If you’re a typical consumer who usually purchases suitable PC products within the $500 or below range, there’s no need for concern. However, if you’re someone who strays above that, you might be in for a rough time.
There have been numerous rumors about AMD’s shifting strategy for its desktop graphics cards under the upcoming RDNA 4 architecture, particularly about AMD potentially scrapping plans for high-end enthusiast-level products in the Radeon RX 8000 series. Now, we finally have some clearer information.
At a press Q&A at IFA 2024, Jack Huynh, AMD’s senior vice president and general manager of Computing and Graphics, addressed the company’s evolving strategy for competing with Nvidia. According to Huynh, AMD’s aim is to increase its market share against Nvidia. However, given the performance and reception of the Radeon RX 6000 and 7000 series, this new approach tells a different story.
AMD plans to forgo releasing a “halo” class flagship product, the kind that typically garners the most amount of attention but is rarely purchased by most PC enthusiasts.
AMD has taken this approach before, intentionally shifting focus away from large-scale, high-end products to concentrate on mid-range offerings that cater to the majority of consumers. This strategy became evident after the release of the Radeon 300 series, which included the cutting-edge “Fury” graphics card featuring HBM (High Bandwidth Memory) for significantly improved memory bandwidth compared to traditional GDDR.
However, these cards underperformed in the market. The R9 Fury X, while more advanced, was overshadowed by NVIDIA’s GeForce GTX 980 Ti, which offered 6GB of GDDR5 memory. In contrast, the Fury X, despite its superior technology, was constrained by its 4GB HBM capacity.
For the following 400 series, AMD chose to exit the high-end market entirely, avoiding the costly production of large-scale dies. Instead, they focused on delivering powerful and affordable mid-range products aimed at the broader consumer base.
This shift led to the creation of the Polaris architecture and the RX 480, which was designed to compete with NVIDIA’s GTX 1060. The RX 480 excelled effortlessly in this space, offering a highly affordable 4GB version for $200 USD and an 8GB model for just $240.
In contrast, NVIDIA’s GTX 1060 6GB was not only weaker but started at around $250, with their “Founder’s Edition” priced at $300. To match Polaris’ aggressive pricing, NVIDIA released the GTX 1060 3GB, which reduced CUDA cores from 1280 to 1152 and came with only 3GB of memory. This version was priced at $200, the same as AMD’s fully capable RX 480 4GB.
After stepping away from the high-end market, AMD returned with the RX VEGA series, though the launch was met with lukewarm reception. These GPUs used HBM2 memory with 8GB capacities. The RX VEGA 56, featuring 56 compute units (CUs) and priced at $400, was positioned as a rival to NVIDIA’s GTX 1070, but despite being slightly faster, the GTX 1070 was $20 cheaper. Meanwhile, the higher-end VEGA 64, with 64 CUs, competed with NVIDIA’s $600 GTX 1080 but was priced $100 lower.
NVIDIA responded by releasing the GTX 1080 Ti, featuring the high-performance GP102 core at $700. The 1080 Ti dominated the market with a significant 25% performance lead over AMD’s VEGA 64, which unsurprisingly struggled to gain traction.
AMD later rebounded with the launch of the Radeon VII, named after its new TSMC 7nm process. Essentially a die-shrink of the VEGA 10 core, it reduced the die size by nearly a third. The Radeon VII boasted 16GB of HBM2 VRAM and doubled the bus width, becoming the first GPU to reach 1TB/s in memory bandwidth.
Priced at $700 to compete directly with NVIDIA’s GTX 1080 Ti, the Radeon VII traded performance blows with NVIDIA but often came up short in head-to-head comparisons.
By the time AMD launched the Radeon VII, NVIDIA was already rolling out its RTX 2000 series, with the RTX 2080 based on the mid-range TU104 core priced at $800, and the 2080 Ti commanding a staggering $1,200. As consumers largely ignored the Radeon VII, AMD once again shifted focus to the mid-range market, having struggled to produce a full-scale lineup for years.
This led to the RX 5700 XT, initially announced at $450 but quickly reduced to $400 after consumer pushback. It became a “cheap and affordable” 251mm² powerhouse that outperformed NVIDIA’s $350 RTX 2060 and even rivaled the $600 RTX 2070.
The success of the RX 5000 series forced NVIDIA to introduce its “SUPER” lineup, releasing the RTX 2060 SUPER for $400 and the RTX 2070 SUPER with more CUDA cores for $500. AMD’s RDNA architecture was a massive success, reshaping the mid-range GPU market.
Then came the RX 6000 series, powered by the RDNA 2 architecture, featuring a complete lineup of competitive GPUs that challenged NVIDIA’s RTX 3000 series across the board.
NVIDIA appeared so concerned that they launched the RTX 3080, featuring the high-end GA102 core, at $700 but with a limited 10GB of GDDR6X memory. AMD’s 6800 series, particularly the 6800 XT, came with 16GB of GDDR6 memory starting at $580, with the 6800 XT priced at $650.
This card frequently traded blows with the RTX 3080 across various titles, with no clear winner emerging. A similar dynamic played out between AMD’s $1,000 RX 6900 XT and NVIDIA’s $1,500 RTX 3090, although the RTX 3090 often edged ahead in performance, especially with its 24GB of VRAM.
Despite AMD’s success in delivering more affordable and competitive products, NVIDIA maintained a stronghold on consumer mindshare. Even with AMD finally offering competent alternatives, many consumers continued to prefer NVIDIA, often ignoring the fact that AMD’s products were superior in performance and specifications, providing enhanced longevity thanks to a greater quantity of VRAM.
The release of NVIDIA’s RTX 4000 series has widely been regarded as one of the worst GPU generations to date. NVIDIA significantly inflated prices while cutting back on core quantity, offering lower-grade dies marketed as higher tier SKUs. This left consumers paying higher prices for subpar hardware.
