Modern NVidia processors allow you to create high-performance graphics cards. Such devices are capable of solving tasks of increased complexity and have a high speed of operation. Within the framework of this material, the current model range of semiconductor chips, which underlies the accelerators of the 20XX line, will be considered.
Modern NVidia GPUs are manufactured using 12nm technology. This improves their energy efficiency and reduces electricity consumption. At the moment, chips have been developed only for accelerators of the middle and premium classes. The most budget microprocessors for low-performance discrete graphics cards have not yet been presented.
At the same time, mid-level adapters are based on TU-106-400 and TU-106-300 chips, which are respectively used in the RTX 2070 and RTX 2060 accelerator models. In turn, for high-performancegraphics accelerators also provide two models of microprocessors. One of them is TU-104. This chip is used only in RTX 2080 series adapters. But the more productive TU-102 microprocessor is used in two models of graphics accelerators: RTX 2080Ti and Titan RTX. For the foreseeable future, this four-chip lineup will be supplemented by lower-end solutions.
Chip for mid-range graphics accelerators
TU-106-300 is NVidia's lowest performing processor. The GeForce RTX 2060 is based on it. The recommended cost of such accelerators is $350. This chip includes 1920 stream processors and 30 ray tracing units. The number of TMUs and ROPs is 120 and 48, respectively. The clock frequency of this chip can vary from 1365 to 1680 MHz. All this provides the computing performance of this device at the level of 6.5 TFLOPS.
The TU-106-400 chip boasts a slightly higher level of performance. It is used in the RTX 2070 series of graphics cards. Such devices are priced at $500 or more. This microprocessor includes 2304 stream processors and already 36 ray tracing hardware modules. In this case, there are 144 TMUs, and 64 ROPs. The clock frequency range of the device is limited to 1410 MHz from the bottom and 1620 MHz from the top. Its declared performance is 7.5 TFLOPS.
Microprocessor for high performance graphics cards
BThe RTX 2080 graphics accelerator is based on the TU-104 chip. The estimated cost of such devices is $ 700. In this case, the number of CUDA cores has been increased to 2,944. In turn, in this case, there are already 46 tracing modules. The number of TMU units has been increased to 184, and ROP is still the same 64 pieces. This chip model operates at clock frequencies from 1515 to 710 MHz. At the same time, its operating speed is 10.1 TFLOPS.
The processor for NVidia RTX 2080 Ti and Titan RTX video cards is the same. Its marking is TU-102. In the maximum configuration, it includes 4,608 CUDA modules, 72 ray tracing elements, 272 TMU elements, and 88 ROPs. In the younger model of the RTX 2080 Ti graphics adapter, some of these elements are simply disabled, and in hardware. Recommended clock rates are 1350-1545 MHz. At the same time, the speed of such devices is 13.5 TFLOPS. The price for such accelerators starts from $1,000.
As noted earlier, modern NVidia processors for mid-range and premium graphics accelerators provide a high level of performance and are capable of solving any tasks. They implement ray tracing technology at the hardware level. In some cases, it should further increase the performance of such adapters, but so far it has not yet become widespread. According to experts and specialists, the future belongs to it, and it will gradually become more widespread.
Previously considered graphicdevices of this series were introduced quite recently. Therefore, this line of video cards will be supplemented with new devices and will remain relevant over the next year. And after this time interval, such accelerators will cope with the most resource-intensive tasks for another 2-3 years without any problems.
So it's a worthy investment that will increase the speed of any PC. But still, it is most reasonable to use such accelerators only in combination with the most productive computing processors, which include the latest i5 and i7 modifications from Intel or Ryzen 5 and Ryzen 7 from AMD. In all other cases, the hardware resources of the graphics accelerator will not be fully loaded.