HP TECH TAKES /...
Exploring today's technology for tomorrow's possibilities
Subscribe
Which is the Best NVIDIA Graphics Card for Gaming?
Jolene Dobbin
|
May 9, 2019
When coming across component page after component page on the internet, it’s easy to get burned out trying to decide which is the best graphics card to buy.
In this article, we’ll try and ease some of those decision-making pain points by helping you understand some of the features in NVIDIA graphics cards. Then you can make a more informed choice when you’re shopping for the best NVIDIA graphics cards for laptops and desktops to meet your PC gaming needs.
What is a GPU?
The graphics processing unit (GPU), which is also called the graphics card or video card, is a very important part of a gaming system. And for some graphics-intensive applications, it may be more important than the central processing unit (CPU), which is often referred to as the brain (or heart) of a computer.
The GPU performs rapid math calculations while freeing the CPU to perform other tasks. It speeds the process of creating and rendering images, animations, and video.
Two types of GPUs exist, integrated (or embedded) GPUs that reside on and share memory directly with the CPU, and discrete GPUs that have their own memory on their own card.
Types of GPU cores
Cores convey sets of instructions that enable software programs to perform specific functions and process data.
Processors like CPUs have both single and multiple cores, which are used for sequential processing. A GPU, on the other hand, was built for multi-tasking. It has hundreds or thousands of smaller cores that handle thousands of threads (also known as instructions) simultaneously.
1. CUDA cores
A leader in graphics card design, NVIDIA created CUDA (Compute Unified Device Architecture), which is a specialized programming language that uses the GPU to perform tasks faster and more efficiently. CUDA cores are the GPU’s equivalent of CPU cores. They’re parallel processors, with several thousand cores, that perform numerical calculations in fractions of seconds.
CUDA cores are responsible for handling all of the data that flows through the GPU, bringing it power and speed. Graphics settings that are affected the most by CUDA core counts are those that require the most from the GPU including lighting, shadows, special effects, complex simulations, virtual environments, and more.
2. Tensor cores
Built on NVIDIA’s Turing architecture (see below), tensor cores are designed for performing complex calculations, especially those within artificial intelligence (AI) and deep learning applications, in a single operation - enabling massive increases in efficiency and throughput.
They’re used specifically for deep learning super-sampling (DLSS), NVIDIA’s proprietary technology that provides enhanced images without having to tap into a lot of the GPU’s key processing capabilities. DLSS forces games to render at lower resolutions while using AI to make it appear as if the images were rendered at higher resolutions.
It does this by using some anti-aliasing techniques. Anti-aliasing is the technology that removes edges or “jaggies” (what’s often referred to as the “staircase effect”) from video games. making them appear smoother.
3. Ray-tracing (RT) cores
Ray-tracing (RT) cores are designed to carry out ray-tracing techniques. Ray tracing is a rendering capability that realistically simulates a scene’s lighting and objects by rendering physically accurate and lifelike shadows, reflections, and more.
It forms an image by tracing the path of simulated light, creating levels of realism that are much better than those created using traditional rendering techniques. Because ray tracing is very computationally intensive, RT cores are designed to accelerate some of the most time-consuming components of ray-tracing calculations.
What about the architectures?
NVIDIA graphics cards, such as the GTX series and RTX series, are built on the Pascal and Turing architectures, respectively.
1. NVIDIA Pascal architecture
The NVIDIA Pascal architecture is one of the more powerful computer architectures ever built inside a GPU. It offers substantial improvements over older architectures in performance, power consumption, and heat generation.
The Pascal architecture delivers extraordinary performance of more than 5 teraflops for high-performance computing (HPC) workloads and a 7x increase in deep learning capabilities compared to previous-generation GPU architectures. [1]
2. NVIDIA Turing architecture
The NVIDIA Turning architecture is currently the world’s most advanced GPU architecture and the first GPU that’s capable of real-time ray tracing. It improves performance significantly more than previous generations (such as Pascal) with enhanced graphics capabilities and new programmable shading technologies. This enables fluid interactivity with large models and scenes, improved virtual reality (VR) experiences, more realistic details, and smoother rendering.
A multi-purpose GPU architecture, Turing combines multiple processor cores, including both tensor cores (for AI acceleration) and RT cores (for real-time ray tracing) performing real-time ray tracing, AI, and pixel processing simultaneously.
GeForce GTX 1060, 1070, and 1080 series graphics cards
The NVIDIA GeForce GTX 10 Series graphics cards are some of the best NVIDIA GPUs, powered by the Pascal architecture to deliver up to 3 times the performance of previous-generation graphics cards. In addition, they offer innovative gaming technologies and enhanced VR experiences.
1. GeForce GTX 1060
The GeForce GTX 1060 takes advantage of the Pascal architecture to enhance speed and power efficiency. It has a base speed of 1,506 MHz and boosted speeds of 1,708 MHz, a memory bandwidth of 192 GB/s, and 1,280 CUDA cores.
The GTX 1060 will play most games at Full High-Definition (FHD) resolution (or 1920 x 1080 pixels, which is referred to as 1080p gaming) resolutions, as well as some in the Quad HD (QHD) resolution range, which is 2560 x 1440 pixels (also called 1440p). The GTX 1060 is a solid mid- to budget-range card, going for around $200 to $300.
2. GeForce GTX 1070
The GeForce GTX 1070 offers GDDR5 video memory of 8GB, a clock speed of 8 Gbps, and a 256 GB/s memory bandwidth. Like the GTX 1060, it has a base speed of 1,506 MHz, and boosted speeds of 1,683 MHz, but with more CUDA cores (1,920 in fact).
The GTX 1070 falls somewhere between high- and mid-end graphics card, retailing in the $400 price range, and is ideal for gaming at FHD resolutions. It also works well with games playing in QHD.
