NVIDIA Desktop GPUs NVIDIA offers top-of-the-line desktop graphics cards in the GeForce RTX series. In addition to powerful hardware, NVIDIA also offers first-class software support. Regularly updated graphics drivers ensure optimal performance at all times. The modern GeForce RTX series also features artificial intelligence computing cores that can increase performance by tens of percent in supported games or applications or allow you to perform local inference on your own hardware. Floating point Nowadays, we measure the raw performance of graphics cards using the FLOPS unit. It indicates how many operations per second a graphics card can perform, and allows us to compare them. However, these values can vary depending on the bit size of the floating point (or FP). The larger the bit size, the more accurate the calculations, but also the fewer there are. Since gaming computers need precise values to run games smoothly, they mainly use 32-bit FP (or FP32). On the other hand, Deep Learning does not necessarily need such precise calculations, which is why AI graphics cards use FP8. This drastically reduces the accuracy, but on the other hand, increases the number of calculations. Graphics memory VRAM, nowadays primarily of the GDDR type, is synchronous memory, similar to standard RAM. However, in the case of graphics memory, it is typically memory chips with faster throughput and multiple data transfer rates. The result is much faster buffering of data that the graphics card or coprocessor calculates and passes to the processor. CUDA technology Professional application users can use CUDA graphics stream processors thanks to the CUDA architecture. This allows the raw performance of the graphics card to be used for specific calculations, which can speed up work many times over using a classic processor, which is significantly limited by the lower number of cores. Active The active type of cooling is widespread mainly due to the reliable method of reducing temperatures, it consists of air flow cooling, most often created by one or more fans and its goal is to help create a negative pressure inside the case, which significantly improves the airflow that removes heat from the graphics card out of the case. We solve the active cooling of graphics cards into axial, common for traditional gaming graphics cards, and radial, which uses longitudinal air flow, this type of cooling is used, for example, by TURBO versions of graphics cards. PCI Express PCI Express is an interface that usually takes the form of an expansion slot to ensure modularity of the entire system, whether it is GPUs, network cards, controllers, M.2 drives or other expansion cards. It is true that the newer the generation and wider the interface, the higher the performance and throughput. Most modern graphics cards use 16 lines to connect to the processor. Currently, the most current generation is PCI Express generation 6.0 with a speed of 7.5 GB/s per line.
