Wednesday, October 14, 2015

What you should know when picking the right GPU

Picking the right GPU

Graphics card technology can be confusing, so let’s break it down

The market for graphics cards is large and, in having two large companies dominate, similar to the CPU market in many ways. Buying a graphics card can be even more confusing, though, because of generational overlap (and the occasionally resurfacing) of technology; not to mention the fact that the capabilities of similarly named models can differ wildly. This has been even made worse by the fact that Radeon recently rebranded its entire line, meaning there are two different numbering systems for just one company.

For all those reasons, the best way to make sure you find the graphics card you want is to go right back to the basics and learn how to read the specifications for yourself. So, to make sure you can do that to the necessary extent, here are our explanations of graphics card specs and what you need to look for when you’re trying to buy a new one.

Picking A Manufacturer

There are two major companies producing graphics chipsets – AMD and Nvidia. Both manufacturers has its own line of cards, each with their own advantages and disadvantages. As a consumer, the good news is that neither manufacturer is substantially better than the other in any universal way. The bad news is that this means you have to base your choices on lots of small differences instead, as with CPUs, on a couple of big ones.

As a general rule, Nvidia cards are slightly faster, but AMD cards are slightly cheaper, so you pay about the same amount for performance. At the very top end of the market, Nvidia cards are faster – but then there’s over a thousand dollars of extra money separating the fastest AMD card – the Radeon R9 295X2 – and the fastest (and faster) Nvidia card, the GeForce GTX Titan Z. At the point where most consumers are going to look for cards, you could flip a coin to choose and not end up in any major problems. Functionally, performance will be indistinguishable from game-to-game.

If you’re running an AMD-based system already, it’s more likely that your motherboard and CPU will have features that can combine with AMD’s Radeon graphics cards, and the same is true for Intel systems and Nvidia GeForce cards – but essentially, the two brands are equivalent to one another.

Sponsored Products Related To This Article

We’ll get into the specific technicalities further on in this article, but for now we recommend that you start making your choice based on what the best card you can fit in your budget is. With graphics cards, you can usually squeeze out a fair chunk of extra performance by spending about $50 more, so establishing a point at which you will absolutely not pay any more is the best way to make choosing a card easier. Otherwise you’ll end up with a $550 power-draining beast that is nonetheless technically better value than the $240 card you started looking at.

Once you’ve pinned down the maximum price you want to spend, you can start investigating the actual features of the card instead. Does it have the right ports you want? Does it have one or two fans? Is it double-height or not? Those factors are all more important to you, ultimately, than the manufacturer you choose.

It’s also worth noting that the structure of the graphics card industry is a little unusual. Although AMD and Nvidia design and produce the GPU chips that power their cards, both companies sell those GPUs to other manufacturers, who produce their own versions of the cards based on AMD and Nvidia’s specs. The original design of the card is called a ‘reference’ version, but card manufacturers – such as Sapphire, XFX, MSI and the like – may add features not seen on the reference board. Extra ports, additional fans, and factory overclocks are just three things you might get from one manufacturer but not at another. So even after you’ve picked a card model, you may find yourself comparing two competing versions of the same product. In general, the differences between these cards is miniscule – we recommend you stick with a manufacturer you’re familiar with unless you have good reason not to, just for peace of mind.

Graphics Card Hardware

To select a graphics card, you need to be aware of how the hardware relates to its performance – and there are many factors affecting this.


The graphics card’s chipset refers to the reference design it is based on. All cards with the same chipset have broadly similar capabilities and can be directly compared with one another in ways that other cards can’t. If they have the same chipset they run on the same architecture, so it’s comparatively easy to figure out when one of the cards is better – it’s just the one with higher numbers!

If nothing else, it’s worth paying attention to the chipset’s process design, which is given in ‘nm’ (nanometres). As with CPU architecture, the smaller this value, the more energy efficient it is, and the faster the GPU can run.


A graphics processing unit (or GPU) is the processor of your graphics card, and the bit that performs the rendering and calculations for graphics. Most cards quote two speeds: the Base clock speed (which the card normally runs at) and the Boost clock speed (which it temporarily switches to at times of high load). These values are given in MHz, similar to normal CPUs, and dictate the number of updates that the GPU can make per second.

The confusing thing about clock speeds is that they can’t be directly compared to check performance unless the chipset of the cards is the same. A card with an 800MHz clock speed can be faster than a card with an 1100MHz clock speed if the architecture is sufficiently different!

If the chipset is the same, you can use clock speed to compare card performance to an extent, however. Cards with a higher clock speed perform more operations per second, so it’s not uncommon for manufacturers to perform a ‘factory overclock’ and run it slightly faster than the reference design – particularly if they add extra cooling to increase the hardware’s temperature tolerance.


Memory is also important in determining speed. It’s something of a misconception that extra graphics RAM will markedly improve performance – having the minimum amount of necessary RAM available is important, but if you have more than that amount it won’t have any real effect – so don’t go for 8GB unless you have good reason. More important is the type of RAM. GDDR5 is faster than DDR3, but also a more expensive.


This determines the amount of data that can be shifted in a single memory cycle, and is affected by two factors: the memory clock rate and the memory bus size. Unlike GPU clock speeds, you can compare them across different cards as long as the RAM type is the same. The memory bus size is given in bits, and more bits means better performance. Most modern cards are likely to use at least a 128-bit memory bus, but you may see as high as 512-bit, with performance directly proportional to size. As a rule of thumb, you can quickly compare available memory bandwidth by multiplying the bandwidth by the clock speed. The bigger the amount that comes out, the better.

