Death By Decapitation: EV8 is Chopped
Monday’s announcement was certainly excellent news for Intel’s IA64 product planners and strategists. In one fell swoop their most technologically powerful competitor in the 64-bit MPU field was instantly neutralized. Up until the axe fell the majority of Alpha logic and circuit designers were deeply involved with the hardware implementation of the EV8 microarchitecture, the most ambitious microprocessor yet disclosed. This work stopped immediately, and as a result the EV8 will never see the light of day. Compaq claims to be committed to the multi-year effort to bring systems built around the still formidable EV7 to market and take the final Alpha through one process shrink. But with the end of the architecture in clear sight, the EV7 will likely have little impact on the competitive landscape no matter how much it eclipses the performance of competing processors.
The simple elimination of Alpha on its own would have been a positive development for the remaining RISC players, IBM and Sun. It removes a potentially powerful competitor as well as provides an opportunity to pick up disgruntled Compaq customers. But Compaq’s course of action makes all Alpha IP, and potentially most of the accumulated Alpha design experience, available to Intel at little cost. The direct processor implementation know-how that the Alpha design group encompasses may be of marginal immediate value to the IA64 effort, because Alpha’s minimalist RISC and IA64’s EPIC architectural and design philosophies are diametrically opposed. But what they lack in practical experience in designing arcane VLIW hardware is counterbalanced by bringing in fresh design approaches, methodologies, thinking, and mindsets to bear on the problem (Intel engineers freely acknowledge they learned a lot of valuable lessons and insights from absorbing the StrongARM design). The Alpha compiler group could be more important to Intel in the short term in making IA64 more competitive. It is not widely understood, but even out-of-order execution superscalar RISC processors like Alpha benefit greatly from the type of compiler-based code optimization and instruction scheduling that is absolutely vital to EPIC processors.
The real question is how many of these freshly minted Intel employees will stay. It is quite obvious that Intel wants to avoid the mass exodus that accompanied the StrongARM design team transfer. They have extended generous and accommodating terms to the Alpha design group and clearly want to retain as many of them as possible, and keep them together as a cohesive team to work on future IA64 processors. Intel’s success in retaining these highly sought after engineers will not be apparent for months if not years. But there doesn’t seem to be anywhere near the level of acrimony and bitterness that accompanied the DEC settlement four years ago. Another important factor may be the scarcity of fresh opportunities in areas like network and embedded control processor design due to the current worldwide recession in the semiconductor industry. AMD is not likely to benefit from the current situation like it did four years ago because the Alpha hardware designers that stayed did so in many cases because they didn’t want to do x86 MPU design.
Discuss (16 comments)