Lines and marking of NVIDIA video cards

Often before making a choice, it is difficult to understand its positioning in the line, as well as its features and performance. This article aims to shed light on this issue and tell you about the main lines of NVIDIA graphics cards, their generations, purpose, labeling and performance.


This material will not consider cards until 2010 and the Fermi microarchitecture, since they have already lost their relevance today. The only exceptions are top-end solutions for their time, but even now they are only simple average ones and are not sold anywhere. Also, the ION line will not be considered here, since at the moment computers with these cards are not sold.

To find useful information on NVIDIA graphics cards, we recommend using their website.

We advise you to take a look at the technologies used in modern video cards from NVIDIA.



This family is the most basic for NVIDIA. Its representatives are installed both in powerful gaming PCs and in simple office laptops. Video cards from this family satisfy 90% of the needs of ordinary consumers. The rest of the families are created for enthusiasts, for professionals and the corporate segment, or even for quite unusual tasks at first glance.


The generation and microarchitecture of the video chip are reflected in its code name. So:

  • Gf fermi
  • Gk kepler
  • GM Maxwell
  • GP Pascal

You can read about the first cards released on Pascal by following the corresponding links: GTX 1060, GTX 1070, GTX 1080, TITAN X, and now mobile GTX 1080/1070/1060 have already arrived.

GeForce Generation Microarchitecture Series

Fermi 400/500
Kepler 600/700
Maxwell 800/900
Pascal 1000

But this table needs to be corrected several times:

Labeling exclusions Models Microarchitecture

605 Fermi
610 Fermi
620 Fermi
630 * Fermi
640 * Fermi
645 Fermi
730 * Fermi
750 Maxwell
750 Ti Maxwell
640M LE * Fermi
670M Fermi
675M Fermi
710M Fermi
720M Fermi
820M Fermi
870M Kepler
880M Kepler
920M Kepler

* means that an exception exists for some modifications of these video cards.

  • GT this letter combination reflects low-performance video cards, they cannot be considered as gaming ones.
  • GTX this index denotes mid-range and high-end video adapters that are well suited for gaming.
  • M mobile video card (they are much weaker than their brothers without this letter)
  • X marking a more powerful video card in mobile solutions
  • LE this is the version of the card with a lower clock frequency for mobile adapters
  • Ti designation of a more productive version for desktop cards

It should be noted that the more efficient version differs not only in overclocking, but also in the core components (unified shader units, texturing units, rasterization units).

Perhaps, starting with the Pascal generation, mobile video cards will no longer be equipped with the letter M, since they use almost the same chips.

The numbers following the generation indicate the position of the model in the line.

Interesting fact: 90 means 2 chips 80 in SLI mode (working in pairs).


This is a sub-line of GeForce, because they have a GTX index. First you need to figure out the positioning of this ruler. These are the fastest and most expensive video cards at the moment. But this price is really too high for this level of performance. The thing is that they are also positioned as powerful professional video cards for mathematical calculations and FP 64 calculations (double precision floating point calculations). This is a kind of SUV in the world of video cards and you can work and play. The exception is Titan X, which lacks stars from the sky in FP 64 calculations and is in fact just a very expensive video card with a huge set of video memory.

In this line at the beginning of 2016, there are only 5 video cards and almost all of them are in the reference design (versions from third-party manufacturers).

TITAN, TITAN Black Edition and TITAN Z belong to Kepler, TITAN X Maxwell, there is also TITAN X in Pascal (the differences are in the consoles: the first has the full name NVIDIA GeForce GTX TITAN X, and the second just NVIDIA TITAN X).

You can read about all models on the NVIDIA website.


This family is intended for professional use. These cards are very well suited for complex 3D applications and computational simulations. These cards render real movies with special effects. These cards are not suitable for games. Even the most expensive solutions will outperform the average GeForce gaming graphics cards. The thing is that these video adapters are designed for double precision floating point calculations (FP 64), and single precision (FP 32) is enough for games. But in these more accurate calculations and applications using OpenGL drivers, the superiority of Quadro is simply colossal, because even cheap Quadro (although they are not so cheap) make the most powerful gaming video cards (with the exception of some Titans). If the application does not have any specific optimization for the capabilities of Quadro cards, then the result is decided by the number of stream processors and memory bandwidth, in which gaming video cards are in full order.

