I just had a look in to hte A2630 manual and I`m right. The RAM on the card is mapped into the 8 MB adressspace.
Back then 8 MB was an enormous amount of RAM so only few would even reach that barrier due to cost of RAM alone.
What was done by more modern accelerators was that something simmilar to addmem was build into the cards ROM that assigned adresses outside of the 8 MB window to the installed RAM and tells the CPU where to find it.
The drawback of that is, that RAM outside of this 8 MB windown can not be seen directly from any ZII card, so no ZII DMA would be posisble into that RAM.
Today this "drawback" is totally neglectable as cards that use ZII DMA a re few and slow as hell anyway, plus a 030 CPU can shuffel data much faster around then any ZII DMA possibly could.
Back in the time the A2630 was made the world was a little bit different, so having the RAM in th ZII adressspace was more important then being able to add more then 8 MB to the system.
What I meant with "real" accelerator was that the A2630 is quite ancient,
having maximum 4 MB 100ns ZIP-RAM and all.
There are only two and a half reasons to use an A2630 today:
1a: You want to be as compatible as possible while having a little bit more horsepower
1b: You are a collector and want to have a real classic vintage machine with Kick 1.3, A2091, A2058 and all
2: You don`t have the money to efford a more modern accelerator :-(