Discover How DirectStorage Enhances Performance on AMD CPUs and NVIDIA GPUs

SSDs have been around for quite some time and although they significantly reduce loading times compared to HDDs, there is a limit. Most games and applications load almost at the same time, regardless of whether we have a SATA 3 SSD or an M.2 PCIe 5.0 SSD. The full speed of the SSD isn’t utilized unless we employ technologies like DirectStorage. Now, DirectStorage has been tested alongside AMD CPUs and NVIDIA GPUs to see if it improves performance and reduces loading times.

When the PS5 and Xbox Series X/S consoles were launched in 2020, they were the first to include an SSD, finally enjoying near-instantaneous loading times that we had on PCs for quite some time. While many games have near-instantaneous loading times, others still require a considerable wait on both PCs and consoles. SSDs are limited by how they communicate with the rest of a computer’s components, meaning that even if we buy a faster disk, we may not notice any difference.

DirectStorage performance on AMD CPUs and NVIDIA GPUs has been compared

Microsoft DirectStorage was created to avoid this issue, as it allows for the utilization of a faster SSD to reduce loading times and increase performance. It requires a relatively modern PC, with an NVIDIA RTX 20 GPU or an AMD RX 6000 or later. Another test of DirectStorage has now been carried out, comparing the performance difference when activated on AMD CPUs and NVIDIA GPUs. Starting with the first scene without DirectStorage, we have a bandwidth of 0.787 GB/s, a loading time of 108.38 ms, and a 7% performance loss due to asset streaming. The same scene with an AMD Ryzen 7700X CPU using DirectStorage yields 4.6 GB/s, 34.99 ms, and 5% loss.

In the second scene, without DirectStorage, the results are 2.23 GB/s, 1237.8 ms, and a 6% performance loss. With the CPU and DirectStorage, the results improve to 3.82 GB/s, 832.51 ms, and 8% loss. The third scene, without DirectStorage, shows 2.33 GB/s, 1063.7 ms, and 9% loss, while using the CPU achieves 3.5 GB/s, 799.56 ms, and 8% loss, respectively. Finally, the fourth scene without DirectStorage provides 0.445 GB/s, 383.2 ms, and 2% loss, compared to 4.06 GB/s, 78.02 ms, and 5% loss with the CPU.

The Ryzen 7700X improves loading times by 38%, while the RTX 4090 by 77%

As we can see, using the AMD CPU with DirectStorage provides a substantial bandwidth improvement, reducing loading times by an average of 38% across the four scenes. However, the CPU still doesn’t match the performance of a GPU like the RTX 4090 used in the tests. To avoid overwhelming with numbers, let’s just look at the third test with the RTX 4090: it achieved 29.48 GB/s bandwidth, a loading time of 163.1 ms, and a 25% performance loss due to asset streaming.

On average, the GPU managed to reduce loading times by 77% and significantly increased bandwidth. Everything appears to be better using the GPU until we encounter the performance impact graph. Here, the use of the GPU results in an average additional 10% performance loss. In cases like the second and third scenes, performance loss is nearly triple that of the CPU.

In summary, DirectStorage improves performance on both AMD CPUs and NVIDIA GPUs, with the latter offering faster loading times but at the cost of additional performance loss.