MIPS Technologies Inc has formally introduced the long-gestating R10000 family of 64-bit RISCs, claiming that the 275MHz part puts its architecture back out front again, besting the fastest of Digital Equipment Corp’s current crop of Alphas. The R10000 integrates 6.8m transistors, and the 275MHz version is rated at an estimated 12 SPECint95 and 24 SPECfp95. The initial version of the 200MHz R10000 processor from NEC Corp and Toshiba Corp is now moving into production. Large integer software, such as database applications, show a 1.5 to 2.0 times performance increase over the fastest existing MIPS general purpose processor, the 250MHz R4400 processor, the company says. Floating point-intensive applications that include single precision, scalar and irregular vector computing, including graphics rendering, are two to three times faster than R4400-based systems. The R10000 features superscalar Andes Architecture with Non-Sequential Dynamic Execution Scheduling. The four-way superscalar part fetches four instructions and issues up to five instructions per cycle and has five independent fully-pipelined, low-latency execution units. The on-chip cache has 32Kb instructions, 32Kb data.