The MIPS result in SysInfo is rubbish and can't be compared to a 68060.
The Sysinfo results are real and you can compare Phoenix with any other 68K with them.
As probably all people know Sysinfos benchmark is difficult for Super Scalar CPUs.
This means the benchmark is written in such a way that a Super Scalar CPU can not make full usage of a possible 2nd instruction execution per cycle.
But this drawback does affect Phoenix and 68060 the same way.
The Phoenix Core can execute several instrutions per cycle.
While this features is NOT enabled in the current Vampire loads, Phoenix can do this.
Execution of 2 instrutions per cycle can be enabled in bigger FPGA systems
Like for example in the ApolloPhoenix card or in the NATAMI.
I don't know why the system libraries still appear in "SLOW RAM".
This just depends on the mapping options used.
Phoenix has some sort of tarnsparent MMU feature.
Phoenix supports automatic MAPROM - loading of Kickstart to fastmem and "MMU" mapping it to Kickstart memory region.
Phoenix also supports re-mapping of memory slices.
I've configured it to map 1 MB of fastmem to "slow-ram" area.
This is good for game compatibilty as some games expect to find memory at $c00000.
Memory could also be mapped to Zorro 2 and or Zorro 3 range.
This way up to 64 MB are then visible and usesable in the Vampire cards
and 128 MB are useable on the Phoenix card.