Rock

Perhaps you wondered a little bit about the fact, that i didn´t wrote about the reports about the delayed Rock processor. I wrote an article, but this contained some internal information, as but i´m still not an official leak (i should talk with Jonathan about that ;) ), i just deleted it. Okay, the delay of Rock was a rumour long before and even the sparrows and the nightingale sang that from the roofs. As I wrote before: That isn´t much a problem, as we have two strong new procs in this year: The Quadcore Jupiter Sparc64 and the Victoria Falls boxes. Those offerings were the reason we didn´t got butchered by the analysts. It isn´t the situation where we wasn´t able to compete on the proccessor level at UltraSPARC III vs. Power4 times (interestingly we won many customer benchmarks even at that time despite of the disadvantage in the proc field. A good point for the old speaking: You buy systems and knowledge, not processors). That was really a hard time for us. Don´t want them back … Today: Single proc Niagara 2 is able to compete with low end Power systems for many workloads, the M-class systems are competitive with the mid and high end offerings of IBM. And VF and Jupiter will make the situation even more comfortable soon. Victoria Falls will open a complete new market for us this year. Think about a 256 thread OLTP system. Think about a 256 thread webserver. Think about a 256 thread Lotus server. Think about an 256 thread SAP application server. But i wrote about this earlier. Furthermore: Delays in the processor business are normal, not an exception. Think about Power6 (it came late and despite of reports of IBM friendly commentators the 4.7 Ghz is still largely missing in action). Think about Itanium (“The next version will kill all other procs! Really!”). An delayed or underdelivering Tukwila would be a disaster for HP. And the problems with AMD are in my opinion the most sad story of the business. I´m still strongly opinionated in the direction of AMD as the much better x64 architecture (and Quickpath and integrated memory controllers in next-generation Intel CPUs support that, they just call it different than AMD) , but AMD leaved out a window of opportunity to hurt Intel´s marketshare with the delay of Barcelona. I never talk about competitors proc slipages at customer sites (such arguments transform to boomerangs to easily most of the times), i only hate benchmarking with procs unavailable to the masses … Some commentators (interestingly not the media itself, those comments come from the “reader comments”-section) suggest that we kill Rock in favour with a longer agreement with Fujitsu. Well, this is utter nonsense from outsiders. We did the tapeout, we have running and stable silicon, we have just a few things left to solve - in short: We did the hard part already, why should we kill of Rock now? Furthermore, we need Rock in the long term. This proc is important to keep the SPARC franchise healthy. Well, i would have prefered Rock at the original schedule …. because of the fun to shock the competitors with the performance of the proc, but hey … the difference won´t be that extremely large (you know, competitors doesn´t sit still), but we will be able to kick some serious butts to have fun with the systems ;) And remember: Delays in the proc business are normal, not an exception. That´s true for every architecture.