How is a video card made? From VRM to VRAM, the anatomy of a GPU

A look under the body of modern gaming video cards to learn how to choose the ideal one for our PC.

phone book What is a video card like?  From VRM to VRAM, the anatomy of a GPU

Before embarking on what for many represents the most difficult challenge in the world of assembly, namely the choice of components, it is necessary to take a step back and try to acquire the necessary tools to know “to read“the data sheet of a video card.
In fact, a very strong heterogeneity exists in the GPU market which, however, allows us, with due care, to really be able to choose the right one for our pockets and our configuration.
We often hear about “video cards for FullHD at 60 FPS” you hate “GPU for 30 FPS in 4K“, extreme simplifications which, however, make the concept quite well. Each video card currently on the market has specifications that make them more or less adequate to perform a certain type of work.

Are GPU and video card the same?

The first aspect to be explored concerns precisely this acronym, very often used improperly to mean something that is not. To be clear, the GPU is not a video card just as a CPU is not a computer. As mentioned in our previous study, the GPU is the beating heart of a video card but it is not the whole. A video card is in fact a small system, separate and separable from a computer, which houses inside a series of fundamental components capable of communicating with each other in order to obtain an output, or video signal.
The GPU is the beating heart of the video card, made up as a whole by a printed circuit, a graphics processor, VRAM, VRMs and the actual power supply, as well as video outputs and a communication channel with the PC consisting of the PCI Express port.


The GPU, or Graphics Processing Unit, is to the video card what the processor is to a PC. But a certain amount of caution is needed before venturing into comparisons as easy as they are distant. CPU and GPU are profoundly different in architecture, function and potential.
The fundamental difference lies precisely in the architecture. A CPU is an extremely performing computer, capable of hosting a limited number of cores, which in turn are made up of billions of transistors. The cores of a processor are limited in number but capable of executing extremely complex instructions in series. The potential of a CPU is in fact expressed in the profound heterogeneity of complex tasks it can perform and not in the amount of calculations it is able to perform simultaneously. This is instead what characterizes the GPU.

SEE ALSO  Telegram is becoming more popular among mass media, pushing television and Internet sites into the background

A graphics processor differs from a CPU in that it has an innumerable number of tiny cores. In modern GPUs the most adequate scale is that of several thousand cores, able to perform simple and repetitive calculations, in parallel and at very high speeds. In the latest generation of NVIDIA GPUs, these small computers are called CUDA Core. In AMD graphics processors, on the other hand, the compute units are called Stream Processors. However, the two terms are not interchangeable and comparable.

It will therefore be quite useless to compare the number and frequency of the Stream Processors of an AMD GPU with the CUDA Cores of NVIDIA, because they completely differ in architecture. For this reason, in recent years, manufacturers have begun to use, and sometimes abuse, another unit of measurement, certainly more functional for the purpose, but that is not necessarily a litmus test: we are talking about FLOPS, that is the quantity of floating point operations per second (FLOPS) that a processor is capable of.


Performing many calculations per second is therefore the raison d’ĂȘtre of a graphics processor, which requires this particular specialization to animate every single pixel of our monitors with hundreds of frames.
To make the visual experience as fluid and complex at the same time, the GPU needs a space to temporarily allocate a part of the calculations it is able to perform.
For this reason a video card is equipped with a variable amount of VRAM memory, which has multiple purposes. First of all is precisely that of framebuffer, or a transit area in which already processed information is stored and intended for video output.
The other essential function of VRAM is that of preload portions of the graphic material useful for final processing, such as game textures.

The question at this point arises: is it possible to say that it is the VRAM and not the GPU that affects the performance of the video card? The answer is obviously no. Over the years VRAM has become a real workhorse in marketing campaigns. In reality this value, in order to be useful in a manner consistent with its potential, it needs a calculation capacity that is supportive of it. In most cases, in fact, it is the manufacturer itself to insert an amount of VRAM memory adequate to the capabilities of the GPU.

