The calculation power gap between 2304 and 2048

The double precision values of 480 and 580 are basically the same, which are 1 / 16 of the single precision, about 360gflops. In fact, these two cards are not easy to use. The double precision of 280x is three times that of this, about 1000gflops

Differences:

1, different performance

2048 stream processor and 2304 stream processor, the performance gap caused by the pure stream processor gap, so 2048sp is generally regarded as the high frequency version 570

2. The speed of updating is different.

one is that AMD's architecture and salary can not give full play to the advantages of multi stream processors (580 stream processors far exceed 1070, but the performance is backward). 2048sp, on the other hand, has some updates

extended data:

580 graphics card technology improvement:

semiconctor chip manufacturing technology improvement, through the upgrading of the proction process of core components and the optimization of semiconctor characteristics. Under the premise of still based on TSMC 40nm process, gf110 can run at higher frequency, and can give way to 512 CUDA cores and 16 polymorph engine chips to meet the needs of mass proction

through these two improvement measures, the per watt performance of gf110 is about 20% higher than that of gf100. As a flagship proct, it took half a year to upgrade. I'm afraid that some of the so-called "new architectures" in the past did not have such an increase