We will not touch on maps until the Fermi microarchitecture.

The first letter means the microarchitecture of the chip:

  • she’s not there Fermi
  • K Kepler
  • M Maxwell

There are also letter indices at the end of the model name:

  • M designation of a mobile video card
  • D another set of outputs. In the case of the K2000, instead of two DisplayPort and one DL-DVI ports, the D version has two DL-DVI outputs and one mini-DisplayPort.

The numbers indicate the position of the video card model in the line (more is better).


It has long been known that video cards (GPUs) do mathematical calculations much faster than processors (CPUs). The thing is that in these operations, the number of cores and the parallelism of calculations are of great importance. There are much more cores in video chips than in CPUs. If we take CUDA, then in one video card there can be up to 3072 pieces! This feature is associated with the role of the video card in the computer; it needs to do many simple actions in parallel and in a very short time. Tesla is a family designed specifically to speed up math calculations. Such cards do well with both FP 32 and FP 64 calculations. They are used in scientific centers and on servers, because per unit of consumed energy they will do more useful work than a processor. Interesting fact: the cards of this line do not have video outputs. The first letter means generation

  • K Kepler
  • M Maxwell

The numbers indicate the position of the chip in the ruler (bigger is better). Here we will not analyze the letter indices and marking details. Our goal is to introduce you to this family and tell you a few words about it. There is little information about the entire family due to its application only in narrow areas and the difficulty in acquiring.

Interesting fact: There is such a video card NVIDIA Quadro M6000 with 24 GB of video memory!


This family was created for the corporate segment. Previously it was part of the Quadro family and was also designated by the letters “NVS. These video chips are designed for business applications (financial, corporate, ECAD), multi-monitor solutions. For example, they are used for digital signage. They feature a large number of display ports on some models and a very low total cost of support (TCO). They do not shine with performance and they use not such fast DDR3 memory. Heat dissipation does not exceed 70 W. For comparison, the most powerful NVIDIA NVS 810 model has only 512 CUDA cores, TDP 68 W and 4 GB DDR3 memory, but as many as 8 Mini DisplayPort 1.2 outputs.


A family of systems on a chip (SoC) for mobile devices (there is a good article about SoC on our site). Within the framework of it, the first dual-core and quad-core solutions were presented. At the time of their release, they are top-end solutions in terms of graphics, but in the processor part, things are pretty good. At the moment they have their own Denver core designs instead of the “classic Cortex. There are two versions of Tegra K1:

  • 2 Denver cores
  • 4 cores Cortex-A15

Tegra K1 was built on Kepler microarchitecture, and Tegra X1 on Maxwell. Oddly enough, but Tegra X1 uses 4 cores Cortex-A-53 and 4 cores Cortex-A-57 (big.LITTLE technology). The advantage is that there are exclusive projects and ported computer games made only for Tegra-based devices, due to their power and company connections. Nvidia also has its own tablets and laptops that have some interesting technology. For example, broadcasting games from a PC to the screen of your Tegra-based mobile device. Tegra-based devices are a great help for mobile gaming.

Interesting Facts:

Tegra 2 was the first 2-core chip for mobile devices.

Tegra 3 repeated the success of the previous chip, but with 4 cores.

Tegra K1 switched to the microarchitecture of the Kepler graphics core and 28 nm technical process and came close in performance to the PS3 and XBOX 360.

Tegra X1 adopts big.LITTLE technology, switched to Maxwell microarchitecture and 20 nm technical process. Became the first system-on-a-chip to achieve 1 teraflops of FP16 floating point computing. To demonstrate the demo of Unreal Engine 4 “Elemental, he spent 10 watts, Xbox One 100 watts, and the top-end PC of 2012 300 watts. This does not mean that it can compare with the top-end PC in 2012, but only demonstrates a huge increase in the efficiency of video chips.


We hope that our article was clear and interesting for you, and also helped you understand the lines and markings of NVIDIA video cards. If you have any questions and inconsistencies, then first look into the Introduction, if the problem is not resolved and questions remain, then you are welcome in the comments!


5/5 - (1 vote)

Leave a Reply

Your email address will not be published.