AMD is Self-Destructing: Financials, Reorgs, and Public Relations

AMD is indisputably the most threatened silicon manufacturer in the gaming space. It is also critical to the stability of competitive advancement. The company exhibited a year-over-year revenue decline of 22% in 4Q14, with its PC segment – not differentiated between CPUs and GPUs – falling an additional 15% over its previous quarter. The PC segment lost approximately $56mm in 4Q14, following a $17mm loss in 3Q14. More recently, JPR showed a 26% quarterly decline in GPU sales for AMD, a bigger hit than the usual mid-year 6.86% decline. A lack of specificity between GPUs, CPUs, and APUs ensures limited visibility to the company's anchors, making it difficult to determine which segments are the worst-performing.

In spite of this – or perhaps spurred-on by it – AMD has rebranded itself several times over, promising “transformation” of the company and defining a set of “five pillars.” That promised transformation came in the form of layoffs and the departure of newly-hired executives who were billed to enact game-changing improvements for the company. Regarding AMD's five pillars, as defined by now-defunct CEO Rory Read, none included a direct mention of “gaming” at the top level; instead, the five pillars consisted of ultra-low power devices, embedded devices, semi-custom solutions, professional graphics, and dense servers. None of these explicitly and directly pertain to PC gaming, though semi-custom solutions can be deployed in what ultimately become gaming devices (see: consoles).

At the time of announcing its pillars (2H14), the company indicated that it was on-target to generate nearly 50% of its total revenue from the five high growth markets it had defined. This, backed by AMD's definition of “transformation” – which shifts the company away from conventional CPU engineering and into growth markets – seems to highlight a diminishing focus on gaming hardware.

AMD suffers a competitive disadvantage on all fronts, expounded by inconsistent driver support, facility-limited fabrication abilities, uncompetitive performance-per-watt in discrete markets, questionable software and API investments, and innumerable other deficiencies. The company is split thin between significantly larger vendors, and the strain bleeds red through its engineering departments.

On top of all this, a series of relations pitfalls and an exclusion of review media ensure limited visibility to a company desperate for positive attention. AMD's actions are those of a panicked, cornered animal, striking out at all in range without regard for grander strategy.

Perspective – How Big is AMD?

Intel is a large company. With record earnings per share set in 3Q14, Intel holds near-complete dominance – 98.3% of all server processor sales (Mercury Research) or, at worst, 92% (Gartner, 2013) – in enterprise market segments; Intel is also the single biggest CPU manufacturer on the planet, owning and operating its own fab plants. The company's server division accounts for approximately 75% of its total revenue, and it isn't relinquishing much control in that market other than to recent advances made by ARM.

Intel is also the only gaming-dwelling CPU manufacturer to own fabrication facilities, each of which can cost upwards of $5 billion. We can’t directly compare the company against AMD as their investments differ in many aspects, but let’s set the stage for relative size of each of the big players.

Depending on which metric is used to define size, ARM appears impressively resource-efficient given the above table. Intel is a steady giant. NVidia has steadily increased its R&D investment, surpassing AMD in recent quarters.

AMD, meanwhile, has axed employee count and R&D spending year-over-year since 2011. AMD spent an approximate $1.7B on R&D in 2009, plummeting to $1.4B in 2010. The company's R&D budget grew in 2011, hovering around $1.45B, before falling to approximately $1.2B in 2013 (Morningstar). In recent years, the company has only struggled further with R&D in the face of a 22% YOY decline in 4Q14 (mentioned above).

AMD CEO Lisa Su, who recently replaced Rory Read, carefully stated that AMD would “increase [its] R&D investments in enterprise, embedded, and semi-custom” (4Q14 Earnings Call). With seemingly endless hits to gross R&D spending, that money has to come from somewhere – it's being shifted from elsewhere in the budget, and AMD doesn't cater to many markets outside of “enterprise, embedded, and semi-custom,” especially if we're looking at their five pillars.

AMD's first quarter of 2015 didn't look much better. CPU and GPU revenue (which are merged in the charts) plummeted 38% YOY, with emergent markets falling an additional 7% (including embedded, semi-custom in consoles, and enterprise).

Cornered

Long-standing rival Intel has held onto the x86 and desktop markets with relatively unwavering resolve, aside from a brief, shining period during the Athlon 64 era. Intel no longer considers AMD competition, concerning itself instead with rising giant ARM and its own struggle to grasp at footholds in emergent and mobile markets. Intel’s CPUs dominate between 92% and 98.3% of the server market and 82.7% of the desktop market, leaving the remainder to what's left of IBM (enterprise), rising ARM clusters (enterprise), and AMD (client & enterprise). Intel is estimated to hold approximately 80% of the notebook CPU market, further highlighting its industry-wide dominance. Looking strictly at gaming saturation, Steam's user-fueled hardware survey reveals a 74.4% / 25.5% CPU marketshare split, favoring Intel.

Unfortunately for AMD, Intel is not its only concern.

GPU rivals nVidia (51.8% total gaming marketshare) and Intel (19.2% total gaming marketshare) have both begun chipping away at AMD's thin 28.6% total gaming marketshare. Accounting strictly for add-in board sales and ignoring Intel’s IGP and AMD's APUs, the numbers paint a bloodbath: nVidia holds approximately 75.98% of the discrete GPU market, leaving the remaining scraps (24.02%) to its struggling competitor. This picture has only grown more gruesome with time, evidenced by record-setting strides made by nVidia's 15% GeForce revenue hike in 2014. NVidia's 75.98% add-in board marketshare hasn't always been this high, either; the company grew to 65% AIB marketshare around this time last year, gaining an additional ~11% of total marketshare in the time since (JPR).

