ISA is still a significant factor for many server users and even in embedded systems there is some attraction to not needing to use cross compilation (i.e., x86 compatibility has some value). Even those who could somewhat easily transition to ARM from x86 are likely to wait until support is more common and risk is lower.
x86 offerings could have a slightly higher cost/power per unit of performance (or higher per thread performance) to stay on message that ARM is "better" if compatibility (or high single-thread performance) is not important without hurting x86 sales too much.
In some ways, it is the OEMs that will have to sell the hardware. Traditional x86 server vendors might be more inclined to deal with AMD for ARM processors than a company new to servers. Being a supplier with more diverse offerings (as well as motherboard compatibility) may give AMD an advantage.
AMD might not have had much choice but to offer ARM processors simply to appear relevant in the server market. In addition, if AMD did not enter the ARM server market before it became significant (if that happens), those who entered first would have a significant competitive advantage. On the other hand, if AMD abandonned x86, their unique assets related to x86 would become valueless and near-future x86 sales would shrink. If they already need to develop x86 cores for the personal computer market, using these designs for servers does not add much cost.
On the other hand, who can guess what the AMD leadership are actually thinking? They might be just trying multiple directions in the hope that something works.
You make some good points. It appears that they are planning on keeping both the internal ARM design for the high end and the ARM design at the low end. But that is what is confusing. How will they differentiate their ARM offerings vs. their x86 ones. It appears that they are just putting both out there and hoping for the best.
I receive the impression that the use of ARM processor designs was a time-to-market choice and that later AMD processors, at least for high performance designs, would use internally designed cores.
With respect to the common motherboard strategy, this seems to present AMD as a more attractive choice than other ARM vendors in that inventory issues are reduced (cost and risk). This might allow AMD to reacquire some server share from Intel in x86 and take a significant portion of any ARM server market.
For servers, competing with Intel is nearly inevitable. It seems the only way to avoid that would be to choose a specialized portion of the larger market which Intel does not view as worth pursuing. However, that severely limits revenue, which is significant for the high fixed costs of processor design and manufacture. Even targeting a niche would not prevent competion from companies using ARM.
It seems that AMD is, to some degree, focusing on more specialized markets, which is somewhat appropriate for a smaller, "hungrier" company. AMD cannot be expected to compete head-to-head with Intel given Intel's advantages in revenue, process technology, and vendor relations. With the increasing credibility of ARM as an alternative to Intel, AMD's position as a second source for x86 is weakened, so relying on x86 alone is more dangerous.
AMD's historically inconsistent execution may also encourage more numerous smaller projects rather than a few high-effort projects. Intel's advantages also work better for high effort projects (which tend to be high volume) and Intel is more likely to pursue any market that would justify higher effort.
The above is just speculation based on general concepts and some common knowledge of the field.
What is AMD' strategy? I am confused on what they are trying to achieve. They will have low performance ARM and x86 and also high performance ARM and x86. ARM and x86 processors will be interchangeable. They will use CPU IP from ARM and also design their own. They will not only compete with Intel but also all other ARM players.
It looks like a "let's try everything and something will work out" strategy.
Not to belabor the point but the original K8 architecture that Keller came up with was completely abandoned. The K6 team with Greg Favor (K6 architect) had also left so Fred, Kevin McGrathe and the Austin team (Scott White and others) made 64b extensions to the K7 design which ended up becoming Opteron.
Well, having been around the processor and CPU industry for decades one keeps track of CPU designs, their architects and so on. The Opteron design was based on K7. Opteron was the 64b version of K7. I believe there were numerous articles on this at that time. Keller did not have anything to do with Opteron.
January 2016 Cartoon Caption ContestBob's punishment for missing his deadline was to be tied to his chair tantalizingly close to a disconnected cable, with one hand superglued to his desk and another to his chin, while the pages from his wall calendar were slowly torn away.122 comments