Dear Tench:
1) The SPEC scores were suspicious as well. 20% increase with no change except compiler. That kind of increase shows that all of the increase was optimization, not that the underlying CPU was any better. The fact that it could not produce working code, EVEN AFTER INTEL TRIED TO, shows it produces scores only on SPEC and look alike applications (the narrow balance point I speak of). That is not my definition of a good compiler. When comparing sub optimal code in all of Moldyn, which was never optimized for either CPU, Athlon beat P3 even more.
2) Intel is suspiciously not allowing any benchmarks of their P4 to get out. We have only one data point. That datapoint shows a severe reduction in IPC. The CPU keeps getting pushed back ala Itanium. Now we know for sure it is because Itanium does not run at the speeds marketed previously. Thus, the simple conclusion is that the same is happening to P4. Many more samples of Athlon were shown much earlier than the supposed release of P4. Also many who do see it, keep saying it is underwhelming.
Thus, it is either the sheer speed that is lacking or that the benchmarks currently used show little, if any, improvement. Reading between the lines, a 30% to 40% degradation is quite likely as the promotions of low processing and high bandwidth applications seem to indicate. My K6-3/400 is quite capable of playing DVDs in software (read MPEG-2 conversion) at a variable rate between up to 1.35 MB/sec or 10.8 Mb/sec. That is faster than 10 BaseT ethernet. Far faster than 56 Kb/sec. You would not say that a K6-3/400 is a fast CPU by today's standards. The benchmarks of choice for P3, Quake, and MM encoding, are not even mentioned (really bad performance with those?).
This is Intel's IDT. Where are the P4s to try out? Why can't we see benchmarks or even side by side comparisons to the shipping P3 1.0 GHz much less the limited 1.13 GHz? Either the P4 will not be released within 4 months or they do not like any comparisons against the current samples. When we see some benchmarks, then you can say it is better than the theory. But, the definitive answer will come only when it is looked at by the world at large.
Pete |
