Intel Needs AMD

They’ve been helping them for decades – Intel needs AMD to keep the FTC off of their back and have for years. How do they help? Look at pricing. The per chip cost on a Core 2 is about $50 – they could sell the E6300 at $80 and still make a heaping profit and totally crush AMD. They’re two years ahead of AMD on process technology (the only one that exceeds Intel is IBM, but not by much), and they have double the per wafer yield (over 80%). Intel is positioning itself as the premium product supplier to AMD – and they’ve suceeded. Intel platforms have superb reliability compared to AMD (mostly the fault of the chipset, but I digress). Also, Intel has had technology sharing agreements with AMD for years upon years, and they keep making more. Yet somehow I don’t see a single AMD technology, EM64T aside, that Intel has adopted. Yet if you look at the RAM standards, AGP, PCI Express (yes, PCI-SIG was founded by Intel), USB 1.1 and 2.0, integrated math coprocessors, out of order execution, decoupled pipelines, cooling technology including fan control, the ATX motherboard standard, clock generators, the final demise of ISA, Serial ATA, and even integrated Wi-Fi and softmodems were all pushed by Intel and have all since been integrated into AMD products at some level.

IBM was not the driving force behind the K8, they were the driving force behind the process technology and that’s it. They still are to this day the driving force behind AMD’s processes, even moreso than they were when the K8 launched. And even now the K8 has scaling issues that have never been resolved, so they never penetrated the Big Iron market where the highest margins are (Big Iron = 8 CPU or larger, single OS image). Additionally, the K8 was designed by the old DEC team that worked on the Alpha processor (remember that?), not any IBM alliance. If you look at an architechural level, the K8 is very different than any IBM processor, such as the PowerPC or POWER5 line.


If you think Intel’s cache is what’s proping them up, look at the numbers on the Core 2 when it had a smaller cache (and there are smaller cache versions available). They’re not that far off, and are highly competetive with AMD’s product lineup, exceeding their line in performance per dollar and performance per watt at every level. And keep in mind that the Core 2 is based on a chip that was first launched in 1994, which keeps the transistor count low and allow them to impliment additional cache.

“The connection between the companies? We share underlying breakthroughs in areas of chip design and manufacturing, such as transistors, chip connection, packing, and lithography, which stand to benefit both companies.”

Exactly as I said, they’re colaborating on process technology, not on chip design. If you read further down you see where the chips get used by IBM, they sell machines with AMD processors, like they have been doing for the most part since the early 1980’s where roughly 20% of PC-ATs shipped with AMD processors.

As far as performance, we’ll wait and see, but they’re still stuck on the HyperTransport bus that doesn’t scale well at all.