Guide: Choosing Between Integrated or Dedicated Graphics

Notebooks have evolved at an unprecedented rate. While improvements in processor technology have certainly paved the way for OEMs to fill the internals to the brim as much as possible, one component gets to decide the purchasing decision more than anything else these days - the graphics processing unit (GPU).

With software, (including OSs) relying heavily on the GPU to push pixels to offer richer user experiences, it has become all the more important to pay attention to the built-in graphics capabilities of your next notebook or ultrabook. What makes this even more intimidating is that there are more choices for notebooks and ultrabooks these days than you might imagine.

In this article, we will break down the pros and cons of both integrated and dedicated graphics processors to help you decide which best suits your needs before buying your next laptop.

AMD laptops are generally more affordable than their Intel counterparts and you should be considering them if all you need is a machine that you can depend on for your day-to-day tasks without breaking the bank. However, we would suggest that you wait till the latest AMD mobile APU (Accelerated Processing Units) based notebooks launch in Q3 2017 if you’d like to go with Team Red.

Integrated graphic cards covering a large bandwidth of performance capabilities

All in all, opt for integrated graphics solutions if you value portability and battery life at much more affordable price points over improved performance and you have no need to play the latest games or play older games at full details at 60 frames per second.

A few words of caution, though: check the specifications of the laptop or ultrabook and make sure the integrated GPU can work for your needs. Remember that the integrated GPU uses system memory as a frame buffer and ultrabooks generally come with Low Power DDR (LPDDR) RAM and Ultra low voltage (ULV) CPUs, which can bottleneck the integrated GPU. It is always a good bet to know the class of processor and type of RAM used so that you can directly correlate that to the integrated graphics performance.

If you are looking at anything remotely graphic intensive, such as multiple-timeline editing in Premiere Pro, applying Photoshop filters across multiple layers, 3D modelling, or a decent gaming experience, you are better off with dedicated GPUs. They come with dedicated graphics memory that frees up system memory for other tasks. Just like CPUs that come with dual or quad cores, dedicated graphics cards come with many cores that enable parallel processing of visual data. NVIDIA calls them CUDA (Compute Unified Device Architecture) cores while AMD refers to them as ‘Stream Processors’. A GPU can contain hundreds or thousands of CUDA cores or Stream Processors depending on the requirement which is directly proportional to raw power. Dedicated notebook GPUs have evolved very much over the past few years with the latest NVIDIA notebook 10-series GPUs based on the Pascal architecture showing comparable performance to their desktop brethren.

While AMD is yet to revamp its line of Radeon notebook GPUs, you can find models which include the Radeon R5 to R9 which offer similar performance to an NVIDIA GeForce GTX 965M or 970M. Notebooks featuring the NVIDIA GTX 1050 and 1050 Ti are more than capable for video editing and moderate gaming. You will have no hiccups in playing titles such as Gears of War and Overwatch in high settings at 1080p, but you can forget about 4k gaming or playing the latest AAA titles on maxed settings if you want a smooth 60 frames per second. Thin notebooks such as the Aero 15 feature a GTX 1060 and now with Nvidia’s Max-Q initiative, a GTX 1080 can be fit into lighter and more portable laptops. The Asus Zephyrus ROG announced at Computex 2017 is one such example which implements Max-Q along with a GTX 1080 card — but it is important to remember that a “Max-Q” card will not be running at the performance level of its laptop or desktop analogue. Most of the laptops that feature the 10-series Pascal cards, come with a Core i5 or Core i7 HQ series processors and DDR4 memory clocked at or above 2400 MHz to make sure that the GPU is not bottlenecked. The GTX 1060 and above GPUs are also Virtual Reality (VR) capable which is something to look for if you are using VR headsets such as the Oculus Rift or the HTC Vive.

Dedicated graphic cards are starting where integrated GPUs give up

While a laptop with a dedicated GPU sounds like utopia, dedicated graphics cards have their own share of cons – the biggest trade-offs being significantly reduced battery life if the dedicated GPU is used over the integrated one and increased thermal output. Dedicated GPUs need their own cooling systems and more often than not, these notebooks tend to heat up rapidly especially during intense gaming sessions and 3D rendering work. To further muddy the waters, some manufacturers fail to properly engineer their products to handle the combined heat output of their components, especially with small and slim laptops. Rather than using the appropriate cooling solutions, some OEMs resort to thermal throttling, which severely impacts performance. One of the many examples of this is Dell’s XPS 15 series, which are frequently seen to throttle down to 800MHz under sustained gaming, reducing performance to a slide show. And then there is of course the portability aspect. Do not expect battery power to last for more than a couple of hours with high-end configurations with no support for switchable graphics.

