How to understand the graphics card specifications

With the announcement of each new graphics card, all its characteristics and specifications are revealed, a lot of data that some users will not know how to interpret, so they cannot assess what the new card is capable of offering them. To solve this problem we have prepared this post in which we explain all the important specifications of a graphics card.

Index of contents

Graphics card models

First of all we are going to talk about graphics cards for laptops, they are easy to identify because they almost always have the tag "M" in their name. We almost always say because with the arrival of the GeForce 10 series this tag has been removed that we will find in its previous generations and in all AMD cards. Nvidia and AMD are the leading manufacturers of PC graphics cards.

The next step is to identify the generation of the card, for this we will look at the first numbers, the higher they are, the more modern the card and generally more powerful. Examples:

• GeForce GTX 10 60GeForce GTX 6 60AMD Radeon RX 5 80AMD Radeon RX 4 80

The following numbers represent the order or tier of the graphics card and the higher the more powerful it will be. Examples:

GeForce GTX 10 80

GeForce GTX 10 50

AMD Radeon RX 5 80

AMD Radeon RX 5 60

Buying these numbers is only reliable if we are talking about cards from the same manufacturer.

Connection to motherboard

Today all graphics cards use the PCI-Express 3.0 x16 connection to communicate with the motherboard so at this point there is not much to say. We may find an older card with a PCI-Express 2.0 x16 connection, if so it will be more than enough for it.

We recommend reading the best graphics cards on the market

If we have to pay attention to connect it to a PCI-Express 3.0 x16 port on the motherboard since sometimes they also put PCI-Express 3.0 x8 ports that are physically the same. The second one we will only use in case of putting two cards. We can differentiate them with the naked eye by looking at the contacts.

Maximum image resolution

The maximum resolution represents the highest number of pixels that the card can draw on the screen, this depends on the connection and the highest is usually for DisplayPort, the most used connection is still HDMI. A pixel is each of the points that make up the image, there are millions of them. There are also DVI and VGA connectors although they are used less and less.

Core speed

The speed or frequency of the core is represented in MHz or GHz and indicates how fast the card works, the higher the speed, the higher the performance generally. Higher speeds mean more power consumption, so more powerful cards consume a lot of power to run. Two speeds are normally indicated, the base and the turbo.

Graphics card chip size

The size of the chip or GPU represents how large it is physically, this size is measured in mm2. The bigger a chip is, the more complex it is, the more elements it contains and therefore its performance is greater.

Stream Processors (AMD) or CUDA Cores (NVIDIA)

These represent the number of execution units that are inside the GPU, these units are the ones that do the work and therefore the more of them there are, the more powerful the card will be. AMD and Nvidia use very different designs, so comparing this data is only reliable if we are talking about cards from the same manufacturer.

Generally AMD needs more than Nvidia to achieve the same performance, for example the GeForce GTX 1060 has 1, 024 CUDA Cores while the Radeon RX 580 has 2, 048 Stream Processors and its performance is very similar.

ROPs and TMUs

They are the crawling and texturing units respectively, these units are in charge of putting the pixels on the screen, applying the textures and various additional tasks. We can say the same as in the case of Stream Processors and CUDA Cores.

AMD vs Nvidia: the best cheap graphics card

Texture and Pixel Fillrate

The Texture Fillrate indicates the pixels that are textured and rendered per second, on the other hand the Pixel Fillrate measures the number of pixels that the GPU can draw per second. The higher they are, the more powerful the card will be in general. They are measured in GTexel / s and Gpixel / s respectively.

Graphics card TFLOPs

TFLOPs represent the maximum power a GPU can offer and take into account frequency, Stream Processors / CUDA Cores (NVIDIA), and ROPs and TMUs. It measures the operations that the card can do per second, the most powerful cards reach 12 TFLOPs or something more.

WE RECOMMEND Game Ready 388.71, the new NVIDIA drivers are available

Amount and type of memory

The memory of the graphics card is used to store the data it is processing and access it very quickly, as we increase the resolution and the graphic detail increases the consumption of this memory, so it is important that it does not fall short. The amount of memory required varies depending on the power of the card, many times several versions of the same card are offered with different amounts of memory, in these cases it is safest to opt for the version with the largest amount, although sometimes it is excessive and does not provide benefits. The amount of memory is measured in GB and ranges from 2 GB on the lower end cards to 12 GB on the higher end cards.

Another important fact is the type of memory that is related to its speed, if we order them from fastest to slowest we have:

1. HBM / HBM2GDDR5XGDDR5GDDR4GDDR3

Memory frequency, interface and bandwidth

The memory frequency is measured in MHz or GHz and its interface is divided into bits. Both data represent the speed at which the GPU can access the stored data and is just as important or even more than the amount.

To understand it easily we can imagine a highway in which the number of lanes are the bits and the speed of the cars is the frequency. The more lanes (bits) and the higher the speed of the cars (frequency), the more number of cars can circulate each second.

The current GDDR memories reach speeds of 11, 000 MHz and interfaces of up to 512 bits, in the case of HBM and HBM2 they reach approximately 1, 500 MHz and 4, 096 bits.

Bandwidth takes into account the memory interface and its speed, it is measured in GB / s and it is the value that really matters in terms of performance. The best cards can exceed 500 GB / s

TDP, consumption and power pins of the graphics card

The TDP is a measure of the heat generated by the card during its operation and is closely related to consumption, although it is not the same, both are measured in W. The motherboard can only give 75W of current, so the most powerful cards need auxiliary connectors, there are 6-pin that can give up to 75W and 8-pin that can give up to 150W. The most powerful cards can reach 300W or even slightly exceed them.