GetToKnowYourGFX_intro2.jpgImageWhen we take a look at the situation a few years back, a large number of educational texts concerning the basics of motherboard functioning meant that even irregular followers of the IT industry eventually bumped into one of them. On the other hand, graphics cards are somehow always left out, reviewed, but without much explanation for various expressions, acronyms and terms that make most “newbs” startled. Having understood this, we’ve decided to deal with graphics card in the theoretical sense a little. This is, after all, one of the vital (THE vital, if you’re asking the author of this article) system components, the elements of which should be known at least roughly. Since it would be an incredibly arduous task to take pictures of each element on every type of card, we’ve chosen a single referent model - Zotac GeForce GTS250. This card has a rather typical design, therefore ideal for our intentions.


First of all - the cooler, since this is the thing that you’ll notice first. Ten and more years ago, the cooler, or should we say, cooling system in any form, wasn’t mandatory, and chips used to be “naked”. Today, that is completely unimaginable, with this many elements placed into a single GPU, so even the cheapest and weakest of models have some sort of cooling system, at least a passive piece of aluminium. Our GTS250, for example, has a pretty serious cooling system, consisting of an aluminium cooling profile and a fan. Fans come in many variants, from the classic ones to turbine ones. As you may have already seen, ours is the latter. Of course, things get ugly with so many components cramped in one place, so manufacturers usually armour their coolers with plastics, which often serves an additional purpose. The plastic lid directs airflow over the ribs of the cooling system, sending hot air outside the case, in order not to additionally heat other components. This isn’t exactly a fixed rule, and the lid is sometimes present merely for the aesthetic function, especially with double-slot coolers (those occupying the adjacent slot as well as its own). Alongside everything already stated, the cooling profile is often made of copper, one of the best heat conductors, and heatpipes are an increasingly common phenomenon on graphics cards as well, spreading the heat evenly along the surface of the cooling system.


Graphics card with the cooling removed

DVI outputs



Before we take the cooler off, let’s have a look at another visible feature - graphics outputs, used to transmit image to other devices. This is the only part of the graphics card located outside the enclosure. The usual connections are D-Sub, i.e. VGA, which is an analogue monitor output. Since such monitors are losing ground to newer standards, digital outputs on graphics cards are getting more and more numerous. You’ll hardly find a graphics card on the market that doesn’t have a DVI output, for example, and our referent GTS250 has two of them. Another possibility is HDMI, one of the younger interfaces (the youngest if we’re looking at the most recent 1.4 standard, certified last summer), which some monitors prefer over DVI, and which is practically the only one installed in LCD and plasma TVs, while new devices, such as Blu-Ray players, A/V receivers or even new generations of DVD players contain an HDMI input/output. One huge advantage of HDMI is that it can transfer both video and audio through a single cable, which is of immense importance when interconnecting home devices such as TVs, but other devices as well. It’s good to know that DVI and HDMI are partially compatible, which means that video and audio will transfer both ways with zero degradation in quality. Outputs come in various combinations, which means that you could stumble upon a graphics card with an optical audio or an S-Video output. The final and least common output is DisplayPort. Unlike old connections such as D-Sub, S-Video and similar, this connection type is in expansion. It is genetically the same age as HDMI, and shares a multitude of features with it, such as audio and video transfer through the same cable, but is nevertheless missing support for Dolby TrueHD and DTS-HD Master Audio, as well as a few other, less relevant things. The list of manufacturers supporting DisplayPort is longish, and the finest example is ATI, with their Radeon HD 5870 containing six (!) mini DisplayPort connectors. Besides a few technical advantages, the fact that DisplayPort is a royalty-free standard is perhaps the main impulse behind so many manufacturers supporting it.

Below the cooler

Having removed the cooler, we face the PCB and all of its elements. If you haven’t managed to pick it up from motherboard reviews, PCB stands for Printed Circuit Board, with the role of housing and interconnecting electronic elements, as well as provide power input, using prefabricated pathways.


PCI-E voltage connector

SLI connectors



You’ll immediately notice that one of the chips is different from all the others in its size and looks. This is the GPU - the graphics processor. When we refer to various codenames, such as G92, G80, GT200b, RV770, Cypress, Hemlock etc. this is the chip that we’re addressing. Naturally, various GPU models can be very different in appearance, but they are always easily distinguished. Stream processors, texture and ROP units, implemented video capabilities and various engines are packed among thousands and millions of transistors situated in these chips. All this complexity and high clocks are the main guilty party for the massive cooling systems needed to keep them under control.


