The most annoying thing here is the complete lack of information on caches.

If it's the same as K1/M1 then ok, 512 MB L2 per 4 core cluster is simply inadequate for a normal Linux workstation workload.

I don't know where they (or SpacemiT) get the "30% faster than A55" claim because I certainly haven't seen any such thing. On some micro benchmarks such as my own primes one the K1 comes in at slightly more clock cycles than the U74 in JH7110, but pulls slightly ahead on high clock speed. Both are right around A55 performance, and 30% faster than A53.

On real-world tasks such as building software (e.g. Linux kernel, gcc) all 8 cores on a K1 working together come in slightly slower than the 4 cores on the JH7110 -- not much in it, call it the same. If you use `-j4` then it's miles slower.

I'm pretty sure that's down to the smaller caches. Or possibly TLB misses. U74 has 40-entry L1 Data and Instruction TLBs, and a direct-mapped 512-entry L2 TLB -- bigger than A53, smaller than A55. I haven't seen any information on the X60 core's TLB structure.