The GPU Wars Are Heating Up

AMD Seeks To Challenge NVIDIA — How Does Its New RDNA Architecture Stack Up?

After AMD released it’s long-awaited Navi GPU, NVIDIA leaked that it will be releasing the “Super” version of its existing RTX lineup sooner rather than later.

The Super announcement was PR at its finest. The world’s preeminent GPU maker wasted no time attempting to overshadow AMD’s rival chip, but the question remains: should NVIDIA be worried about AMD’s Turing competitor?

AMD CEO Lisa Su claims the RDNA architecture is a completely new design built from the ground-up and on par with the way its CPU designers built its Zen chip. The 7nm Navi 10 GPU is the first RDNA-based graphics chip, and the full core – used by the RX 5700 XT – comes with 40 compute units, 2,560 cores across two shader engines, 64 render output units, and 160 texture units. It’s 251 mm² and has a full 10.3bn transistors inside it. Navi has a GDDR6 memory controller, though is compatible with HBM if needed, and has PCIe 4.0 support too. The PCIe 4.0 support is especially important because the faster transfer rates between storage and the CPU/GPU allows more data to be processed faster.

NVIDIA told journalists at E3 that it still holds the title of “most power efficient GPU on the planet,” citing the Turing architecture’s economical performance per watt advantage even on the less power-efficient node. That’s a claim that will undoubtedly be challenged by AMD, which cites the move to 7nm as a major reason why the 5700s should, theoretically, use less power to get better performance than their NVIDIA rivals.

Turing is a 12nm part and we’re quite proud of that. Turing remains the most power efficient GPU on the planet.

— Jeff Fisher, SVP of GeForce

One thing that stands out about the Navi release is that AMD doesn’t always claim to beat Intel or NVIDIA when the comparison is core to core or compute unit to compute unit. What AMD does emphasize, however, is that Navi can give users a lot more raw power at a much lower price.

As far as performance goes, AMD is claiming a 25% increase in instructions per clock for the RDNA architecture versus its GCN architecture (RDNA contains elements of GCN, but is noteworthy for its performance improvements). It’s also promising a 50% overall performance gain when utilizing the same power and the same general configuration too.

Ultimately independent tests are needed to compare AMD’s apples to NVIDIA’s apples, but the AMD’s increasing relevance in the GPU world has clearly ruffled NVIDIA’s feathers. Whether or not Red Team can cut into Green Team’s market share is still yet to be seen, but the innovation in the GPU realm will benefit consumers and enterprises alike, especially with technologies like PlasmaENGINE finding revolutionary new uses for the powerful processing units.

Write a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.