Overclock.net › Member Blogs › An Interpretation Of Amds Long Term Strategy

An Interpretation Of Amds Long Term Strategy

First off, something to satisfy Intel die-hards (also known as stating the very obvious facts that are true as of today): Conroe is one hell of a hammer punch that hit AMD right in the solar plexus. Intel's only significant rival in the consumer CPU marketplace has lost its wind and is gasping because of the power of the blow.

This is undeniable, even after the much-ballyhooed Phenom Quad-Core line stumbled out of the gate. In all fairness, Phenom isn't the Conroe killer that AMD's most ardent fans hoped it would be, but to dismiss it as a complete failure is to simply admit you're drinking Intel's Kool-Aid.

Phenom is bugged with L3 cache errata; it's got problems overclocking stably, and it simply cannot touch C2Q in terms of raw processing power and speed when overclocked. At "equivalent" clock speeds, the difference is much smaller, but the present issues with overclocking mean that, in terms of being attractive to that tiny enthusiast market segment that has arrogant delusions that it dictates the course of the market, it's a pig with makeup on compared to even less than Intel's best.

As ever, it's a race between two, and on this particular track, Intel just got the hole shot and beat AMD to the first corner.

But races are not always won on the first corner. If there's an immutable rule in racing, he who leads at the end of the final lap wins.

AMD has got some catching up to do, but it won't get to the top simply by leading the MHz race.

In a way, it's a repeat of the infamous GHz Wars of several years ago, when Pentium III/4 was the defending champion, and AMD was clearly in second place. Then K8 was born, and after a few growing pains, became a legend amongst PC enthusiasts. Intel's Superman had a taste of AMD Kryptonite, and thinking that the answer was to try to fly faster than a speeding bullet, Pentium 4 ran into the limitations of its design and was consigned to trailing K8 at the end of the last lap. Whether with one core or two, AMD opened up a can of whoop-tushie on Intel.

The giant woke up, and used its vast resources (which has always been its greatest advantage over AMD) to design a much more efficient competitor; it pulled the plug on the Pentium line and introduced the mighty Core design.

AMD, meanwhile, had not exactly stood still. Although Socket AM2 is nothing but a refresh of K8 (albeit an advance over S939), it also made another interesting move which indicates that it isn't content to just compete in the CPU market. AMD acquired ATI, the Canadian GPU manufacturer, and absorbed a significant amount of know-how in the process.

The ATI-AMD merger meant that AMD now was fighting a war on two different fronts, facing two mighty enemies: Intel was the principal opposition in the CPU stakes, and nVidia was the rival that mattered in both the GPU and motherboard chipset stakes.

That's a huge bite, and it remains to be seen if it's more than AMD is able to chew.

In terms of long-term strategy, it's early days yet. And after a little more than a year of sucking both Intel's and nVidia's exhaust fumes because of necessarily conservative product roll-out strategies (because of the absorption of ATI) and clumsy PR moves and product launches, the puzzle pieces are starting to fall into place.

It may admittedly sound absurd to hardened PC enthusiasts, but the signs point to the idea that AMD is no longer primarily interested in ruling just the consumer CPU market. I'm sure that they would like to be on top of the desktop market sector, but if the present Phenom line doesn't have it in it to overtake Intel's best, it's not a complete loss. I think there's a bigger target that AMD is actually aiming at.

I believe AMD's long-term strategy is focused on developing an entirely integrated computing platform. While this may sound counter-intuitive to traditional PC enthusiasts, who are used to the PC's modular design characteristics, in some ways this is a more interesting technical challenge, a more sophisticated approach.

AMD is not interested in owning the bragging rights to the fastest CPU on the market; it wants the bragging rights to having the fastest and most efficient computer. The strategy is to develop a computing platform that is overall the fastest available. If any of the individual components are faster than the competition, that's fine; what's most important, though, is to develop the sum of the parts so that it outperforms anything else out there.

AMD has grown up as a provider of technical solutions, so therefore their goals and aspirations have also matured. It's an ambitious long-term strategy, a huge calculated gamble, and a substantial risk.

