GPGPU Benchmarks
For GPGPU benchmarks we are utilizing Aida64’s GPGPU benchmarks.
In this first graph, we are looking at the data transfer speed that can be made to the video cards memory. This innately tests PCI-Express bandwidth performance. You can see that with PCI-Express 4.0 the two AMD video cards can read and write much faster to GPU memory, 25GB/s. The two NVIDIA GPUs are around 12GB/s. If an app needs to direct access video card memory it will be able to do so faster with the XFX Radeon RX 5700 XT THICC III Ultra.
This next test shows Single-Precision and Double-Precision FLOPS value in GFLOPs. This test shows that the two AMD video cards are ahead in both single-precision, and even more so in double-precision GFLOPs. The XFX Radeon RX 5700 XT THICC III Ultra is even ahead of the reference AMD Radeon RX 5700 XT. The higher clock speeds provide a higher single and double-precision performance. The double-precision performance is literally doubled that of the GeForce RTX 2070/SUPER FE video cards.
In this test, we are looking at the Single-Precision Julia test and Double-Precision Mandel test. The results here are interesting. In Single-Precision Julia the GeForce RTX 2070 SUPER FE and GeForce RTX 2070 FE are much faster than the two AMD video cards. However, the tables turn when looking at the Double-Precision Mandel test. In Double-Precision the AMD video cards are a lot faster than the GeForce RTX video cards. The XFX Radeon RX 5700 XT THICC III Ultra is faster than the reference AMD Radeon RX 5700 XT, showing the increased clock speeds help here.
Now we are looking at encryption performance using the GPU. Higher is better here. In terms of SHA-1 Hash, the two GeForce RTX video cards wipe the floor in performance compared to AMD. However, the two AMD video cards are faster at AES-256. The XFX Radeon RX 5700 XT THICC III Ultra is faster than the AMD Radeon RX 5700 XT, especially in SHA-1. The improved clock speeds help here also.