Work RAM High & Low

What is the difference between High-RAM and Low-RAM?

Q)

What is the difference between "High RAM" and "Low RAM"? I would appreciate it if you could answer from a hardware point of view, such as wait cycle and RAM speed ...
Also, I think that the cache will be effective when actually using it, but how effective is the cache on those RAMs?

A)

First of all, regarding the addresses where both are located, WORKRAM-H uses synchronous DRAM, from 06000000H to 1Mbyte, and WORKRAM-L uses DRAM, which is from 00200000H to 1MByte.

This is the address at the time of cache, and at the time of cache through, WORKRAM-H is from 26000000H to 1Mbyte, and WORKRAM-L is from 20200000H to 1Mbyte, respectively.
(See HARDWARE MANUAL ■ 1.2 SCU Mapping )

In the read operation, in the case of a cache hit, WORKRAM-H can read without weight, but in the case of a cache miss, access is required for 16 bytes.
With the synchronous DRAM used in WORKRAM-H, access is performed at extremely high speed when a cache miss hit occurs, and 16 bytes of information can be acquired in about 7 clocks.
However, the space of WORKRAM-L takes 14 (clocks per long Word) x 4 clocks (56 clocks).

When writing, it takes 4 clocks for WORKRAM-H and 14 clocks for WORKRAM-L for 1 longword (7 clocks for 1 word), but it is continuous because it has a write buffer for 1 longword. In case of writing, it will be faster than this.

As mentioned above, since the cache speeds of WORKRAM-H and WORKRAM-L are different, it is recommended to put the program on the WORKRAM-H side as much as possible. Others, WORKRAM-L will not be able to use the DMA transfer of the SCU (transfer destination, transfer source both), so pay attention please.

This is because the space of this WORKRAM-L is outside the control of the SCU.
As a method for speeding up SH2 programs, it is advisable to write programs with pipeline processing in mind and efficient use of cache.