For example, until now, NVIDIA’s 60-series cards typically used a 106-class core. However, the RTX 4060, priced at $300, was the first to feature a 107-class core, which in previous generations would have been reserved for a lower tier 50-series GPU.
If you wanted a true 106-class product, or an authentic 60-series GPU from the RTX 4000 lineup, you had to spend $400 or $500 on the 4060 Ti, which came with either a disappointing 8GB of memory or a slightly better 16GB option.
When NVIDIA announced the RTX 4080 at a staggering $1,200, it was met with laughter, quickly followed by anger. Adding fuel to the fire, NVIDIA introduced a so-called “RTX 4080 12GB” model for $900, which featured significantly fewer CUDA cores, 7,680 compared to the 4080’s 9,728, a 20% reduction.
It was also limited by a 192-bit memory interface, whereas the true 4080 had a 256-bit interface. In reality, this model was more akin to a 4070 or 4070 Ti class card, but NVIDIA tried to pass it off as an RTX 4080 at an inflated price.
Thankfully, the “RTX 4080 12GB” was pulled before launch, and the card was later rebranded as the 4070 Ti, released with the same specs for $800.
Meanwhile, AMD responded with its RX 7000 series, following NVIDIA’s aggressive pricing strategy. This led to some of the highest GPU prices in history, with fewer cores relative to flagship models than ever before. However, AMD offered compelling alternatives, such as the $500 RX 7800 XT, which outperforms NVIDIA’s RTX 4070 while costing $100 less and offering 4GB more memory.
The RX 7900 XT similarly outperformed the RTX 4070 Ti, while AMD’s flagship RX 7900 XTX managed to surpass NVIDIA’s RTX 4080, initially priced at $1,200, by offering a discount of over $200.
While AMD has provided consumers with somewhat more affordable alternatives to NVIDIA’s offerings, they have not made a strong effort to significantly undercut the competition and capture a larger market share.
Despite RDNA 3 being a better value overall, the situation remains less than ideal, as AMD has largely mirrored NVIDIA’s new pricing strategies rather than significantly challenging them.
Jack Huynh, AMD’s senior vice president and general manager of the Computing and Graphics Business group, outlined the company’s goal of increasing its GPU market share from around 19% to as high as 40% or even 50%. Achieving this would require a strategy focused on aggressively undercutting the competition.
Since enthusiast-class graphics cards represent a smaller segment of the market, gaining such a significant share would depend heavily on selling a larger volume of mainstream gaming GPUs.
Take the RX 6700 XT, for instance. Initially priced at $480, it was competent but overpriced. It wasn’t until the RX 6000 series was succeeded by the current 7000 series that the 6700 XT became a $300 contender while stocks lasted.
The 7700 XT, its successor, aimed to compete with NVIDIA’s 8GB 4060 Ti but was priced with a $50 premium. The original $450 MSRP of the 7700 XT was simply too high for what was ultimately a mid-range product.
AMD’s most apparent strategy, and one it seems to be pursuing, is to focus on the mainstream market by delivering the best possible GPUs at affordable prices for gamers. This approach allows NVIDIA to continue dominating the high-end enthusiast market, which yields substantial profits despite relatively low sales volumes.
While specifics on AMD’s future plans remain under wraps, Huynh made it clear that AMD’s near-term strategy doesn’t involve directly competing with high-end GPUs like the rumored Nvidia RTX 5090. Huynh emphasized the importance of gaming to AMD’s future.
Rather than taking on NVIDIA head-to-head, AMD aims to leverage its strengths in providing superior mid-range options. According to Huynh, there is much to look forward to in AMD’s future graphics card lineup.
Rumors long since suggested that AMD would forgo large-scale dies with the RX 8000 series, potentially missing the opportunity to directly challenge NVIDIA’s high-end products that attract few gamers. Instead, AMD is likely to focus on providing high-performance GPUs tailored to gamers within the $250 to $500 range.
To achieve this, AMD could prioritize producing smaller dies, which are easier to manufacture in larger quantities, cost less, and can be priced more aggressively while still ensuring healthy profit margins. This strategy could be effective, provided AMD avoids directly mimicking NVIDIA’s pricing tactics that create the illusion of value.
Recent GPU rumors indicate that AMD’s upcoming Navi 48 core will feature 64 Compute Units and continue the evolution of RDNA 4 architecture. This new architecture will include Software Prefetch and more flexible scalar loads, making GPUs more CPU-like as they handle a broader range of compute applications. Additionally, RDNA 4 will support AI features such as FP8 and sparsity, and improve cache controls to better align with RDNA’s complex cache hierarchy.
Such a design could allow AMD’s top RDNA 4 graphics card to deliver performance comparable to the Radeon RX 7900 XT. If priced around the $500 range, it would significantly undercut NVIDIA’s offerings and provide exceptional value.
This is especially relevant given rumors that NVIDIA might drop the GB104 core in favor of a smaller, lower-grade GB105 core to create a misleadingly wide performance gap between the RTX 5070 and RTX 5080, which may not offer substantial improvements over their predecessors.
AMD appears less interested in competing with NVIDIA’s high-end models like the RTX 5080 and beyond, particularly as NVIDIA has sharply increased prices and shifted focus to data centers. The RTX 5000 series is expected to see even higher prices, with NVIDIA seemingly intent on maximizing profits with subpar hardware.
As a result, AMD is likely to dominate the $400 and below price bracket, a market segment they have traditionally excelled in. Meanwhile, gamers seeking high-performance options at reasonable prices might find compelling choices with RDNA 4, as AMD aims to offer solid performance without the exorbitant pricing seen in the high-end segment.
