Nvidia Geforce – Everything you need to know about the blue team’s line of gaming GPUs

A bit about the history of the Nvidia Geforce line

Geforce is the name of a line of graphics processing units (GPUs) designed and developed by Nvidia. This name originated from a contest in 1999. The “Name that chip” contest was organized by Nvidia to find a name for the successor product line to the Riva TNT2. There were more than 12,000 entries and 7 winners.

Up to the RTX 30-series, there have been a total of 17 generations of Geforce products. If talking about discrete graphics card products, its only remaining competitor is AMD’s Radeon line, after companies such as 3dfx, Matroxx,… went bankrupt one after another.

Generations of Geforce

Below, I will only mention GPU lines for desktops. And I will only focus on GPUs sold in retail form, not mentioning OEM codes.

Laptop GPU codes will only be listed from the Geforce 700-series generation onwards.

The origin of it all: Geforce 256

Geforce 256 is the first generation of Geforce GPUs launched by Nvidia. Launched on August 31, 1999, this GPU is 20 years old this year. At that time, its launch created a seismic shift in the graphics industry. Because it was the first graphics card to support hardware image conversion, lighting and shaders. However, games supporting those technologies were released about a year later. Early Geforce 256 cards used SDR type VRAM, which was later converted to DDR type.

Nvidia stopped supporting this line of graphics cards on April 14, 2005, with the last driver version for Windows 2000/Windows XP 32-bit. However, those drivers can still be installed for Windows 7. But Aero effects will not be supported.

This generation’s opponents include:

ATi Rage 128 and Rage Fury MAXX

3dfx Voodoo3

Matrox G400

S3 Savage 3000

Geforce 256 technical specifications

2nd generation: Geforce2

Launched on September 7, 2000, Geforce2 is the successor of the Geforce 256 series. In this generation, Nvidia has switched to designing dual texture processors on the same pipeline. Since then, the texture fillrate has doubled compared to its predecessor. The manufacturing process has also been scaled down, from 220nm to 180nm. More prominently, we have a Geforce2 Ti GPU that has been shrunk even further, down to 150nm.

The Geforce2 line has many variants, the most powerful of which is the Geforce2 Ultra variant. Its non-MX variants were succeeded by the Geforce3 series. However, the Geforce2 non-MX line is still retained and has been reduced in price, to serve as a counterweight in the mid-range segment. And later, the entire Geforce2 line was completely replaced by the Geforce4 line.

This line was discontinued on November 20, 2006. Its competitors include:

3dfx Voodoo5

ATi Radeon R100

PowerVR Series 3

Geforce2 series specifications

Subsequent Geforce lines, from Geforce3 – Geforce6, brought only a few improvements. Accordingly, the next generation of prominent GPUs will be Geforce8

Legend of the past: Geforce 8 series (8xxx)

The legendary Geforce 8 (original name: G80) was officially launched on November 8, 2006. At that time, it was the first GPU line to support Direct3D 10. And this was also the first GPU to use architectural code naming after famous scientific giants. Specifically, Geforce 8 is manufactured on Tesla microarchitecture, named after physicist Nikola Tesla. The process has also been scaled down to 90nm. The Tesla architecture is also Nvidia’s first architecture to use unified shaders. And, after the Geforce 7 series, the Geforce 8 series is Nvidia’s next generation brought to laptops

Initially when it was first launched, the new Geforce 8 series only had the code 8800GTX. The GTS version debuted a few months later. And it took up to 6 months for popular card codes to be released. A year later, Nvidia launched a remake of Geforce 8, based on the G92 GPU and 65nm process.

Support for the Geforce 8 series was discontinued on December 14, 2016, ending more than 10 years of “roaring” as a monument.

Geforce 8 series technical specifications

Legendary successor: Geforce 9

Continuing the legendary Geforce 8, we have a new legend: Geforce 9. This version is manufactured on a reworked Tesla architecture, with 65nm and 55nm processes.

The biggest highlight is that, along with the G92 GPU, Geforce 9 has brought PCI Express 2.0 support to all codecs. And so far, PCI Express 2.0 has not been much saturated by newer standards.

Along with Geforce 8, Geforce 9 was also discontinued on December 14, 2016.

Geforce 9 series technical specifications

At this point, I would like to skip the Geforce 100-series line. Because this line is just a rebranded Geforce 9 line.