Theoretically, switchable graphics are great. They let you use the integrated graphics for basic tasks and automatically switch to the dedicated GPU on-demand. NVIDIA calls this technology, Optimus while AMD calls it Enduro. The idea is to maximize battery life when not running graphic intensive applications. You might want to check if the notebook comes with Optimus enabled. The integrated graphics, in  this case, becomes the main display controller while the dedicated GPU is ‘masked’ behind it. Therefore, the dedicated GPU has no direct access to the display. High-end gaming laptops with switchable graphics tend to come with a hardware switch called the multiplexer (MUX) which lets you switch between integrated and dedicated graphics. Some laptops also directly connect external monitors to the dedicated graphics card. Notebooks enabled with NVIDIA Optimus mostly lack a dedicated hardware multiplexer. Optimus is also counteractive to NVIDIA’s own G-Sync technology. G-Sync enables adaptive refresh rates (up to 120 Hz) on compatible displays to minimize screen tearing and other artefacts during intense gaming sessions. Since G-Sync requires a dedicated access to the display, notebooks that come enabled with NVIDIA Optimus cannot support G-Sync on the primary display. The only option in that case, would be to hook up an external display via DisplayPort or HDMI to enable G-Sync (if the port is directly connected to the GPU).

So, are eGPUs finally the answer to our notebook graphics woes? Not entirely. Remember that although many notebooks these days come with Thunderbolt 3, it is important to note the transfer rate of the port. A Thunderbolt 3 port with 4 dedicated PCIe lanes (full 40Gb/s bandwidth) is much more suited for eGPUs than ones with just 2 lanes as the bandwidth is effectively reduced in half. Additionally, the TB3 interface itself causes a performance hit even under the most optimal conditions, as it does not directly connect to the CPU. This will likely be improved in the future, as Intel has announced plans to integrate the TB3 controller directly to the CPU. Another major point of consideration is price: often you need to pay separately for the eGPU case and the video card which can get expensive. eGPUs must be considered only when your notebook comes with just integrated graphics or low end dedicated graphics. Installing a high-end graphics card such as the GTX 1080/Ti in an eGPU to augment a notebook that is already running say, a GTX 1060 is not worth the extra dough and moreover, results in compromised performance. To quote from the conclusion of our Razer Core analysis:

All in all, the Core does what it says it does: it allows the user to attach an external graphics card to a computer with a TB3 port. However, the limitations of TB3 and the device need to be kept in mind to avoid unrealistic expectations of parity with desktop-gaming setups. It will get you close enough, though — for a price. 

Today's integrated graphics, beginning with the Intel HD 620, are enough for most people who don't play triple-A games, game competitively, edit video in 4k, nor work with CAD. The Intel Iris Pro series of GPUs are a significant improvement on the HD 620 and will boost the speed of professional tasks, but will still provide only a lackluster gaming experience for demanding titles.

For those who want to play the latest games, 3D model, or edit 4k video, a dedicated GPU will significantly increase performance. Don't just buy anything with the highest numbers, however. First, make sure that you aren't overpaying for a card beyond your needs: a GTX 1060 is perfect for 1080p gaming at high details. Second, check that the laptop you are looking at can actually run its CPU and GPU at the speeds advertised. Lastly, whether you want a laptop with G-sync, Optimus, or a MUX will depend on your needs regarding battery life, response time, and color accuracy.

External GPUs are a great way to boost the performance of your laptop with integrated graphics, but they are of questionable use (and value) for laptops with a GTX 1060 or better. If money is not an issue and you will only be needing that graphical power in one location, then augmenting an ultrabook with an eGPU might be the most convenient solution.

Helpfull articles:

• Benchmark list of computer games running on laptop graphic cards - What GPU do I need to play my favorite PC game?
• Benchmark list of all mobile and desktop graphic cards - Compare desktop and mobile video cards according to their performance.