3. GeForce GTX 1080
An advanced gaming GPU, the NVIDIA GeForce GTX 1080 is one of the most powerful in the GTX 10 series. It features 2,560 NVIDIA CUDA cores, GDDR5X high-bandwidth video memory of 8GB and a memory speed of 10 Gbps, with a base speed of 1,607 MHz and boosted speeds of 1,733 MHz.
This card runs approximately 30% faster than the GTX 1070 and 1060 cards delivering breakthrough power, efficiency, and performance. The GTX 1080 is considered a high-end card running in the $500-plus price range and is perfect for gaming at QHD resolutions.
GeForce RTX 2080 Ti, 2080, and 2070
The GeForce RTX series graphics cards are the first gaming GPUs based on the NVIDIA Turing architecture. The highly advanced Turning-based RTX series graphic cards run powerful AI algorithms in real time to create crisp images and lifelike shadows, refractions, reflections, and other advanced and special lighting effects that were not possible on previous architectures.
This GPU architecture is equipped with all the cores - CUDA, tensor, and RT - and it’s the first to support real-time ray tracing (RT). The real-time RT is carried out by the RT cores, and its performance is expressed in a new metric called Giga Rays per second (or GR/s).
Armed with GDDR6 memory (currently the world’s fastest), RTX series graphic cards let you play games on maximum settings with high frame rates which are usually those higher than 60 frames per second (fps).
The Turing architecture performs 6 times faster than Pascal for games using ray-traced graphics. NVIDIA claims that on existing games Turing will deliver 2 times the performance of Pascal, and 4K gaming at 60 fps, on even the most demanding triple-AAA (AAA) titles. [2]
Resolutions for 4K are Ultra High Definition (UHD) or 3840 x 2160 pixels - more than 8 million pixels. And 4K means that it offers 4 times the resolution of standard HD.
4. GeForce RTX 2080 Ti
Arguably the best premium GPU in the market today, the GeForce RTX 2080 Ti is truly revolutionary in terms of performance, speed, and realism. It has incredibly fast 14 Gbps memory speed and memory bandwidth of 616 GB/s as well as a huge amount of 11GB of GDDR6 memory.
While its base speed of 1,350 MHz and boost speed of 1,545 MHz are not all that high - and lower in fact than the GTX 1080 - it does have a whopping 4,352 CUDA cores, 544 tensor cores, and 68 RT cores speed along at 10 GR/s.
With the additional memory and CUDA cores, the RTX 2080 Ti is especially suited for high-performing, GPU-intensive 4K gaming. The RTX 2080 Ti can handle 10 GR/s and 78 trillion RTX-OPS (what NVIDIA refers to as the performance available when rendering next-generation games with hybrid graphics).
All of these features enable both ray tracing and DLSS in games that support it. And that’s the caveat: there are only a handful of games using ray tracing currently available.
The RTX 2080 Ti is definitely not cheap, though, as it runs well over $1,000 at the moment. But as a high-end card that’s packed with screamingly fast performance features and very cool cutting-edge graphics capabilities, it may just be worth your investment for soon-to-be-released and future gaming titles for years to come.
5. GeForce RTX 2080
The GeForce RTX 2080 is only a slight step down from the RTX 2080 Ti. While the RTX 2080 Ti performs approximately 20% faster at 4K than the RTX 2080, it still packs a punch with the same memory speed of 14 Gbps, and with faster base clock and boost speeds of, respectively, 1,515 MHz and 1,710 MHz.
It features 8GB of GDDR6 memory and a high memory bandwidth of 448 GB/s. It has 2,944 CUDA cores, 368 tensor cores, and 46 RT cores, an average of approximately 30% less than the RTX 2080 Ti.
Like the RTX 2080 Ti, the RTX 2080 will be able to handle most 4K games at 60 fps. According to NVIDIA, it’s capable of 60 trillion RTX-OPS and 8 GR/s. Considered a high- to high-mid-range card, the RTX 2080 retails in the $700 to $800 price range.
6. GeForce RTX 2070
The GeForce RTX 2070 is similar to the RTX 2080 Ti and RTX 2080 in terms of power, efficiency, and speed. Like its siblings, it has a 14 Gbps memory speed, along with 8GB of GDDR6 memory like the RTX 2080. Its base clock and boosted speeds of 1,410 MHz and 1,620 MHz aren’t that far removed from those of the RTX 2080 either.
It isn’t quite as powerful as its brethren, however, as it only offers 2304 CUDA cores, 288 tensor cores, and 36 RT cores - almost half of that of the RTX 2080 Ti. So, unlike the RTX 2080 Ti and 2080, it won’t be able to push the same boundaries of playability at many games’ highest settings.
But it will enable a decent 60 fps and, probably more importantly, the remarkable ray tracing graphics capabilities as the others in the RTX series.
The GeForce RTX 2070 is a powerful yet cost-effective alternative to both the RTX 2080 Ti and RTX 2080 in that at approximately $500, it only costs about half as much as the RTX 2080 Ti.
Understand your gaming requirements
Choosing the right graphics card for your gaming needs is highly subjective. Are you upgrading your gaming system or just parts of your system? Are you buying an entirely new gaming desktop or gaming monitor? Are you building a computer gaming system from scratch? Are you looking to play one specific game? Or do you want to be able to play a range of different types of games? Is your budget very restrictive, or is the sky the limit?
These are all questions to consider as you go down the GPU path. No matter your choice, NVIDIA most likely has a card that will be just right for you.
[1] NVIDIA; Pascal Architecture
About the Author: Jolene Dobbin is a contributing writer for HP Tech Takes. Jolene is an East Coast-based writer with experience creating strategic messaging, marketing, and sales content for companies in the high-tech industry.