Standards Support

As well as the hardware capabilities, graphics cards listings will show the technologies they support, some virtual, some physical. Knowing what they mean will help you decide how relevant they are to you. For example, all cards will quote, somewhere, which version of the two main graphics APIs they support. The latest version of DirectX is 12 and the latest version of OpenGL is 4.5, though graphics cards still only support OpenGL 4.4 as standard. Don’t worry too much if the card you’re looking at only supports a slightly older version. Most of the time this won’t result in any notable performance decrease, it’ll simply block off access to some of the latest effects and shaders, but it’ll take years to go entirely out of date.

Support for multi-card operation – called SLI by Nvidia and CrossFire by AMD – is only of any importance if you plan to run multiple cards in tandem. To run cards like this, the GPU needs to be the same (or have very similar underlying technology) which means that most SLI/CrossFire users buy two cards together. Some low-end cards explicitly will not support SLI and Crossfire, but in general, any card and motherboard at the mid-to-high end of gaming will.

The number and type of hardware interfaces is also worth paying attention to. Here’s what you might find on a modern card:

Dual-Link DVI-I / DVI-D refers to a DVI interface with extra pins that allow it to reach resolutions of 2560 x 1600, instead of the single-link maximum of 1920 x 1200. DVI-I is a combined analogue/digital port, while DVI-D is digital-only. Your card may have both, or just one.

HDMI is a high-definition audio/video interface available in several different versions. The most recent pair – HDMI 1.4 and 2.0 – both support 4K video and 3D video, but HDMI 2.0 is the only to support 4K in 3D. HDMI 1.4 also only supports 24Hz refresh rate for 4K and 3D video, whereas HDMI 2.0 supports the full 60Hz.

Display Port is a video interface designed to replace VGA and DVI. With adaptors it’s backwards-compatible with both ports, and can also be used to carry other data such as audio and USB signals.

Usually, it’s possible to utilise several of these ports at once – sometimes up to four, depending on how many there are – so check the maximum number of supported screens. Note that some chipsets support more screens than the card can physically accommodate!

Current Models

Selecting the best chipset to look for within your budget means you’ve always got at least two decent options – one Nvidia and one AMD. It’s also important to be sure you’re actually getting a substantial upgrade from any integrated technology. So with that in mind, here are the best chipsets at common ps, and how they compare to one another.

Budget Cards: Under $130

The GeForce 700-series may have been succeeded by the 900-series, but if you’re looking for a gaming-capable budget buy then the GeForce GT 740 remains your best choice. Specifically, we liked the look of the Gigabyte GT 740, with its 1072MHz clock speed, 1GB of GDDR5 RAM, max resolution of 4096 x 2160 and Gigabyte WindForce cooling system. Interfaces include dual-DVI, VGA and HDMI.

The only sticking point is that it’s noticeably slower than the AMD equivalent, the R7 250X. It does have a significantly lower TDP (65 watts) making it excellent for low-power systems, but in almost every other way (including price) an R7 250X is better.

In terms of the Radeon’s previous numbering system, the R7 250X falls somewhere around a GDDR5 HD 7750 (which it’s a little faster than) – but 7000-series cards are largely off the market now.

Casual Gaming Cards: Under $200

The best Nvidia cards at this price are the GeForce GTX 950, which you can find with 2GB of GDDR4 memory. The best Radeon cards are the R7 370, which also have 2GB of GDDR5. The Nvidia cards are actually a little cheaper (but only by around £10), but this makes their general performance superiority much better to take.

gtx 950 and gtx 960 graphics cards

Released this August, the GTX 950 is a cut-down GTX 960, and offers almost as good performance at a much lower price. It uses less power than its AMD equivalents despite performing much better, and it’s capable of running even new games in full HD. For this type of use, the R7 370 just about loses out, and when you’re looking at gaming that’s enough to turn us off.

Mid-range Cards: Under $300

Again, Nvidia are the clear winner at this price with the Geforce GTX 960, the 4GB GDDR5 versions of which top out in the mid-£190s. AMD, by comparison, offers either the R9 380 – which is more like $270, or the R9 390 which is priced around $420. If you could still get the R9 280X, Nvidia might have had a fight on its hands, but it’s disappearing from shelves fast. Besides which, the GTX 960 is still less power-hungry, cooler and quieter.

Despite similar model numbers the R9 380 is about 10% less powerful than the R9 280X in just about every way, which means it can’t compete with the GTX 960.

High-end Cards: Under $500

Finally, Nvidia has a proper fight on its hands. At this end of the market, the Geforce GTX 970 is Nvidia’s best card, priced at $440. The Radeon R9 390 is priced bang on $470, but the performance is proportionally better and it has double the memory. If raw performance matters most to you, the Radeon card is worth every penny. But other factors may sway you on the GeForce.

For example, it’s a much less power-hungry card, and it runs much quieter as a result. If you aren’t doing 4K gaming you don’t need the extra capacity, and might prefer the extra bucks. The Radeon is effectively future-proof, though, and unless you really need that extra bucks (in which case, we humbly suggest a $470 graphics card isn’t in your best interests) then the Radeon ultimately wins out.

Unfortunately, the number of different models of card available make it difficult to be more specific that this, not least because availability varies wildly – but hopefully knowing what chipsets to look for can help you start figuring out which specific card fits your needs.

By MicroMart

Sponsored Products Related To This Article
PC matter, learn about your pc needs, software and peripheral devices
Share This
Subscribe Here


Post a Comment