In November last year, Microsoft officially launched
DirectStorage 1.1, which is available to any developer who wants to use the technology in their games. This API allows the graphics card to read data directly from the NVMe SSD, bypassing the CPU and memory, which can greatly reduce the loading time of the game. In the version 1.1 update, the GPU decompression function was added, which can use the powerful parallel execution capabilities of the GPU to free the CPU from such repetitive and complicated tasks. With high-bandwidth NVMe devices, the data loading time of the game can be greatly reduced.
PC Games Hardware tested the three graphics cards of AMD, Intel, and NVIDIA, and compared their DirectStorage decompression performance, which one is better? The test uses Core i9-12900K as the platform and reference object respectively tests PCI-E 4.0 and PCI-E 3.0 NVMe SSDs, and SATA 6Gbps SSDs, and the graphics cards used in the tests are AMD Radeon RX 7900XT, Intel Arc A770 and NVIDIA GeForce RTX 4080.
Interestingly, the performance of Intel’s graphics cards in terms of DirectStorage decompression is significantly better than AMD’s and NVIDIA’s. When Arc A770 is paired with PCI-E 4.0 SSD, the transfer rate reaches 16.8GB/s, followed by NVIDIA’s RTX 4080, which also has a bandwidth of 15.3GB/s, and AMD’s RX 7900XT also has a bandwidth of 14.6GB/s, both of which are much higher than those used 5.2GB/s when the Core i9-12900K performs this task.
In fact, no matter which graphics card, the actual loading time will be greatly reduced, from 5 seconds to 0.5 seconds. It can be seen that the efficiency of using GPU to perform the decompression task is very high, and it does not matter what GPU you use, they are all capable of this task, but it should be noted that this API is completely invalid for SATA SSD.