Think about things for a moment: Neither of AMD's rivals is capable of producing a complete computing package. Intel has CPUs and motherboard chipsets, but has no real expertise in a discrete graphics processing solution. nVidia might presently have the best GPUs and a supreme ability to design motherboard chipsets, but they don't produce CPUs.

AMD has the complete triumvirate of CPU, motherboard chipset, and graphics processor expertise. If all three subsystems are designed to work completely in concert, performance would theoretically be optimized both in terms of peak output and overall efficiency. Compatibility issues would be eliminated, and there should be platform-specific optimizations inherent in designing an integrated platform. In other words, there would be no compromises insofar as making the best available parts work together.

Look at things this way: Can you run SLI out of the box on an Intel chipset motherboard?

The strategy will hopefully benefit the enthusiast sector, but I foresee AMD's real targets to be the server/enterprise market and the laptop sector. The desktop market segment, of course, is a plum prize as well, but as long as price remains attractive, there will always be people who buy AMDs for their home desktop. This is a salient fact all-too-easily ignored by many enthusiasts who are only interested in the top-of-the-line. Enthusiasts with the financial power and the wherewithal to purchase nothing but the best are, in fact, the tiniest minority.

Phenom is an opening salvo, and though its initial performance has disappointed the enthusiast sector, it is emphatically not a design failure. Some of its features are especially indicative of the path AMD will likely take, specifically the very intriguing Split Power Planes which would allow for more efficient power management for both the CPU cores and the IMCs. The ability to intentionally shut down certain circuits in the cores when not on load (dubbed "CoolCore" by AMD) is also a signature of AMD's intentions to optimize CPU efficiency.

Moreover, a consistent trend of all processor product lines is that as time progresses, so does the performance of a product line. The improvements are a result of incremental refinements to design details, materials utilized, and manufacturing techniques. Processors evolve over time.

AMD's GPUs are admittedly also racing from behind, but it's undeniable that this part of the puzzle is also progressing. The 29xx-series were disappointing in that their performance wasn't within sniffing distance of nVidia's best, but the revamped 38xx-series looks much better. For sure they are far less power-hungry compared to their 29xx-series siblings, but in terms of price for performance, they presently trump nVidia.

More than just the discrete graphics processor, though, AMD now also has a pathway to an integrated graphics solution. Some of the rumored solutions are quite radical, actually: Dubbed "Fusion," AMD's integrated graphics solution would not be a part of the motherboard core logic array, but be a part of the CPU itself.

This is most interesting, not because of how the GPU side of things will benefit; in fact, I think the opposite would be true. Industry experts such as Alex St. John have said the GPUs are far more powerful and are capable of more complex operations than even the best CPUs, so cross-breeding the two types of processors will improve CPU performance significantly. St. John has even said that the day is coming when the CPU will be relegated to attending to the OS, while GPU-type processors will do all the heavy lifting. If St. John's analysis is correct, AMD's strategy suddenly becomes brilliant instead of the failing gamble that it looks like to some today.

The chipset sector is probably the most disappointing piece of the puzzle at the moment; nVidia is the undisputed king on the AMD platform side, while Intel's chipsets are quite competitive with nVidia's. AMD's chipsets aren't the enthusiasts' choice with an Intel processor because most dual-GPU Intel boards (if not all) can run Crossfire. This is where the whole package really needs to catch up, and quickly.

This is not a blind stab in the dark as far as analysis goes; the fact is, AMD has already introduced the first iteration of this optimized platform to the retail channel. The so-called "Spider" platform is the first clear indication of AMD's strategy.

I am not Nostradamus; all I've really done is look at what's happened in the past and tried to guess AMD's long-term strategy. It's a game of chess, really, an endurance race. Every move does count, but some moves count for more than others. To lead the race at the start of it is fine, but what matters most is who leads at the end. AMD's most ardent fans, I'm sure, hope that all the moves, and all the hard miles that are yet to be run, all end up in a victory.

As a fan of computing technology, all I'm sure about at this point is it'll be fun to see if I've read the chess board correctly.

As always, thanks for reading. Comments and discussion are sought and welcomed!

Comments

There are no comments yet
Overclock.net › Member Blogs › An Interpretation Of Amds Long Term Strategy