Almost seven years ago today, AMD unveiled its “small die strategy.” Following its failure with the Radeon HD 2900 XT, AMD decided to try something new in an attempt to catch up to and eventually surpass NVIDIA. Instead of following NVIDIA into ever-increasing die sizes and all the associated hazards, the company would put more effort into developing smaller GPUs for the more lucrative sub-$300 market.
Meanwhile, AMD would use multi-GPU technology (known as CrossFire) to deliver even better performance at a total cost that is comparable to NVIDIA’s flagship cards in the high-end markets.
Although AMD’s early efforts were fruitless in terms of dethroning NVIDIA’s dominance, products like the Radeon HD 4870 and Radeon HD 5870 were huge spoilers, providing a significant portion of NVIDIA’s flagship performance with GPUs that were smaller, manufactured for less money, and used less power.
While AMD’s GPU designs ostensibly stopped being informed by the compact die concept earlier this decade, in practice, it has been doing so for quite some time. Despite being over 100mm2 smaller than NVIDIA’s flagship GK110, AMD’s Hawaii dies (the company’s largest as of 2013) measured in at 438mm2.
To bring things up to date, this month represents a momentous occasion for AMD with the release of their new flagship GPU, Fiji, and the flagship video card based on it, the Radeon R9 Fury X. Fiji is not just another high-end GPU introduction for AMD (their third on the 28nm process), but rather a watershed moment in the company’s history.
In addition to being a performance play, Fiji also features cutting-edge memory and power optimization technologies. Simply put, it may be the final 28nm GPU, but it is certainly not the least significant.
Despite the fact that AMD’s new Fiji graphics processing unit (GPU) is anything but small (the company has pushed its design to the reticle limit), I bring up the small die strategy to show how the GPU industry has evolved over the past seven years and how AMD has had to adapt.
NVIDIA has kept pushing huge dies since 2008, but they’ve gotten smarter about it, releasing more energy-efficient GPUs that have made it difficult for a nimble AMD to beat NVIDIA’s prices.
However, as rendering techniques grow less and less AFR-friendly, the viability of dual GPU cards, the foundation of CrossFire and SLI, has declined. Finally, from a business perspective, AMD’s GPU and APU sales are currently anticipated to be lower than NVIDIA’s GPU sales, a situation that hasn’t occurred in over a decade.
It’s not that I have anything against AMD as a company, by the way. Is nothing if not the underdog that keeps surprising us with what they can accomplish with less, but this history is crucial to understanding where AMD has come from and why Fiji is, in many ways, a revolutionary GPU for the firm.
There is no coming back from this; AMD is taking aim at NVIDIA’s flagship with the largest, most powerful gaming GPU they can design, thereby killing the small die concept.
What are you hoping to accomplish? To challenge NVIDIA’s dominance in the performance stakes, which the latter has had for too long, and to provide an alternative flagship card that can compete with NVIDIA’s best.
AMD must overcome a number of obstacles before it can reach its goal. Between GM204 and GM200, AMD has its work cut out for it in competing with NVIDIA’s Maxwell 2 GPUs, which are excellently executed, very efficient, and incredibly powerful. With Fiji and the R9 Fury X, AMD is tackling head-on performance, power consumption, and form factors.
Simultaneously yet, there has never been a more even playing field. TSMC’s 28nm process is currently in its fourth year and has a significant amount of time left. Because both AMD and NVIDIA have had so much time to fine-tune their products for what is now a very mature process, any advantages for being a first-mover or an aggressive player have evaporated.
Now that the 28nm process has reached its limit, NVIDIA and AMD must rely on their engineers and their architectures to determine who can create the best GPU within the constraints of the technology.
Since GPU production technology has stalled at the 28nm node, it’s difficult to discuss the GPU issue without also discussing the manufacturing scenario. However, the current scenario on the manufacturing process side has had an astonishing, unparalleled effect on the technological and architectural growth of discrete GPUs, which has compelled AMD to adapt its business operations to meet market demands.
Consequently, Fiji is not simply a step toward huge GPUs that can compete with NVIDIA’s best, but also a representation of the considerable efforts AMD has made to continue boosting performance despite manufacturing limits.
Now let’s get started with our examination of the Radeon R9 Fury X. This month sees the release of AMD’s new flagship card, which is powered by the company’s latest and greatest graphics processing unit (GPU), Fiji.
The R9 Fury X is competitive because it features 4096 SPs and the first implementation of High Bandwidth Memory. In the following pages, we’ll go more into the card’s architecture and other characteristics, but for now, just know that it has a tonne of shaders, an even greater amount of memory bandwidth, and is designed to provide AMD’s finest performance to date.
Although AMD’s other three Fiji-based parts—including the R9 Fury X—will be released in the following months, this month is all about the company’s flagship graphics card.
At its initial price of $649, the R9 Fury X is priced similarly to its main rival, the GeForce GTX 980 Ti. NVIDIA’s GTX 980 Ti, which was released at the end of May, is essentially a preemptive strike against AMD’s R9 Fury X, as it provides a performance so close to that of NVIDIA’s top GTX Titan X that the distinction is debatable.
The Fury X is the closest AMD has come to an NVIDIA flagship card in quite some time, and while a victory over the GTX Titan X would be wonderful, a victory over the GTX 980 Ti is all that’s needed.
And last, from a business point of view, AMD will target multiple markets with the R9 Fury X. The combination of decreasing prices for 4K monitors, the availability of 4K Freesync monitors, and AMD’s superior performance at 4K resolution makes it a priority for the company to compete with NVIDIA’s GTX 980 Ti in this regard.
Though the flagship VR headset, the Oculus Rift, won’t arrive until Q1 of 2016, we expect AMD to make a big deal out of the R9 Fury X’s virtual reality capabilities. And now, after more than three years since the first release of the Radeon HD 7970, the Fury X is poised to bring the kind of generational performance enhancements that normally justify an upgrade to the market for AMD’s first 28nm card.