Geforce 200-series

Just 4 months after launching Geforce 9, Nvidia is preparing to launch Geforce 200-series. This generation is produced based on the 2nd generation of Tesla architecture. The manufacturing process also ranges from 65nm to 40nm.

Starting with this generation, Nvidia has adopted the GPU bin method. For example, the GTX 280 and GTX 260 are both based on the GT200 GPU. But Nvidia differentiates them by testing for errors during function implementation. If a GPU does not meet GTX 280 specifications, it will be retested and labeled GTX 260 (fewer SPs and ROPs and narrower memory width. And at the end of 2008, Nvidia released a new version of GTX 260). GTX 260, with increased SP number compared to the old version.

Nvidia stopped supporting this GPU line on April 1, 2016. However, this day is actually April Fool’s Day, so many users think this is just a joke from Nvidia.

Geforce 200-series specifications

I will skip the Geforce 300-series line because it is just a rename Geforce 200-series line. And that line is only sold through OEMs. Thereby, the next generation of major GPUs will be:

Geforce 400-series

Released on April 12, 2010, Geforce 400-series is the premise for Nvidia to introduce a completely new architecture – Fermi. In fact, it is expected to launch in November 2009. But after some delays, it was revealed on March 26, 2010 and went on sale shortly after on April 12.

Nvidia claims that Fermi is a major new step for them after the Tesla architecture. GF100, Fermi’s first GPU, is quite powerful: 512 SPs, divided into 16 clusters, each cluster has 32 SPs and comes with 3 billion transistors. It is manufactured on TSMC’s 40nm process. And it is also Nvidia’s first GPU to support OpenGL 4.0 and Direct3D 11. However, no products based on the completely “unleashed” GF100 core have been released by Nvidia.

The reason I call this GPU line a “portable oven” is because the temperature on some codes is too hot. It’s so hot that you can fry an egg on the GPU.

Nvidia officially stopped supporting this generation, along with the Fermi architecture, in April 2018. It will switch to Legacy mode and be maintained until January 2019. Previously, they also announced that after Driver 390.xx version, they will no longer support 32-bit OS.

Geforce 400-series specifications.

Except for the GT 405, which still uses the Tesla architecture, all other GPU codes are based on the Fermi architecture.

Fermi “unleashed”: Geforce 500-series

This generation of Geforce 500-series is a refreshed version of the 400-series generation. First launched on November 9, 2010 with the GTX 580, the Geforce 500-series is the first generation to feature a fully unlocked Fermi GPU. And similar to its predecessor, it will support DirectX11, OpenGL 4.6 and OpenCL 1.1.

After GTX 580, Nvidia also released a number of other specialized versions:

January 25, 2011: Launched the GTX 560 Ti, hitting the mid-high range segment. With 30% more performance than the GTX 460 and positioned between the HD 6870 and HD 6950, it will replace the GTX 470.

March 15, 2011: GTX 550 Ti officially launched. Although it is based on the GF116 GPU, Nvidia named it GTX 550 Ti, not GTS 550. Performance is about 12% stronger than AMD’s HD 5770.

March 24, 2011: GTX 590 launched. Possessing 2 GPUs, it can easily support Nvidia’s 3D Vision technology without hindrance

Like the 400-series, Nvidia officially stopped supporting this generation in April 2018. It will switch to Legacy mode and be maintained until January 2019.

Geforce 500-series specifications

Next new architecture: Geforce 600-series

Launched on March 22, 2012, Geforce 600-series is the basis for Nvidia to introduce a completely new Kepler architecture. This name is named after the German mathematician Johannes Kepler. However, only mid-range and high-end codes use Kepler. Low-range codes will still be based on the somewhat aging Fermi architecture.

Along with that, the following new features are applied only on GPUs using Kepler:

PCI Express 3.0

DisplayPort 1.2

HDMI 1.4a

Encode H.264 by hardware

NVSurround, supports 4 2D screens and 3 3D screens at the same time

CUDA Compute 3.0

Nvidia GPU Boost

28nm process

TXAA

Currently, there is no information about stopping support for the Kepler architecture on PC. Therefore, Desktop GPUs using Kepler are still supported by Nvidia.

Geforce 600-series specifications

“National” card line: Geforce 700-series.