AMD has erupted in the news cycle lately, with an endless stream of refreshes and a new GPU architecture.

Fiji Wasn't Enough

Despite a long-past history of fierce competition between AMD and Intel, and then-independent ATi and nVidia, AMD has continually fallen behind in all departments. The company is now worth less than one-third the $5.4B it spent to acquire ATi in 2006.

Fiji promised the world; the Fury X promised to exceed the same-priced GTX 980 Ti from competitor nVidia, but failed to do so in nearly every regard. And it's not for lack of trying, either. Fiji is an impressive architecture. The shader array is massive, the COMPUTE output is excessively high, the TDP gap has shrunken considerably with the addition of a liquid cooler that mitigates power leakage, and HBM is the way of the future. AMD continued ATi's heritage of introducing new memory technologies first. Unfortunately, none of these items were enough to achieve parity with nVidia, and the limited availability of the Fury X only further damages its reach.

Fury X also saw limited distribution to press, something that became more of an issue with the Nano. We received two Fury X samples on loan from a reader, as AIBs were too low on quantity to distribute.

This rolled into Fury (non-X) and, more recently, the R9 Nano. The Nano is something of an enigma; it's $650 – the same as the Fury X and GTX 980 Ti – but is priced strictly on the basis of its small form factor. The Nano is less powerful than the Fury X, less powerful than the GTX 980 Ti, and is priced the same as both because it's small. That's a niche market to begin with, especially given SFF case adoption of PCI-e riser cards for better size compatibility. What AMD needed was a device to go blow-to-blow with nVidia's high-end, but it has instead created another production-limited, high-cost GPU with a questionably small target market.

AMD has further restricted its reach by limiting press samples to almost none – Tech Report, Hard OCP, and TechPowerUp have all reported a lack of sampling. AIBs have been unwilling to supply the R9 Nano for review, eliminating the route most reviewers take for limited sample cards. It is our understanding that AIBs are prohibited by AMD from distributing review cards of some models to reviewers. This reeks of low faith in product and of “taking it personally” when it comes to critical reviews, spelling a misguided attempt at “damage control” from historically fair press. It is this fingers-in-ears approach to third-party analysis that ultimately kills a brand.

Then again, the company's developer relations strategy isn't much more friendly; AMD – particular its VP of Alliances Roy Taylor – has lashed-out in response to GameWorks' use in The Witcher 3 and PhysX in Project Cars. The company boldly told Ars Technica that GameWorks “completely sabotaged AMD's performance,” saying that “it's wrecked our performance, almost as if it [were] put in to achieve that goal.” At the time of this statement, it had been approximately six months since the last stable (non-beta) AMD driver update, which was Catalyst Omega. Catalyst Omega promised extremely frequent updates when we were on-call with AMD's Robert Hallock in December, though those plans never came to fruition.

Also around the time of AMD's insinuation of foul play, AMD GFX users found that lowering the tessellation count of HairWorks drastically improved AMD performance – something AMD could have issued from the get-go, though didn't provide the solution until after user discovery. AMD later updated its drivers and saw improved Witcher 3 performance.

A New Restructure

AMD seems to be in constant internal flux. During the process of finalizing this article, AMD announced that it would be forming a new “Radeon Technologies Group” separate from its previously-joint CPU & APU group. Raja Koduri has been named Chief Architect of the Radeon Technologies Group, and is “responsible for overseeing all aspects of graphics technologies used in AMD's APU, discrete GPU, semi-custom, and GPU computer products.”

This decision to separate groups which were once intentionally merged and obfuscated is a curious one. AMD is in trouble right now, and the impressive Fiji arch didn't perform well enough in the real world to improve its market grasp. We're curious to see whether AMD can recover its position competently enough to stabilize, though it'd be wise to remain alert for potential external acquisitions or splits within the company.

A Wake-Up Call

The new Radeon Technologies Group doesn't cry “recovery,” but we can hope. AMD and ATi have a history of introducing new technologies to market and advancing the industry through healthy, fierce competition. There's still some good under all the refreshes, rebadges, and odd PR. Fiji is good architecture. HBM is where everyone's going. AMD has some genuinely clever marketing joining its CPU, APU, and GPU technologies, migrating various items between the three; dynamic voltage and improved power management got moved from APUs to Fiji, for instance, and that's been a huge factor in minimizing the card-to-card TDP gap to something near parity. Going to liquid for the Fury X dropped TDP significantly, as did moving to HBM. AMD clearly can take criticism -- heat, power, and noise -- and turn it into an impressive collection of technology (despite sub-par performance, we still have to admire the Fury X for its improvements). The company consistently falls short of greatness, though, and its outward stance of hype and aggression doesn't help.

AMD's relations strategies are misguided and toxic, especially with AMD's Roy Taylor publicly insinuating that some publications are "unfair." On the game maker side, cries of foul play about the year's biggest RPG – targeted at nVidia or not – only aim to drag-down the reputation of its developers, and that's a losing strategy in a market that's about to dive deep into engine-level hardware optimization.

AMD is panicking. The company is bleeding money out of nearly every one of its market segments, and it's struggling to refresh existing brands fast enough to remain relevant. It is important for AMD to stay in the game in some form, acquired or not, to keep its competitors in check. Unfortunately, AMD has recently demonstrated nothing but a lack of both awareness and organization. It is as if all the individual components operate with zero communication; there's all the parts of the body, but no central nervous system.

- Steve “Lelldorianx” Burke.