AMD has launched into a new era of PC graphics with RDNA 2, which is the technology behind Big Navi, but Nvidia’s Ampere architecture also offers great advances over its previous generation Volta. Driving such powerful features as ray tracing at high frame rates and 4K resolution requires a deep understanding of computer hardware design. When processing such vast amounts of data, memory access bottlenecks have to be considered, heat managed and power draw minimized.
Nvidia raised the bar with its GeForce RTX 30 series, offering up to 8K resolution while still rendering 60 frames per second. Nvidia was the first to offer real-time, hardware-based ray tracing with its previous generation RTX GPUs that were released in 2018, so the leading graphics chip maker has had more time to refine that technology. Previously, ray tracing was a burdensome process that was measured in seconds, minutes, or even hours per frame, rather than frames per second. AMD moved from its GCN architecture to Navi in 2019, but couldn’t match Nvidia’s first Generation RTX GPUs. This set up anticipation for the RDNA 2, which came to be known as Big Navi.
AMD rose to the challenge and Big Navi also boasts real-time ray tracing. Each Compute Unit (CU) has this new ability and AMD’s Radeon RX 6900 leads the line up with 80 CU. Nvidia’s top GPU has 82 ray tracing cores, but performance won’t be equal between AMD’s and Nvidia’s hardware, so the similar number may be meaningless. As the pioneer of hardware-based ray tracing, Nvidia likely has an edge with its second-generation improvements. In fact, AMD shared some results of DirectX performance on the RX 6800 that indicates Nvidia’s RTX 3080 has higher performance ray tracing with fewer cores. However, that isn’t the only important metric in gaming. There are relatively few games that support this advanced rendering. The vast majority of games are made with shaders that simulate reality and, in some cases, create a stylized look that provides an aesthetic that wouldn’t be possible otherwise.
There is more to a great graphics card than ray tracing and both AMD and Nvidia have broken new ground with advanced memory access techniques, shaders, and sampling methods to increase the quality of rendered images at maximum speed. When the GPU can process such vast quantities of data at high speed, supplying that data becomes a challenge. Nvidia uses faster throughput GDDR6X RAM and reduces memory requirements with Deep Learning Super Sampling, which is abbreviated to DLSS. Nvidia’s Ampere design increased to a 384-bit memory bus to allow more data to be transferred at once. AMD has a similar super-resolution technology in process and it will be part of its Direct X support. These sampling techniques create additional pixels based on a lower resolution render reducing the burden on both the processor and memory. AMD claims its Infinity Cache doubles bandwidth of its 256-bit bus. In part, this is due to a 128-megabyte L3 cache and expanded memory access to all CUs. When used with a Zen 3 Ryzen 5000, AMD’s Smart Access Memory allows the CPU to read and write directly to GPU memory. This combined with its overclocking Rage Mode allows it to match Ampere frame rates in many games, according to AMD.
AMD and Nvidia chose related, but different approaches to solve problems, so direct comparisons are difficult. An AMD CU is not the same as an Nvidia CUDA core. Also, the results AMD shared have qualifications and individual results will vary. Until hands-on reviews and third party benchmarks are posted, the theory behind each design and the claims are interesting, but they don’t reveal how AMD’s design will perform in real-world tests. However, Nvidia GPUs still have very restricted inventory and the RTX 3070 just became available to order, so reviews are limited for Nvidia as well. What is clear is that Nvidia will be facing stiff competition from AMD this year and in the future, which will benefit the industry and gamers. As expected, AMD did not directly challenge Nvidia’s best RTX 3090, but made a value proposition, to achieve similar performance in ideal conditions at 4K resolution and doing so at two thirds the price. The race between AMD’s Big Navi and Nvidia’s Ampere is tighter than ever and this exciting rivalry will continue in the future.