Geforce 700-series was launched on February 19, 2013. Although it is essentially a refreshed version of Kepler, some codes still use the outdated Fermi architecture, and a small number use the excellent Maxwell architecture of that time. The first GPU, GK110, is designed with outstanding computing performance. At the same time, from this Geforce 700-series line, the Titan line was born, with the first representative being the GTX Titan.

In addition, the 700-series is also Nvidia’s first GPU line to support DirectX 12 (on Windows 10). In addition, from this generation, Kepler has added the ability to support Dynamic Super Resolution (DSR).

As of April 2020, Nvidia has stopped supporting driver support for this GPU line. The latest driver version that fully supports this line is 388.71.

However, by 2021, due to the bitcoin storm and supply shortage sweeping. The Geforce 700-series suddenly returned, specifically the GT 710 and GT 730 codes. Therefore, Nvidia released a new driver version on June 10 with version number 466.77.

Geforce 700-series desktop specifications

Geforce 700-series specifications for Laptops

Geforce 800-series

Nvidia does not have any Geforce 800-series GPUs for the desktop. This line is entirely focused by Nvidia on the laptop segment.

Most controversial generation: Geforce 900-series

As a springboard to introduce Maxwell in the High-end segment, Geforce 900-series was introduced on September 18, 2014. They are manufactured on TSMC’s 28nm process. All GPUs in this series support DirectX 12, OpenGL 4.6, OpenCL 1.2 and Vulkan 1.0.

The reason I call this line the controversial GPU line is because of the GTX 970’s VRAM. On the box and advertising information, Nvidia notes that the graphics card using the GTX 970 GPU has 4GB of VRAM. However, in reality it cannot use more than 3.5GB of VRAM. Finally, after a series of lawsuits, Nvidia also admitted that they changed the GTX 970 GPU structure before launch. In addition, they also encounter performance problems when using Async Compute.

Until now, Geforce 900-series is still supported by Nvidia.

Geforce 900-series Desktop Specifications.

Geforce 900-series specifications for Laptops.

Gaming perfected: Geforce 10-series

On May 27, 2016, Nvidia made another huge step forward in the field of graphics, with the launch of Geforce 10-series with Pascal architecture, named after mathematician Blaise Pascal. They will be produced on two processes: 14nm for low-end codes and 16nm for high-end codes. In August 2016, Nvidia and Samsung signed a cooperation agreement, allowing the miniaturization of the process to 14nm to synchronize the entire product range together.

Along with the new architecture, new VRAM types GDDR5X and HBM2 were also introduced, of which HBM2 is only applied on the Quadro line. In addition, Geforce 10-series also supports DisplayPort 1.4, HDMI 2.0b and high-bandwidth SLI (HB SLI). Not to mention, last April, Nvidia released a driver that allows support for DirectX Ray Tracing for GPU codes from GTX 1060 and above. Therefore, it can be said that Geforce 10-series also supports Ray Tracing, but it is quite limited.

​In addition, Geforce 10-series also has outstanding computing and virtual currency mining capabilities, making this a perfect choice for bitcoin “farmers”. Of course, when the bitcoin storm returned in 2020-2021, the GTX 1070 Ti and GTX 1080 Ti suddenly became hot again, and old goods were also collected by workers at a higher price than the new purchase price.

Of course, this card line is still supported by Nvidia.

Geforce 10-series Desktop Specifications

Geforce 10-series Laptop specifications

Since this generation, Nvidia has dropped the M suffix from the GPU name.

Volta series

Volta was intended to be the successor to Pascal, at least on paper. However, there is only 1 product based on this architecture: Titan V. And maybe, because the Turing development process ended faster than expected, Nvidia decided not to launch any Geforce products based on Volta. But in return, Nvidia has given Volta great computing capabilities with the appearance of VRAM HBM 2 and Tensor core.

Nvidia intends Titan V for computing and Deep Learning purposes.

Titan V technical specifications

Redefining graphics: Geforce 20-series

At Gamescom 2018, after many rumors, the Geforce 20-series was finally released. Along with it comes many new things:

New naming method. Now, high-end products will be called RTX. The GTX name will still exist but for mid-range and low-end models.

New architecture: Turing, named after mathematician Alan Turing. Accordingly, this architecture is expected to have much superior computing power compared to its predecessor.

