ZorRam Speed

[desantii]

I am planning to add ram to my 4000T once its fully functional. For noww keeping the 3640 so ram options are limited. Is the 128mb ZorRam faster that the onboard motherboard fast ram?
If so will there be any noticeable spped increase?
 
thanks

[Piru]

I am planning to add ram to my 4000T once its fully functional. For noww keeping the 3640 so ram options are limited. Is the 128mb ZorRam faster that the onboard motherboard fast ram?

It's slower than the onboard RAM.

[Trev]

But will he notice a difference in day to day use? Many have suggested using ZorRAM cards as swap space, which sounds like a great idea to me. The best way to implement that, however.... When the card was first announced, there were several requests for the inclusion of a configurable storage controller (or ROM-based file system) of some sort in addition to the option to autoconfigure as RAM.

[omnicron10]

The memory will have a lower priority than the motherboard memory.  So motherboard memory should be used first by the OS and then the Zorroram.  

Although slower than MB ram and other Fast Slot connected memory it is still faster than MMU mapped swap ram like Gigaram did.  Using it as swap space does not make much sense but using it as a RAM disk does.  

If you were going to use it as just SWAP space why not just use it as RAM?  I am sure the use of the programs to move data in and out of the ZorRam like swap memory should be even slower.

You could try making a RAM or RAD drive that only uses the ZorRam address space (not sure you can even do that) and then using Gigaram or other software that did virtual memory on OS3 to save the swap to the ZorRam RAM/RAD device.  Might be very interesting but I doubt it would be faster than just using it normally.  

You would need a MMU for the Gigaram style solution.

[Trev]

If you were going to use it as just SWAP space why not just use it as RAM? I am sure the use of the programs to move data in and out of the ZorRam like swap memory should be even slower.

Good point. :-) With Exec's default memory management, though, I'm sure there's a point where one may outperform the other.

[Zac67]

Depends on what's actually running. Any code exceeding the cache will get some - permanent - speed penalty in slower RAM. Swapping from slower to faster RAM would make that code ran faster - at the expense of having to copying the memory tiles before execution, so only heavily used code will profit from that (the larger the speed difference, the bigger the trade-off for swapping may be).
Also a bit academic since virtual memory doesn't work so well on AmigaOS.

[mousehouse]

That almost starts to sound like Upper Memory in the old DOS days ;-) Although DOS used segmented / paged memory if I remember correctly...

Didn't DOS Upper Memory work using some XMS handler that copied pages of memory to the DOS usable address space?

[Zac67]

Didn't DOS Upper Memory work using some XMS handler that copied pages of memory to the DOS usable address space?

That was EMS (usually done in hardware) - more memory than could be addressed was paged into segment pages. Older systems used bank switching, 386+ the MMU.

XMS is just extended memory management without (much) help to get to it.