AMD Is Now More Competitive in High-End Gaming -- But Nvidia Still Has Strengths

AMD's new RX 6000 GPUs deliver major performance and power efficiency gains relative to its RX 5000 GPUs. But Nvidia's ray-tracing and DLSS ecosystems still look like important competitive advantages.
Author:
Publish date:

AMD’s  (AMD) - Get Report new high-end gaming GPUs look pretty competitive with Nvidia’s  (NVDA) - Get Report latest offerings -- in certain types of gaming situations, that is.

On Wednesday afternoon, following many weeks of teasers, product leaks and rumors, AMD unveiled the first products based on its gaming-optimized RDNA 2 architecture: The $579 Radeon RX 6800, the $649 Radeon RX 6800 XT and the $999 Radeon RX 6900 XT. The GPUs had been given the codename Big Navi.

As expected, all of the products are desktop GPUs. The RX 6800 and 6800 XT are due to arrive on Nov. 18, and the RX 6900 XT is expected on Dec. 8.

For comparison, the first three gaming GPUs based on Nvidia’s recently-launched Ampere architecture -- the GeForce RTX 3070, 3080 and 3090 -- have list prices, respectively, of $499, $699 and $1,499. Reviews of the Ampere GPUs have consistently shown large performance gains relative to comparable products based on Nvidia’s older Turing architecture.

Likewise, AMD is promising big performance gains for its RDNA 2 desktop GPUs relative to the RDNA-architecture Navi desktop GPUs it launched last year -- at times more than twice the performance when gaming at a 4K resolution. Along with architectural improvements for GPU compute units (CUs), the gains are made possible in part by faster clock speeds and significant increases in the number of CUs.

AMD's RX 6800 XT GPU pitted against Nvidia's RTX 3080. Source: AMD.

AMD's RX 6800 XT GPU pitted against Nvidia's RTX 3080. Source: AMD.

Memory advances also play a role. All three of AMD’s first RDNA 2 GPUs will come with 16GB of GDDR6 graphics memory -- twice as much as what AMD's most powerful RDNA GPUs feature -- with the memory clocked at somewhat higher speeds. In addition, there’s now a 128MB high-speed cache (AMD refers to it as Infinity Cache) placed on each chip.

The GPUs also support a feature called Smart Access Memory that provides a slight performance boost for systems that also contain one of AMD’s recently-announced Ryzen 5000 series desktop CPUs. There’s also a 1-click overclocking option (known as Rage Mode) that moderately boosts performance while dialing up a card’s power consumption, and as expected, AMD has joined Nvidia in supporting hardware-accelerated ray-tracing, which can enable photorealistic visuals for game scenes.

At the same time, AMD promises that its RDNA 2 GPUs will deliver a 50%-plus improvement in power efficiency, with the 6900 XT delivering a 65% improvement. As a result, though each of the GPUs packs a massive 26.8 billion transistors -- 160% more than AMD’s most powerful RDNA GPU, the RX 5700 XT -- their graphics card TDPs only range from 250 to 300 watts.

The power efficiency of AMD's RDNA 2 GPUs. Source: AMD.

The power efficiency of AMD's RDNA 2 GPUs. Source: AMD.

For comparison, the much smaller 5700 XT has a TDP of 225 watts, and Nvidia’s RTX 3070, 3080 and 3090 have TDPs, respectively, of 220, 320 and 350 watts.

During its launch event, AMD showed benchmark results that had the $649 RX 6800 XT holding its own against Nvidia’s $699 RTX 3080 when playing popular games at 4K, and slightly outperforming it when Rage Mode and Smart Access Memory were enabled. It also showed the $999 RX 6900 XT trading blows with the $1,499 RTX 3090 at 4K with Rage Mode and Smart Access Memory enabled.

AMD's RX 6900 XT GPU compared with Nvidia's RTX 3090. Source: AMD.

AMD's RX 6900 XT GPU compared with Nvidia's RTX 3090. Source: AMD.

However, AMD didn’t show any benchmarks for how the GPUs perform relative to Nvidia’s offerings when ray-tracing -- a very computationally-demanding technology that Nvidia dedicates a lot of GPU circuitry to accelerating -- is involved. Admittedly, this might have something to do with the fact that AMD’s ray-tracing game ecosystem is still a work in progress.

AMD disclosed during its event that several recently-launched and upcoming titles, including Godfall, Far Cry 6, Dirt 5 and The Riftbreaker, will support ray-tracing on its RDNA 2 GPUs. But while this should change in time, since both companies are relying on Microsoft’s  (MSFT) - Get Report DirectX Raytracing API, it looks like AMD still has a ways to go before its ray-tracing ecosystem matches that of Nvidia’s RTX-series Turing and Ampere GPUs, which now support ray-tracing for dozens of titles.

Also: Though AMD says it's testing a similar solution, the company hasn’t yet rolled out anything comparable to Nvidia’s Deep Learning Super Sampling (DLSS), which (with the help of dedicated processing cores) can significantly improve game frame rates by using deep learning algorithms to help render scenes. Notably, the increased performance headroom provided by DLSS, which is also now supported by dozens of titles, can make it easier for a GPU to handle ray-traced scenes.

This is a very good time to be selling gaming hardware. And given the performance and power efficiency gains AMD’s RDNA 2 GPUs deliver relative to RDNA GPUs, as well as the shortages that currently exist for RTX 3080 and 3090 graphics cards, it’s not hard to see AMD’s latest offerings selling well.

But at the same time, Nvidia’s ray-tracing and DLSS ecosystems, as well as the performance of its Ampere GPUs when these features are enabled, remain major competitive strengths as Nvidia tries to maintain its current dominant position in the high-end gaming GPU market.

AMD and Nvidia are holdings in Jim Cramer's Action Alerts PLUS member club. Want to be alerted before Jim Cramer buys or sells the stocks? Learn more now.