GeForce GTX 570

When designing the GTX 570, NVIDIA designers followed the same path, reducing the number of stream processors and the bus width. The new product has 15 multiprocessors (accordingly, 480 CUDA cores remain) and 5 memory controllers (320-bit bus width, 40 ROP blocks).

Please note that the GF110 used on the GTX 570 has slight architectural differences from the GF100. The modernization affected texture units and rasterization engines, which received an improved block for rejecting surfaces invisible in the frame (Z-cull). When testing the GTX 580, I did not notice that these improvements led to any increase in performance, but even a “hidden” advantage of 1-2% can be useful in the fight against previous generation video cards and new products from AMD.

To summarize, the new video card can be described as simply as possible. The GeForce GTX 570 is a kind of hybrid of the GTX 480 and GTX 470. It is distinguished from the older model by a smaller bus width and the number of ROP units, and from the younger model by a larger number of stream processors.

As a result of simplifications, the number of rasterization units in the GeForce GTX 570 became 40 versus 48 in the GeForce GTX 480/580, the bus width decreased from 384 bits to 320 bits, and the memory size also became smaller: 1280 MB for the GTX 570 versus 1536 MB for the GeForce GTX 480 /580. Theoretically, such differences between the GeForce GTX 570 and the GTX 480 should have a negative impact on the speed of operation in high screens with anti-aliasing; on the other hand, the clock frequencies of the GTX 570 are slightly higher than those of the GTX 480, which, with the same number of TMU units, allows the GTX 570 to be faster perform texture sampling, and the architectural improvements of the GF110 will be very useful in some cases.

Characteristics of NVIDIA GeForce GTX 570

Name GeForce GTX 570 Core GF110 Process technology (µm) 0.04 Transistors (millions) ? Core frequency 732 Memory operating frequency (DDR) 1900 (3800) Bus and memory type 320-bit GDDR5 Bandwidth (Gb/s) 162.4 Unified shader units 480 Unified shader unit frequency 1464 TMU on conveyor 64 ROP 40 Shader Model 5.0 Fill Rate (Mtex/s) 49400 DirectX 11.0 Memory Capacity 1280 Interface PCI-E 2.0

 The NVIDIA GeForce GTX 570 accelerator inherited its best features from its older brother, the GTX 580, such as GPU GF110, relatively cool temperament and quiet cooling system. In our opinion, NVIDIA engineers have chosen the technical characteristics of GeForce GTX 570 quite accurately, due to which the new product in most cases shows better performance results than GeForce GTX 480, which, together with other bonuses, allows us to confidently recommend the new product as a worthy alternative to GeForce GTX 480.