When IBM Corp and Motorola Inc launched 120MHz and 133MHz versions of the PowerPC 604, no pricing was given and volume production was set vaguely for sometime in the second half of the year. The big news, however, is that the companies seem to have significantly over-estimated the integer and floating point performance of the […]
When IBM Corp and Motorola Inc launched 120MHz and 133MHz versions of the PowerPC 604, no pricing was given and volume production was set vaguely for sometime in the second half of the year. The big news, however, is that the companies seem to have significantly over-estimated the integer and floating point performance of the processor. The 100MHz part was originally estimated to have a SPECint of 160 and a SPECfp of 165. Things have now been scaled back a bit with the announcement of the new versions – the 120MHz part was estimated to have both integer and floating point SPEC ratings of 180, while the estimated figure at 133MHz was 200MHz. But actual measured figures for the new RS/6000 models show the processors falling short of expectations. They also show the extent to which it is unwise to assign SPECmarks to a chip, rather than an entire machine. The new low-end RS/6000 43P machine shows a 100MHz 604 with 256Kb of asynchronous level 2 cache producing SPECint 128.1 and SPECfp 120.2. Doubling the size of the cache and making it synchronous improves the figures to SPECint 140.8, SPECfp 129.1. At 120MHz, and with the 512Kb cache, the machines manage SPECint 157.9 and SPECfp 139.2. The speediest, 133MHz, part is measured at SPECint 176.4 and SPECfp 156.5. In summary, a 133MHz 604-based machine is carrying out floating point operations more slowly than the alliance suggested a 100MHz chip would, while integer figures at 133MHz are worse than we were led to expect from the 120MHz part. It’s also worth noting that the integer scores are consistently better that the floating point figures, which is again a change from the expected results. It’s not all gloom for the PowerPC camp: IBM quoted Pentium figures from Intel Corp that show the machines running 20% faster than a comparable Pentium-equipped box at the same clock speed, while floating point is 40% better. The quoted figures for a 133MHz Pentium is SPECint 147.5 and SPECfp 109.6. There are two possible explanations for the discrepancy between the estimated and measured figures: either the alliance’s estimating methodology is out of kilter, or IBM’s machines simply don’t do justice to the 604’s raw performance. The former would be something of a blow to the alliance, which has made much of its conservative estimates in the past. The latter will not be too palatable to the Power Personal Systems division at IBM. As for the next stop, IBMers say that 150MHz and 180MHz versions of the 604 are planned, but no time-scales were given. Motorola is the one member of the alliance that has yet to announce machines using the 120MHz or 133MHz processors.