its "simple":
MIPS more fool-proof. both firmware/hardware/software developers was hard to break it[with excptly specified code/data pages doing stack overflow wasn't as trivial task as in x86, for example], even intentionally, also it hardened against lazy programming a bit, thus consistent/predictible dataflow - easier.
also many if not ALL simply forgot that MIPS, along with Berkley[presently evolved(ospfied)into Sparc] - oldest RISC architectures, aside PPC.
many features arch got earlier than others do. they got[along with Inmos chips]decent built-in FP unit[some commercially-unsucessful FP units of ancient top-end MIPS FP units - out-gunned only by Cray HW], SIMD, DMA-based/singe int periperal[acorn/amiga-style], decent/impressive Out-of-Order Exec/Sheduler/Decoder/Fetch[despite troubles w/implementing this on arch], decent memory controller, MIPS was actually first[1991?] before Dec Alpha[1992] 64-bit commercially successful chip, until AMD rolled AM64 and etc and etc.
precently MIPS only lack most fat/hot[in terms of transistors/heat]portions, ie OoO exec, fast/wide Floating point unit, probably augmented by DSP/GPU, like in promised Q in Snapdragon 5[transparent offloading FP and partially INT/GPR/ALU workload to GPU w/o re-code/re-compile SW] and future AMD chips.
future Tilera chips will be based on new MIPS arch and become, tnx to both optical internal bus and advanced interconnect[lightweight(in terms of transistors count), but fast/flexible], new 64-bit arch with improved sheduler/exec - something WAY more interesting than present MIPS chips.
presently Tilera chips only limited by C/C++ software engineers/software developers mentality/approach/values. i wonder what happen when/if they adopt Erlang for example. those pesky semaphores/mutexes, shedulers, heap and other lazy/redundant/insecure/not scalable stuff from 50yrs old Unix/C era - heavily bottlenecking that way.
recent MIPS Aptiv cores had COOL Out-of-Order[but quite simplified/lightweight, which is reasonable]sheduler/exec/fetch, nice DSP/FPU, improved memory controller and internal interconnect[which isn't as powerful as Tilera chips, but can help you to make at least 4xcore and 8xcore CPU/controller/ic w/o serious engineering investments] and etc and etc.
ie, MIPS was return back. to basics.
presently in China and India - MAJORITY of ultra-affordable tablets/UMPC/phones/netbooks with price-range from $20 to $70 was MIPS-based.
you can't see such stuff x86-based[even on VIA].
even ARM chips can't do it[fit low-end devices/markets needs].
compare benchmarking of 74Kf and 1074Kf[MP 74K derrivative/model]and ProAptiv and InterAptiv chips on internet for reference.
thats stuff could actually make MIPS handy for destop use AGAIN, aswell as for HPC[just like FR-V and ARM already by Fujitsu and BSC,Samsung respectively].
which actually nearly employed in China for example[
https://en.wikipedia.org/wiki/Loongson],aside microcontrollers for undustrial[including military]use.
you can get simple/short datasheets Directly from MIPS[after registration]
https://www.mips.com/products/processor ... ips32-74k/
IMO, 1.5Ghz RISC chips[15 stage pipeline! way shorter than others. mean - LESS latency!] on 40nm wafer - sound impressive[as perform].
https://www.mips.com/products/processor ... /proaptiv/
http://www.anandtech.com/show/5826/mips ... tiv-series
that beated ARM-A15 into dust in benchmarking/workload !!