There is usually a regularly shaped “suite” of memory chips around the GPU. GTS250 has four pairs of chips. A total of eight chips on separate 32-bit memory buses, used in a standard way (32-bit is the standard chip configuration), add up to a 256-bit graphics card. This can be applied in almost any case, bar when the chips are located on both sides of the PCB, because that means that two chips are sharing a 32-bit bus. Therefore, a double-sided PCB with four chips on each side (a total of eight) would make for a 128-bit graphics card. Another relevant piece of information is whether the memory is SDR (which is highly unlikely, to be honest), (G)DDR(1-4) or GDDR5. The bus is effectively doubled when the latter is used, since GDDR5 does double the effective I/O throughput in comparison with older memory generations. As far as GDDR4 is concerned, this is a standard which was abandoned very fast. ATI implemented it in a few graphics card models, while Nvidia skipped it completely. GDDR4 was thought to offer lower temperatures, required voltages and clocks, but “relaxed” latencies and chip structure, as well as marginal performance improvements, soon pushed it into oblivion.


GPU with the surrounding memory

Graphics Processing Unit (GPU)


Voltage section

The voltage section is a very important part of the graphics card, because it deals with “ironing out” input voltage and providing a stable one; voltage regulators are the key here, and their number depends on the needs of the graphics card. Capacitors and resistors are also present, which only goes to prove how many voltage filtering phases modern graphics cards have. In the case of our GTS250, the GPU has a 3-phase voltage filtering, while one more phase is in charge of the memory. All capacitors are shielded, which should theoretically prolong their lifetime, as well as higher endurance against high temperatures. The higher the quality of the voltage section elements and the better their shielding against high temperatures, the less chance there is for the graphics card to break down due to long-term overclock or high voltages, whether for the GPU or the memory. Naturally, the overclock potential is dictated by the GPU itself, which may or may not be a good sample, and the same goes for the memory, while the voltage section only contributes to greater stability, especially in the long run.

Voltage connectors

Connectors may or may not be present on your graphics card, and they are usually located on the edges of the PCB. Depending on the graphics card’s needs, if the consumption goes over 75 W (PCI-E) or 150 W (PCI-E 2.0), the manufacturer will install molex connectors. Although there are several types of these, by far the most common one is the one you can see in the picture - the 6-pin molex connector. You may stumble upon a 4-pin one here and there, but these are usually found on long-gone AGP models. If required by the card, you may even find two molex connectors on one card, mostly in high-end models craving power, and one of those may very well be an 8-pin, especially on the very strongest cards, since those are the most taxing ones as far as electricity bills are concerned.


4-pin fan connector with PWM regulation


PCI Express connector


PCI-Express and SLI/CrossFire

These terms refer to communication between a graphics card and the motherboard bus, as well as internal communication with another graphics card in the same system. On the bottom of each card, there is a PCI-Express connector, installable at an appropriate slot on the motherboard. Three standards are currently in existence for this sort of connection, one of which is nearly extinct. The latter is AGP, which is all but obsolete and located only in older PCs. The PCI-Express is the standard, and the difference between the first and current 2.0 versions are visually undetectable. The difference consists in a doubled throughput per line compared to the previous version. Maximum deliverable power has also been doubled, which means that a PCI-E 2.0 slot can deliver up to 150 W to the graphics card, unlike the first generation, only going up to 75 W. On the upper side of the PCB, you’ll find a small connector or two, used for interconnecting cards in SLI mode, in our case, or CrossFire, in case of ATI cards. Each of the two technologies has its own way of functioning, but intercommunication via the foldable cable attached to the upper connector on the PCB is typical of both manufacturers. This feature is far from mandatory and present only on some cards, which means that it’s more of a bonus capability. It should also be noted that ATI models aren’t always necessarily interconnected this way to work in CrossFire - sometimes, the motherboard is able to provide enough room for this type of communication.

We hope that you’ve got a far better knowledge of what a graphics card consists of after reading this article. Naturally, different models mean different looks and element placement and configuration, but many of these features truly are inherent to all cards, so you’ll be able to discern the characteristics such as the GPU, memory, voltage connector types etc. easily.