Although there’s no shortage of AMD talk here on Tom’s Hardware, we’d like to take this moment, on the dawn of secondnd-gen Ryzen’s release, to discuss how the red team might just be at a major crossroad. Changes in market strategies among its two major competitors, Intel and Nvidia, have left AMD an opportunity to capitalize on the consumer computing market.
Between this year and last, some significant changes have occurred in the industry that are relevant to AMD’s current situation. Obviously, there was the Meltdown and Spectre debacle, which undoubtedly hit Intel hardest. AMD lost the head of its Radeon Technologies Group, Raja Koduri, to Intel, and Intel subsequently announced its plans to re-enter the high-end discrete GPU market. And Nvidia became the second of two fabless semiconductor companies to break the top 10 list in chip sales.
There are also some broader shifts that have occurred in the industry. Among the three competitors, AMD is the only one that hasn’t seen a significant change in its target market. Consumer desktop and semi-custom projects (i.e., consoles) continue to account for the majority of AMD’s revenue. On the processor front, the company spent the past few years architecting its return to competitiveness in x86. That focus paid off only recently, albeit in a more spectacular fashion than many were expecting. On the graphics front, however, AMD hasn’t been able to replicate that success.
Intel and Nvidia, on the other hand, have both diversified and branched into more lucrative computing markets. Intel is executing on a strategy that it announced in 2016 to shift its focus from PCs to IoT and data-centric markets such as cloud computing and FPGAs. Nvidia has shown a proclivity for datacenters as well, but it’s also investing heavily in developing autonomous vehicle systems and AI. For both companies, their revenue from other sources is close to matching their consumer PC sales.
For years now, AMD has tried to follow in its competition’s footsteps and claim a piece of the datacenter market. Last year’s Vega graphics architecture launched as a compute product before the gaming version of it hit retail, and AMD is marketing Epyc and Vega Instinct as competition for Nvidia’s dominance in the AI market. AMD has made some inroads, and a breakthrough in the datacenter market is what the company ultimately needs, but we think AMD is currently better poised to capitalize on some of the slack its competition has left in the consumer computing market.
Despite Koduri and other recent losses, AMD seems to be currently riding a wave of momentum that is fueled by its Ryzen processors. The second generation of Ryzen is now being launched, and our review of the Ryzen 7 2700X, as well as rumors of an impending response from Intel, show that updated Ryzen is even more competitive than its predecessor. Intel missed the mark with its 2017 desktop consumer processor launches by fragmenting its product lineup. It rushed new six-core processors out to replace its four-core parts, introduced the new and confusing X299 high-end desktop (HEDT) platform whose capability depends entirely on the processor pairing, and it launched the new Coffee Lake mainstream desktop platform, which now seems poised to span three very different chipsets. AMD’s Ryzen and Threadripper offerings, on the other hand, are streamlined and work on platforms that behave as platforms should: providing (more or less) the same capabilities to every processor that they’re compatible with.
AMD’s newfound success in consumer processors and Intel’s intent to re-enter the discrete GPU sector put the two companies in a strangely reversed situation, with each playing catch up on the other’s strength. And, if rumors are to believed, Intel is going to be trying to catch AMD in all graphics markets, including gaming--not just in the GPU compute market where AMD is significantly weaker. This makes sense, because Intel will need a market to offload the development costs of its first discrete GPUs. The consumer market is less sensitive to power and performance, and consequently will be more tolerant to the inevitable teething problems Intel’s early GPUs will face.
The success of Ryzen makes us think that AMD is much closer to catching Intel in processors than Intel is at catching AMD in graphics. However, Intel has, as always, resources and process technology on its side. When it finally arrives in (rumored) 2020, Intel’s discrete GPU would be the only one to potentially benefit from in-house fabrication. Nvidia and AMD are tightly partnered with TSMC and GlobalFoundries, respectively, but both are still fabless companies.
The wild card that can crash both parties is, of course, Nvidia. It poses a threat to Intel’s datacenter presence and to AMD’s already inferior position in the discrete GPU market. For the former, it comes down to shifting market demands. Deep learning and AI are increasingly the target of datacenters, and massively parallel processors like GPUs are simply superior to CPUs in these workloads. The effect is compounded by the fact that Nvidia is heavily invested in designing even more targeted hardware for these workloads. For the latter, it’s a matter of pure technical superiority. AMD’s GPU lineup has been recycling old designs and introducing new, noncompetitive ones on a recurring basis. The most successful design AMD has had since 2014 is Polaris, which is heading into its second refresh cycle this year. The last GPU from the company that beat its competitor at launch was Hawaii from 2013.
In order to make a serious attempt at claiming the dominant position in the consumer market, AMD desperately needs to strengthen its position in graphics. Recent reports show that AMD’s GPU market share might be improving, but we know that the cryptocurrency craze is at least partially responsible, because the company hasn’t done anything in terms of product releases that could justify the change. Meanwhile, Nvidia is just waiting for the right moment to release new GPUs that would put AMD even further behind. Working in AMD’s favor is the consumer goodwill that it can capitalize on, especially among enthusiasts. Neither Intel nor Nvidia will ever admit that they’re skirting evil-monopoly status, but the former’s rushed response to Ryzen and the latter’s GPP program don’t help their cases.
AMD’s first steps to improving its competitiveness in graphics might have started with the hiring of Koduri’s replacements. The latest rumors suggest that AMD is ready to bring a Ryzen-level revival to its graphics cards. Unfortunately, as far as we know, it won’t be happening this year.