In 1995, MIPS released the R10000. This processor was a single-chip design, ran at a faster clock speed than the R8000, and had larger 32KB primary instruction and data caches. It was also superscalar, but its major innovation was out-of-order execution. Even with a single memory pipeline and simpler FPU, the vastly improved integer performance, lower price, and higher density made the R10000 preferable for most customers.
More recent designs have all been built on the R10000 core. The R12000 used an improved process to shrink the chip and run it at higher clock rates. The R14000 bumped the speed again to up to 600MHz, added support for DDR SRAM in the off-chip cache, and increased the SysAD bus speed to 200MHz for better throughput. The most recent version, the R16000, doubles the size of the caches to 64kB for both the instruction and data cache, adds support for up to 8MB of level 2 cache, and bumps the clock rates once again, to 700MHz.
R10000 family memory coherency issues
The R10000 features aggressive out of order execution including speculative execution of loads and stores. On systems with coherent I/O this does not pose a problem. On non-coherent systems this will cause serious memory corruption. Linux therefore currently does not support these systems. Affected systems are the Indigo 2 R10000 and all SGI O2 with R10000 family processors.
The issue exists in all R10000 family processors. R12000 and newer implement mode bit which faciliates a more efficient workaround but for unknown reasons this feature is not being used in the O2.
- R12000 processor manual