If you’re buying a graphics card, you may be wondering what it stands for in gpu. Nvidia’s model numbers are important because they help you determine the performance and features of each card.
For example, the GTX 1050 Ti is a high-end GPU with 20% better performance than the non-ti model.
Model Numbers
Model numbers are often cited as one of the major culprits of poor performance on modern systems. Thankfully, Nvidia and AMD have both made efforts to provide their customers with options in the form of a wide variety of products ranging from entry-level enthusiast-oriented GPUs to top-of-the-line mainstream models.
The best way to start your research is with a simple Google search for the various gpu architectures in question, followed by a quick look at your favorite graphics card retailer. The resulting list should be an excellent starting point to help you make the right decision for your next upgrade.
As far as gpu features go, the most obvious is the number of cores. However, a more important factor is the actual clock speed that each individual core reaches at any given time. This is referred to as the GPU clock speed or simply the GPU frequency and can be compared to your CPU’s clock rate (which may be in the tens of gigahertz) or the memory bandwidth that each core can access.
RTX
Nvidia’s latest Turing graphics cards, the GeForce RTX 2080 and 2080 Ti, set a new standard for gaming. They offer a 30 to 40 percent performance boost over the previous generation. They also feature ray-tracing technology, which can help developers create realistic effects such as hair, fur and physics.
Despite all of this, it’s important to remember that not all games will benefit from ray-tracing. It’s a very powerful tool, but it can only improve the look of games that are specifically designed to use it.
Even then, a game’s engine is a big factor in how much ray-tracing it can use. For example, Shadow of the Tomb Raider – one of the first games that supports RTX – only uses ray-tracing for a small fraction of its scenes. This means that, to make a real difference, a game needs to be built with ray-tracing in mind from the very beginning.
That said, a RTX card might be able to render ray-traced objects faster than a previous generation card by using more cores or other technologies, and that could mean a significant boost in speed. However, it’s important to note that ray-tracing will never replace raster-based rendering in any game, nor is it the most efficient way to render a frame, as all those photons bouncing around will saturate your GPU pretty quickly.
Turing
Turing is Nvidia’s next-generation graphics processor architecture that was recently unveiled at SIGGRAPH and Gamescom. It’s a significant step forward from the company’s current Pascal-based GeForce GTX 1000 series, offering improved performance and more powerful GPU units.
At its Gamescom event, NVIDIA unveiled the GeForce RTX 2080 and 2080 Ti, which feature Turing-powered GPUs that are up to six times faster than Pascal cards. This is thanks to the use of new ray-tracing technology that helps your GPU calculate game lighting much more quickly. The company also claims that the RTX architecture is “a new computing model” that will help deliver more realistic visuals in games.
NVIDIA’s GeForce RTX 2080 and RTX 2080 Ti were designed with real-time ray-tracing and deep learning features in mind, which are crucial for delivering more realistic graphics in PC games. While these features aren’t a requirement for any game, they can drastically improve the quality of your gameplay.
In addition to its RTX cores, Turing also supports more CUDA units than any other GPU architecture on the market. NVIDIA says that the extra Cuda units can allow your card to process integers and real numbers simultaneously, which is a big jump.
Tensor Cores
Tensor Cores are specialized hardware units added to recent NVIDIA GPUs to speed up matrix multiplication-related tasks like convolutions and densely connected layers in neural networks. However, because they are specialized hardware and require a specific programming model, Tensor Cores cannot be straightforwardly applied to other applications outside machine learning.
Nvidia CEO Jensen Huang said at the Gamescom keynote that the next-gen GeForce RTX GPUs will feature a new AI engine based on its Tensor Cores. Known as Deep Learning Super Sampling (DLSS), this technology is expected to greatly boost game speeds by up to two times with uncompromised image quality.
DLSS will be able to “denoise” the grainy artifacts caused by real-time ray tracing’s light casting, which can make the visual experience look less real. It will also be able to help reduce anti-aliasing (AA), which usually consumes a lot of GPU cycles and memory resources.
The company plans to deploy DLSS in a number of its GeForce RTX GPUs, but it’s not clear which ones will receive it first. As the DLSS feature is not built-in to any of the RTX graphics cards yet, developers will need to code it into their titles.
In addition to ray tracing, the GPUs will also support other types of AI, including machine learning. These include artificial intelligence frameworks such as Caffe2, MXNet, CNTK, and TensorFlow. These frameworks deliver faster training times and higher multi-node performance, enabling developers to build faster and more accurate models with ease.