 100 series. All models support DirectX 10.1 and Shader Model 4.1, except the GT 330 GPU.

GeForce 400

The GeForce 400 series is the 11th-generation GeForce GPU family that were introduced in April 2010. It is based on the Fermi architecture. All model were made in 40 nm process.

With the exception of the OEM-only Geforce 405, all models supports DirectX 11, OpenGL 4.4 and Shader Model 5.0. Later drivers also provides OpenCL 1.1 supports for the Fermi-based GPUs.

Oddly, Nvidia released the first-generation Fermi GPUs with at least one SM (Streaming Multiprocessor) disabled.

GeForce 500

The GeForce 500 series, introduced in November 2010, is the 12th generation GeForce GPU family. It is based on improved 2nd-generation Fermi architecture. The improved architecture use less power and this allows Nvidia to release GPU with all SMs enabled (GTX 580).

As with GeForce 400 series, it supports DirectX 11, OpenGL 4.4 and Shader Model 5.0. Later drivers also provides OpenCL 1.1 supports for the Fermi-based GPUs. All model were made in 40 nm process.

GeForce 600

The GeForce 600 series is the 13th generation GeForce GPU family. Introduced in 2012, most GPUs in this GPU family are based on Kepler architecture, which consume less power than Fermi.

The Kepler architecture unifies both core clock and shader clock, abandoning the shader clock found in previous GPU designs. This increases the performance-per-watt of the new architecture. The GPU also introduces PCI-E 3.0 interface supports.

Fermi GPUs were made in 40 nm process, while Kepler and GPUs were made in the smaller 28 nm process.

As with previous two GeForce GPU families, it supports DirectX 11, OpenGL 4.4, Shader Model 5.0 and OpenCL 1.1.

GeForce 700

Introduced in 2013, the GeForce 700 series is the 14th-generation GeForce GPU family. Most GPUs are based on Kelper and Fermi, while GeForce GTX 750 and GTX 750 Ti were based on the newer Maxwell architecture.

Unique to the new Maxwell architecture is the larger cache and more control logics (about four) per SMs, improving energy efficiency.

The GeForce GTX Titan series GPUs has the double-precision FLOP performance unlocked and has twice the memory than the rest of the GPUs in this GPU family.

As with previous GeForce GPU families, it supports DirectX 11, OpenGL 4.4, Shader Model 5.0 and OpenCL 1.1.

GeForce 800 series

Introduced in 2014, it is the OEM-only GeForce family.

It consisted of Kepler-based and Maxwell-based mobile GPUs.

GeForce 900 series

Introduced in September 2014, it is the 15th generation GeForce GPU family. It served as the introduction of the second iteration of the Maxwell architecture, featuring slightly tweaked SMM blocks.

Despite the decrease in memory bus width (256-bit v.s. 386-bit of GK110), the second-generation Maxwell GPUs has the improved delta color compression technology so the bandwidth efficiency is further improved.

The GM2xx GPUs also introduced Multi-Frame Sampled Anti-Aliasing (MFAA), Real-Time-Voxel-Global Illumination and many more. In additions, The GM2xx series GPUs supports encoding video at 4K resolution at 60 fps.

In January 2015, it was discovered by users that the GTX 970 doesn't fully utillizes its RAM. It was later explained by Nvidia that due to the disabled ROP units and reduced L2 cache, it has fewer resources to handles the large 4 GB, so they split it into 2 memomry segments (3.5 GB and 0.5 GB). This makes the GTX 970 the first cut-down variant to not disable a memory sub-controller completely.

The GeForce 900 series support the DirectX 12.0 API with feature level 12_1. It also supports OpenGL 4.5 and OpenCL 1.2. The GeForce 900 series is also the first to have UEFI Graphics Output Protocol.

GeForce 10 series

First released in March 2014, the GeForce 10 series utilize the Pascal microarchitecture, which features dynamic load balancing scheduling system, instruction-level preemption, and triple buffering at driver level. It is also the first series to support GDDR5 and GDDR5X memory.

The GeForce 10 series support the DirectX 12.0 API with feature level 12_1, as well as the Vulkan 1.3 API. It also supports OpenGL 4.6 and OpenCL 3.0.