[ad_1]
Starting from the bottom, in terms of gross power, the RTX 2070 is roughly equivalent to the GTX 1080; the RTX 2080 goes with the GTX 1080 Ti; the RTX 2080 Ti is in a league of its own. The 2070 and 2080 have 8 GB of RAM GDDR6; the 2080 Ti has 11 GB. All three are based on the company's new Turing architecture, which means that they have dedicated cores for IA (Tensor) and Ray-tracing (RT).
Expect a fourth card, probably the RTX 2060, to drastically reduce the entry price in the coming months, followed by a series of reduced options for budget-conscious players (Series 10 has grown to 1030 ). There is also room for expansion: the RTX 2080 Ti Founders Edition can handle 14.2 trillion floating point operations per second (TFLOPS), while the Turing TU102 chip is based on push up to 16.3 TFLOPS. This is achieved through a combination of higher clock frequencies and CUDA cores (the 2080 Ti is equipped with 4,352, the fully configured TU102 with 4,608).
RTX is also coming up with a lot of improvements under the hood. There is a faster cache system with a shared memory architecture, a new graphical pipeline, and simultaneous processing of floating and integer computations. If that does not mean anything to you, do not worry too much: the word "soup" does not just mean that the RTX range has more raw power, but uses that power more efficiently.
And that's the key here. Ray-tracing has made headlines, and I'm intrigued to see how the developers are using it, but it's the efficiency that fascinates me about RTX.
The ultimate goal of a gaming system, whether it's a $ 2,000 gaming PC or a $ 300 Nintendo Switch, is to calculate a color value for each pixel on a screen . Even a simplified guide on how a modern graphics pipeline works, it would take the whole length of a novelty, but here is a summary in three sentences: the processors are not designed to render modern graphics. Instead, a processor sends a plan for what it wants to draw to a GPU, which has hundreds or thousands of cores capable of operating independently on small image blocks. The GPU executes on the CPU plane, running shaders – very small programs – to define the color of each pixel.
The challenge for graphic card manufacturers and game developers is therefore the scale. This $ 300 portable-mode switch typically calculates 27 megapixels per second, which may well be with a three-year NVIDIA mobile chip. If you're aiming for 4K at 60FPS (which many players who buy RTX cards want), your system should be close to half a billion pixels per second. This puts tremendous pressure on a system, especially when you consider that your PC does not simply select these colors and simulates a complex 3D environment in real time as part of the calculations.
There are already many techniques to reduce this tension. We render all or part of a scene at a lower resolution and stretch the results. This is very obvious when you play 720p on a 1080p screen, but less when, for example, a cloud of fog is drawn in a quarter of a resolution. And that's all that NVIDIA's optimizations are: reducing quality in places you will not notice.
The new NVIDIA graphics pipeline can use several new shading techniques to reduce risk. In many ways, this relies on less flexible energy saving measures used for VR, such as MRS (multi-resolution shading) and LMS (shading adapted to the current one). goal). In the image above, you see a GPU split a scene into a grid in real time. The unstained squares are very detailed and shaded in a ratio of 1: 1, just like a classic game scene. Colors do not need the same level of attention. Red squares, for example, are only shaded in blocks of 4×4 pixels, while more detailed but non-essential blue squares are shaded in 2×2 blocks. Given the low level of detail of these areas of the image, the change is essentially imperceptible.
Source link