New process: 12nm FFN

RT multiplier, providing real-time Ray Tracing capabilities

New memory controller, supporting GDDR6 VRAM standard

DisplayPort 1.4a, supports DSC 1.2

VirtualLink VR via 1 USB Type-C port

NVLink

GPU Boost 4

Nvidia has divided the Geforce 20-series GPUs into two versions: A and Non-A. Non-A GPUs will not come pre-overclocked from the factory.

And contrary to tradition, Geforce 20-series does not have any code in the mid-range and low-end segment. Nvidia has positioned the Geforce 16-series for those segments.

On January 10, 2020, Nvidia quietly launched another version of the RTX 2060. That version will use GPU code TU104 as on the RTX 2070 Super and RTX 2080. It is equipped with the RTX 2060 graphics card model EVGA’s KO.

All Geforce 20-series GPU models from RTX 2070 Super and above support NVLink.

Geforce 20 upgrade: Super-series

On July 9, due to the impact of rival AMD on the Navi architecture card line, Nvidia launched the Super version of the Geforce 20-series. This version is basically just 2060, 2070 and 2080 with increased parameters. Among them, the most significant increase is the RTX 2060. When its Super version increased the amount of VRAM from 6GB to 8GB, the clock speed was also pushed higher. From there, bringing the performance closer to the normal RTX 2070.

Geforce 20-series Desktop Specifications

Geforce 20-series Laptop specifications

“Black sheep” in the family: Geforce 16-series

Geforce 16-series was created by Nvidia to fill the mid-range and low-end segment, which is currently being vacated by the Geforce 20-series and dominated by AMD. Although based on the same Turing architecture, the Geforce 16-series does not have an RT core, and therefore does not have Ray Tracing capabilities. But recently, through a driver version, the Geforce 16-series will support DXR, thereby having the ability to Ray Tracing, but it is still quite limited.

On October 29, Nvidia officially launched two upgraded versions of GTX 1650 and GTX 1660, which are GTX 1650 Super and GTX 1660 Super. Both are performance upgrades of the 1650 and 1660 models. In particular, the 1650 Super model is the most comprehensive upgraded version. On March 4, 2020, Nvidia continued to launch the GDDR6 version of the GTX 1650 model.

All Geforce 16-series GPU models do not support SLI/NVLink.

Geforce 16-series Desktop Specifications

Geforce 16-series Laptop specifications

Disappointing launch: Geforce 30 series

Although the Geforce 20 Series did not meet expectations, it still had some success. However, for Nvidia, that is never enough. And on September 1, they launched their latest strategic card: Geforce 30-series with a series of improvements such as:

Manufactured on an 8nm process specifically designed by Samsung

New generation SM

Ray Tracing capabilities are enhanced with 2nd generation RT cores and 3rd generation Tensor cores

Uses GDDR6X memory (except RTX 3070 uses GDDR6)

Uses PCI Express 4.0 protocol

At the launch on September 1, Nvidia announced 3 models: RTX 3070, RTX 3080 and RTX 3090. Of the 3 models released, only RTX 3090 is capable of supporting NVLink. The remaining 2 models will not be supported.

​ However, despite the extremely promising parameters and performance, this Geforce 30 series can be considered a failure for Nvidia. Almost no cards arrived in the hands of gamers in the first month after launch and the number of cards dropped in the following months. All the cards have been collected by miners to mine the Ethereum cryptocurrency, and to make sure they collect the cards, these miners use bots to place orders, causing the stock to run out in just 0.01 seconds. In addition, merchants also see this opportunity, participating in bot competitions to buy cards to resell at triple the price.

The card shortage continued for more than half a year and there were some positive signs when Nvidia launched the RTX 3060 card line with halved cryptocurrency mining performance. However, after only 1 month, Nvidia itself ‘squeezed’ itself when releasing a driver version with unlimited “coin mining” performance for this card, and so the card returned to a state of shortage.

​It wasn’t until June 2021 that light appeared at the end of the tunnel when Nvidia limited mining performance on all newly produced cards. Plus the cryptocurrency market going down makes the situation even better. At the time of writing, the price of new cards has begun to decrease slightly but has been sold at retail, gamers no longer need to build a whole new system to buy cards.

Geforce 30 series specifications for Desktop

Geforce 30 series specifications for Laptops

Epilogue

So I have gone through the entire history of the Nvidia Geforce graphics card line. See you again in the next article. At that time, I will go through the entire history of AMD’s Radeon graphics card line.