AMD will keep making Intel chipsets

In an interview with DigiTimes, Jochen Polster ( AMD’s Worldwide Infrastructure Partnerships sales and marketing VP) stated “AMD will not abandon or limit any opportunity to cooperate with others and the company will still continue developing chipsets for Intel platforms. Our goal is not to amass a huge market share, but to obtain a reasonable share.” in rely to a query as to whether AMD would continue manufacturing chipsets for Intel as ATi did before AMD acquired them.

How much market success there will be I’m dubious, even in the event that AMD’s makes a terrific chipset for Intel’s processors you can be certain Intel will be putting all of their weight on manufacturers not to touch said chipset.
This isn’t the late 90’s wherein Intel’s weight is such that even powerful companies like Asus have to secretly manufatcture motherboards and disavow all support and publically deny they made them as was the case in the early days of the early K7, nor as hostile an environment. No one is terrified of losing Intel’s business and their mighty discounts anymore. Intel can no longer afford to do such things as easily as before, and AMD is now a legitimate alternative in the market for companies to go to.
With many alternative options out there for Intel chipsets however it’ll be easy to accede to Intel’s wishes. Even typically ATi friendly companies like DFI dropped their ATi based Intel motherboards almost immediately after their acquisition by Intel, and naturally Intel dropped all mention of the exiestance of such products from their site.

It doesn’t appear Intel has much options to revoke ATi’s bus license so in the short term such chipsets should remain in existance, but once Intel’s long awaited CSI appears you can be certain Intel won’t be selling a license to utilize the bus to ATi at any cost.

Hell has frozen over

While it may not be nearly as exciting as when the Berlin Wall fell, it is a pretty historic day for AMD. Dell has finally crumbled under the pressure and will adopt AMD processors for its high end 4 socket servers. As you all know, Dell has been historically an all Intel shop, and really only in its first years of existence did it ever use AMD chips as a second source. Since that time Intel has offered Dell outstanding deals to stay all Intel, as well as basically handle all of the processor inventory needs of Dell. This was obviously a very good deal for Dell, as they got matching marketing dollars, first pick of new processors, and outstanding support from Intel. It hasn't always been easy for the two of them, and times have been strained before. The last time there was any significant break between the two was when AMD had first introduced their 1 GHz Athlon processor, and for the next six months supplied 90% of the 1 GHz+ market. Dell was in negotiations to create a product line based on the original Athlon, since it was beating the Coppermine based Pentium !!!'s at the time. They even went so far as to enter into negotiations with MSI to provide the motherboard that would be used. Intel gave Dell a lot of concessions at that time, and they promised that the low quantities of 1 GHz P3's would eventually end. They also promised Dell that they would get first pick of the Pentium 4's that would be showing up in the Fall of that year. So, Dell dropped the idea of an AMD based line, and kept chugging along with Intel only solutions.

Now that we are in 2006, Dell has started to see their server shares erode due to increased competition from guys like HP and Sun. Both of those manufacturers feature a wide selection of AMD server parts, and the four socket variants are selling very well in that space. There are several reasons why the Opterons are really popular there, and mainly it revolves around price, lower power requirements, non-uniform memory architecture (NUMA), native 64 bit support, and the glueless MP that HyperTransport enables. While Intel and AMD are about on even footing in many ways when considering 2 socket architectures (AMD uses HT, while Intel features a dual FSB, so neither really have a bandwidth advantage when it comes to communication between processors and the memory subsystem), the situation is drastically changed when dealing with 4 socket and 8 socket products. Scalability on the Intel side takes a pretty dramatic hit once we go beyond 2S/4P products. Things get pretty bad for Intel once 4S/8P gets thrown into the ring. A lot of the bandwidth between processors and main memory is taken up by the cache coherency protocols, while AMD does not suffer as much due to their more comprehensive MOESI protocol, as well as the three HT links that each processor features.

Dell, throughout the years, has claimed availability as a negative for adopting AMD products, and they were very right in saying so. Until this Spring, AMD only had one Fab producing CPU's, and at the beginning of this year they were selling nearly every good die that came off the line. They simply didn't have the production capacity to fulfill orders from a company the size of Dell. Now that Fab 36 is ramping up, AMD now has a lot more manufacturing flexibility. While Dell will not officially release a AMD product until late this year, they will start stocking up on processors and enable the line to become functional so Dell can have a full blown release. This should mean some nice extra income for the next two quarters until Dell officially unveils the line and starts selling product.

It is interesting that Dell jumped in now, and I think another enabling factor is that Dell is pretty much all DDR-2 now. While current Opterons are all DDR-1, this July we are supposed to see the introduction of the Socket F Opterons. These will be 1207 pin Revision F chips, and it seems very likely that the extra pins could mean expanded HyperTransport links between the processors (16/16 links moving up to 32/32 links). This will give the Opterons a lot more bandwidth to communicate with each other (if in fact true). These chips will also support DDR-2 for server configurations, which will be a definite plus for Dell. AMD is not shirking away from Woodcrest this Summer, and I think that the answers that AMD will provide to this new architecture will be interesting to say the least. Dell is a big win for AMD and their server level aspirations. While Dell will not likely release lower level products anytime soon, AMD could win Dell over as a customer by continuing to support the company well, and produce quality processors that can address multiple markets.