Why is it that every time I read an editorial about AMD, it starts with the author portraying them as a sinking ship and ends with them begging to come out with a product that can compete with Intel’s Core 2 series? Sure, AMD’s Athlon 64 series of processors has been eclipsed by Intel’s latest Core 2 series, but has everyone forgotten about the past? AMD and Intel have changed roles as being the pace setter in the microprocessor industry for years, so in a sense this is nothing new. I hate to say it, but the last couple leadership changes were back in September 2003, with the Athlon 64 and its innovative integrated memory controller and again in March of 2000, when AMD beat Intel to the 1GHz mark. Both of these shifts in the industry happened before everyone wrote a blog or had a website, so maybe everyone has just forgotten the past or just wasn’t involved in the tech industry just a few years ago. It’s a shame really, as not only should the past must be remembered, it’s a critical part of journalism. So, forget thinking about AMD as a sinking ship and I’ll tell you the reasons why.
The past few years Intel and AMD have been in a fierce price war that drove prices down for consumers, but at the same time killed their profit margins, which had a negative impact on their stock prices. In the mean time, consumers have gotten used to lower processor prices and see this as a win-win situation; and for those at the checkout aisle, it is. Since Intel processors don’t have an integrated memory controller, the processor has to access the front side, but to fetch data from the system memory (DRAM) and this action takes place on the Front Side Bus (FSB). To help speed up the Intel Core 2 series of processors, Intel has been increasing the front side bus from 800MHz to 1066MHz to 1333MHz and in the future we will certainly see the move to 1600MHz. Think of this as increasing the speed limit of a major interstate, the faster the speed limit the quicker you can get to your final destination. By increasing the frequency of the front side bus, the system memory can be reached quicker, but why not move the data closer?
Well, Intel is also doing this by increasing their L2 cache. When the Intel core 2 microarchitecture was first released they had an L2 cache size of 2MB on Allendale to 4MB on Conroe. Just this week Intel introduced a quad-core part called Yorkfield that features a massive 12MB L2 cache. This now means that Intel has added 66% more L2 cache per core with the move to 45nm over the original entry level Core 2 processor series and 33% more when compared to the Conroe family of dual-core processors! Since the Intel quad-core ‘Kentsfield’ processors are basically two ‘Conroe’ cores paired together, they add up to 8MB L2 cache. So Intel has added 2MB of L2 cache or 66% more over the past couple years. Something I hear from the community here on Legit Reviews is, “Why doesn’t AMD add more cache?”
Well, AMD has been adding L2 cache. The first Athlon 64 X2 ‘Manchester’ processors featured 512KB L2 cache per core for a total of 1MB L2 cache. AMD then doubled the cache on Athlon 64 X2 ‘Windsor’ series to 1MB L2 cache for a grand total of 2MB of L2 cache. With the upcoming Phenom quad-core series of processors AMD has kept the 2MB L2 cache and has added 2MB L3 cache which is better for their microprocessor architecture. AMD Athlon 64 and Phenom processors have an integrated memory controller, so they don’t have to use a front side bus to access the memory. So, AMD Phenom processors (both quad-core and triple-core) will have 2MB of L2 cache and 2MB of L3 cache. Which is obviously an increase in cache, although it can’t be compared to what Intel is doing. AMD Phenom processors can also take advantage of an improved Hyper-Transport bus if one is used on the 7-series of chipsets as these boards have Hyper-Transport 3.0, which is a speed bump from 2400MHz to 3600MHz. Right about now you might be wondering why I’m rambling about L2 and L3 cache, but it all boils down one thing and the end of the day and its price. AMD Phenom processors may not have the clock per clock muscle to beat out the Penryn and Yorkfield Intel processors, but it costs far less to make them partially thanks to the overall smaller cache sizes.
When it comes to cost, die size plays a critical role and Intel has the 45nm process down pat, but nothing is wrong with AMD’s 65nm process. Sure the smaller the nm the more dies can be put on a wafer thus reducing costs, but remember 12MB of L2 cache versus 4MB of L2/L3 cache makes a difference too. AMD has millions fewer transistors per die, which means lower power consumption, less heat generating transistors and ultimately lower prices. Of course AMD needs to start moving things over to a 45nm process, sooner rather than later, but it doesn’t have to happen today.
The take home message here is that AMD, from my point of view, is not a sinking ship. They just need to make sure they get out their next generation of processors and chipsets out on time. AMD has been delayed getting numerous products out the door as scheduled and that is without a doubt hurting them. Sure AMD will have the first ‘true’ quad-core processor, but how many consumers really cares if something is true or not? Could you imagine walking into Best Buy or Circuit City and hearing a sales associate telling a customer “This one is better because it’s native quad-core, but the other one is quad-core”. That wouldn’t obviously work with the average consumer, which is the direction AMD has been aiming the majority of their marketing and product series. The best marketing tool AMD has is to sell the complete package. By offering processors, chipsets and video cards all under the AMD brand name they have the ability to tweak each component to make sure it’s stable and at the right price point. If AMD falls a bit short on performance they can lower the prices on their components 10-20% across the board making their platform the best bang for the buck. Since their products use fewer transistors and cost less to make, it means that Intel and Nvidia has more to lose, should they choose to engage in another price war. Just this week AMD’s master distributor, ISA Hardware leaked out Phenom pricing and has shows the preliminary prices as $247, $278 and $288 respectively for the processor in a box versions of the the 2.2GHz 9500, 2.3Ghz 9600 and 2.4GHz 9700 quad-core processors. This means that right now BEFORE launch a 2.4GHz Phenom is priced at $288 and a 2.4GHz Kentsfield is $269.
Hopefully after launch the price on the Phenom processors will drop below that of the Intel Q6600 as it needs to be, if they want to be competitive in the enthusiast market.
Lastly, there is always the chance that Intel slips up and if there was to be such a time it would be on their Nehalem based system architecture. After mainstream Penryn and the 45nm Hi-k silicon technology introductions in January 2008 comes Intel’s next-generation microarchitecture (Nehalem) slated for initial production sometime in 2008 (five years after AMD went to an integrated memory controller). Nehalem will be Intel’s first processor with scalable and configurable system interconnects and integrated memory controllers. That will be a major change for Intel and history has shown if a stumble is to happen it is at one of these junctures.