The VRMs

Acronym for Voltage Regulator Module, VRMs are small power systems that limit and stabilize the flow of current coming from the main power supply, so as not to damage the video card.

SEE ALSO  New Fitbit Charge 6: the smart bracelet is now more precise and brings Google Maps to your wrist

The VRMs are made up of three different structures, which cooperate with each other to obtain the final result. MOSFETs are the first element in this “cascade”. They are interposed between the main power supply and the rest of the board components e have the task of delivering the current at the voltage required by the GPU. The Chokes in turn have the purpose of stabilizing the output current and finally there are the capacitors, small cylindrical structures responsible for collecting the current, like small tanks, and for the subsequent final injection into the circuit, in order to eliminate or limit the as possible any voltage fluctuations.

Inbound and outbound communication

Complete all the rendering and of rasterization, the image is sent to the final device, the monitor. For this reason, in the portion that faces the outside of the case we will find a plethora of video outputs. In modern graphics cards we will mainly find a combination of HDMI and DisplayPort outputs.

The video card is a device that allows the externalization of some specific calculations outside the main processor. Current technologies allow fast and efficient communication with the rest of the system thanks to the PCI Express standard.
Currently at the 4.0 iteration, the intergenerational differences concern the data transfer speed and the bandwidth of the bandwidth. To give you an example, PCI Express 3.0 supports a transfer rate of 8GB / s and a bandwidth of 32GB / s, while modern 4.0 channels reach 16 GB / s of speed and 64 GB / s of bandwidth.

As astronomical figures may seem, in reality there are extremely few devices able to settle for a single communication channel, such as some sound cards. There are numerous PCI Express ports of different sizes on all motherboards. Each PCI Express slot will communicate a peripheral with the motherboard chipset or directly with the CPU with a specified number of PCIe channels. In particular, modern video cards will communicate through a x16 slot, that is, consisting of sixteen PCIe channels.

The dissipation system

The aesthetic artifice, as we know, is at the total discretion of the user. Many times, however, it is good to try not to get the pill golden from a particularly aggressive shell or a handful of LEDs. In fact, it is necessary to pay particular attention to what lies behind the dissipation system.
Some manufacturers also add liquid-dissipated video cards with systems to their lineups all-in-one or with waterblock to be inserted into a Custom Loop. The performance of these solutions is notoriously excellent.

The vast majority of video cards, on the other hand, have traditional air cooling. In this area too, a distinction must be made. There are in fact type heatsinks Blower, less and less present on the market, and heat sinks Open-Air.
Blower-type video cards feature a single tangential fan, placed at the opposite end of the board with respect to the video outputs, as its function is to capture air from the surrounding environment and push it longitudinally through the metal dissipation channels, to cool them and allow the heated air to escape through the holes ports located near the HDMI and DisplayPort ports. The usefulness of these systems lies in the fact that, under conditions of airflow limited, do not put heat inside the case or do it in a limited way. The other side of the coin is that they will be slightly warmer than Open-Air systems.

Most of the custom video cards, but now also the reference cards of AMD and NVIDIA, have an open-type cooling system, recognizable by the presence of multiple fans on the body, whose function is to enter larger volumes in the heatsink of fresh air, for better dissipation, however, dispersing the heat into the surrounding environment. Overall, a slight increase in system temperatures could be observed, but with good airflow there should be no particular limitations. Unlike blower systems, the greater volume of air in this case allows the GPU to work at significantly lower temperatures.

SEE ALSO  These Xiaomi phones are the most sought after in Spain: they cost less than 300 euros and are not from this year

A thoughtful choice

The choice of the ideal components requires an essential theoretical basis. Being able to understand the system you want to assemble can be a particularly complex node to solve, but if we know how to choose the right video card, tune the rest of the components synergistically will be child’s play.
On the contrary, if we already have a monitor and a partially configured system, knowing how to choose the right video card for our system will allow us to invest the right capital, especially in a context like the current one, where supplies are extremely limited, or if you have a limited budget available.