Nvidia Uses Different Processing Cores With Different Applications In Its Graphics Processors. The Tensor Core Is One Of These Small Parts That Can Make It Easier For Graphics Cards To Perform Very Heavy And Complex Processing Activities. 

These cores can also help the graphics card to increase the clarity of details and the resolution of game images. Continuing this article from Hardware City, we will introduce and review the use of tensor cores in graphics processors.

Microprocessors can perform mathematical operations such as addition and multiplication on numbers, just in terms of structure and form. Tensor is a mathematical concept that describes the relationship between other mathematical concepts connected.

These numbers are generally shown as an array of numbers; the dimensions of the display can be imagined in shapes similar to the image below. The tensor kernel is briefly used for matrix calculations.

What is Tensor Kernel?

In other words, the term tensor is related to calculations performed in several dimensions, which we call matrix calculations. When you multiply one matrix by another matrix, some addition, subtraction, and multiplication must be done.

For such a story, a large amount of calculation must be done, which a simple processor cannot do, and such processing is considered heavy for a simple processor. The tensor core is one of the most important initiatives of Nvidia company in modern graphics, which can perform heavy matrix processing.

Applications of tensor kernels

The tensor core is widely used in physics, engineering, and mathematics. One of its critical applications is that it can solve complex problems of electromagnetism, astronomy, and fluid mechanics, and it can help a vital technology called Ray Tracing in some parts of gaming.

One of the popular gaming technologies is called Ray Tracing, which is the same as following shadows and lights.

 

Tensor cores can also help to increase the frame rate after enabling the ray tracing feature in the game, which is one of its main features.

Calculations related to deep learning or machine learning are cases where a lot of data or analyses must be processed and extracted. We need new graphics cards to have multiple Tensor cores, although we shouldn’t underestimate Cuda. But now, the tensor is more critical.

Tensor cores alone are a revolutionary architecture in the GPU industry, which were first combined with the NVIDIA Volta microarchitecture; However, their injection in other architectures is also unimpeded.

Tensor Cores are specifically designed for deep learning, and their inference analysis is increased up to 6 times. Also, to advance in the above processes, it has witnessed a 12-fold performance speed, and their processing speed in deep learning and data analysis has finally and at the same time progressed up to 3 times compared to the previous architecture.

Tensor cores are designed for highly specialized applications such as content production, military activities, medicine, aerospace, discoveries, and hundreds of other similar applications. Each Tensor core is capable of processing 64 floating point operations per cycle. The high speed of Tensor cores is not only much higher than the Pascal architecture, but servers that use these graphics accelerators are also up to 47 times faster than similar models that use CPUs for this purpose.

Tensor cores in graphics cards

The presence of tensor cores in graphics cards is not necessary to perform the everyday tasks of these cards, such as graphic rendering and video encoding and decoding. Suppose we want to state the purpose of using tensor cores and DLSS technology in Nvidia’s new graphics cards. In that case, we must say that Nvidia renders a frame at a low resolution when the graphics card is released, and when the rendering is finished, the image resolution increases. Its dimensions reach the dimensions of the monitor. For example, you can render a frame in 1080p resolution using new graphics cards and then resize it to 1440p.

 

When the resolution of the game is high, the textures look much better, and the details of the game are much sharper. Unfortunately, processing all the large pixels requires extensive processing operations.

Summary and answers to frequently asked questions

The new technologies used in graphics cards are not only for gaming but also for neural networks and modeling. In general, Tensor Cores have incredibly high performance, which is improved compared to the previous generation, and its use is summarized in increasing the resolution of images and Ray Tracing.

Does AMD have tensor cores?

Intel and AMD do not have tensor cores in their GPUs. But they have tried to use the same technology in their graphics cards.

What GPU has Tensor cores?

Tensor cores are a proprietary technology of Nvidia, and naturally, the graphics processors of this company will use Tensor cores.

Can tensors be used on GPU?

Yes. Graphics processors or GPUs manufactured by Nvidia are equipped with tensor cores.

What is the difference between CUDA and Tensor?

Coda kernels are responsible for image processing, and tensor kernels are used to perform mathematical calculations.