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 give me an answer from a hardware perspective, such as wait cycles and RAM speed...
Also, I think that caching will be effective when actually using it, but how effective is caching on those RAMs?

A)

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

This is the address when caching, and during cache through, WORKRAM-H is 1Mbyte from 26000000H, and WORKRAM-L is 1MByte from 20200000H.
(Refer to HARDWARE MANUAL ■1.2 SCU Mapping )

For read operations, in WORKRAM-H, if there is a cache hit, the data can be read without waiting, but in the case of a cache miss, 16 bytes must be accessed.
With the synchronous DRAM used in WORKRAM-H, access in the event of a cache miss is extremely fast, and 16 bytes of information can be retrieved in about 7 clocks.
However, the WORKRAM-L space takes 14 (number of clocks per long word) x 4 clocks (56 clocks).

When writing 1 long word, it takes 4 clocks for WORKRAM-H and 14 clocks for WORKRAM-L (7 clocks for 1 word), but since a write buffer for 1 long word is provided, it can be written continuously. Writing is faster than this.

As mentioned above, the cache speeds of WORKRAM-H and WORKRAM-L are different, so it is recommended to place programs on the WORKRAM-H side as much as possible. Additionally, please note that WORKRAM-L cannot use SCU's DMA transfer (both the transfer destination and transfer source).

This is because this WORKRAM-L space is outside the control of SCU.
The best way to speed up SH2 programs is to write programs with pipeline processing in mind and efficient use of cache.