Losing its final major workstation OEM, AMD's CPU business essentially withdrawn from the workstation market ... and the company doesn't seem to care
GPUs but not CPUs. The strategy might have me scratching my head, but when it comes to marketing its wares to the workstation market, that's precisely where AMD's head is at. Yes, AMD is still working a presence as a supplier to the workstation industry, plying its professional-brand FirePro GPUs, but as far as we can tell, the company has thrown in the towel when it comes to selling its CPUs into the same space.
Working hard to increase its foothold in a market dominated by rival Nvidia, AMD is not giving up on the battle for professional graphics hardware dollars. Riding the momentum of its Evergreen generation of GPUs, AMD's workstation-focused FirePro GPUs have made some gains (albeit modest ones) in recent quarters. The company knows it's got a competitive line and is looking to capitalize — if not overthrow Nvidia, at least hang onto the 13% it has managed to hold onto, and maybe gain a few points (and some healthy margins) here and there.
But the opposite appears to be the case in CPUs, where the company has now completely relinquished its hard-won market share back to Intel. Forget fighting the good fight; the company appears completely apathetic about the fortunes or value of entrenching Opteron (or Phenom, or whatever brand CPU) in this vibrant, performance-sensitive corner of the PC market. Truth be told, AMD hasn't really been engaged in the workstation market for some time, sitting back to watch Opteron's penetration in workstations drop from a peak of 3.6% of the worldwide market in Q2 2006 (and a more impressive 9.9% of dual-socket workstations) down to 0.1% in Q3 2010.
Several years ago, the company could revel in the fact that it had gone from zero major OEMs supporting AMD64 in workstations, to counting all but Dell as Opteron workstation vendors. But today, it's essentially back to zero, having lost IBM early on, then taking a big hit when its most enthusiastic supporter, Sun, withdrew from workstations altogether, and most recently when the only remaining serious supplier, HP, discontinued its two Opteron models. So while its FirePro line continues to duke it out with Nvidia in workstation graphics, AMD's role as a supplier of CPUs to the workstation market is effectively over and done. (Fujitsu still sells one outdated model, but literally in the hundreds per year, hardly worth including.)
With Intel back on solid competitive ground with respect to all facets of the workstation business, AMD would face an uphill battle to regain share as a provider of workstation-class CPUs, even if the latter was both engaged in the market and committed to the challenge. But AMD appears to be seeing workstations as anything but a key battleground. And surely, without a fierce determination to win back business, it's not going to.
The Big Question: Why?
Why have AMD's CPUs disappeared from workstations? We see two reasons. One, the company needs to pick its battles carefully, and — right or wrong — it doesn't see workstations as a priority. And two, today's business unit structure at AMD doesn't appear to lend itself to serving the workstation market effectively. Especially compared to Intel, AMD lacks the wherewithal to compete full bore in every segment; so it needs to choose its battles judiciously. Clearly, it needs to protect and regain some lost ground in servers and consumer segments as job one; both represent bigger pieces of the pie than do workstations.
Or perhaps AMD senses the workstation business is a tougher nut to crack for its CPUs than its GPUs, since it's now so far behind Intel (approximately 99.9% as compared to nearly 0.1%) as to make it a lost cause. But that's a tough argument to swallow, given that its positions for CPUs and GPUs aren't all that different, with both respect to technology and penetration in the market.
Yes, it's a long way behind Intel when it comes to CPUs, but it's not like the situation is that different in professional-brand GPUs. Where ATI once held a substantial share (near 30% back in mid-2005), these days Nvidia is nearly as dominant in workstation GPUs as Intel is in CPUs, taking less than 87% (Q2' 2010) compared to AMD's 12.3%. And to flip the tougher-nut-to-crack argument around, there's a valid argument to be made that it would be easier for Opteron or Phenom — or down the line, a Fusion brand — to re-gain workstation market share (now coming from a re-starting point of zero) than push FirePro's penetration past its current position.
But even if AMD wanted to crack the workstation nut, it no longer appears organizationally equipped to do so. Fitting (and looking lost) under servers, workstations look to be the ugly stepchild, with servers grabbing all the attention. The two platforms share a similar architecture and system components in dual-socket (2S) configurations, but where 2S platforms represent the bulk of server volume, they conversely represent a relatively small share (around 20%) of workstations.
So for server management, workstations might appear small potatoes. And building and marketing a single-socket mobile or entry desktop — where the majority of the volume is — is more a prospect for the client-side arm of the business to consider. But then that business unit doesn't seem interested in the relatively low volumes those segments present as compared to desktop and notebook PCs.
Now more than other industry observers, we understand the volumes, margins and expenses being incurred in today's workstation business. The volumes, of course, aren't up to the level of mainstream PCs or servers. But at 3.3 million units representing $8.2 billion (system revenue), they're nothing to sneeze at. Furthermore, the margins are far superior to mainstream PCs, and with the commonality that can be leveraged from both of the other two platforms, the NRE is quite modest. For the right companies, it's a very attractive market to play in — if you don't believe it, just ask Nvidia or HP.
Ironically, AMD itself, at some level, sees the value. Otherwise, why would it continue to push FirePro graphics? Predominantly represented by OEMs, the TAM for FirePro is tied almost 1:1 with workstation volume. So if the volume for workstations is too low for AMD CPUs to bother, why is it enough for AMD GPUs to care?
In fact, when it comes to entry barriers and the dollars necessary to ante into the game, it's arguably easier to push CPUs than GPUs. While the silicon at the heart of FirePro products is the same that ships in consumer products, there's significant engineering NRE and man-hours necessary to design the workstation card variant, develop and optimize workstation-focused APIs and middleware, and test and certify the hardware for key professional applications like Maya and Solidworks. On the other hand, the company can sell the same Opteron and Phenom CPU SKUs into the workstation market that it does for servers and high-end consumer/corporate desktops. And on top of all that, think of the clean, one-stop-shop solution the company can pitch: a combination of workstation-caliber CPU and graphics, a proposition that neither of its chief rivals Intel and Nvidia can match.
Marketing and sales shouldn't be a huge burden, either, as the big workstation names to focus on — AMD, Dell, Lenovo, Fujitsu, NEC — are the same names they already lobby for PCs and servers. And given that Intel owns 99+% of the market, OEMs like HP would welcome an enthusiastic re-entry into the market by AMD. After all, no business wants to be beholden to one supplier, even if it's a supplier of essentially infinite volume, like Intel.
Either AMD doesn't see the opportunity due to organizational blinders, or it does, but feels there are more critical battlefronts that need to be manned for now. Maybe it's just an issue of triage, and the company will more visibly assert its CPU wares in workstations once it sees sufficient progress in other areas first. Perhaps it's just waiting for the first desktop Fusion "Llano" to arrive to then return to market take a bite out of the high-volume entry level, rather than (or, maybe, in addition to) pursuing the higher end market with a Bulldozer-based Opteron. Or maybe the company will simply reassess this now completely unaddressed opportunity, which represents about as low-hanging a fruit as we can imagine: reasonable volumes, superior margins, low entry barriers and only one real competitor (albeit a formidable one).
But whatever the reason, at least for now, without its previous footholds and (it appears) any champions in the company to pick up the torch, it's not looking like AMD's CPUs will be making a significant comeback in workstations any time soon.
Alex Herrera is a long-time engineer and consultant in the areas of semiconductors and computer graphics. Among the hats he wears, Alex is a senior analyst with Jon Peddie Research and author of the “JPR Workstation Report” series reference guide for navigating the markets and technologies for today’s workstations and professional graphics solutions. He can be reached at alex