drkohler said:
You are not making sense on the memory. Clamshell or not, memory bus speed is memory bus speed. I highly doubt anyone is going for mixed memory. You are just asking for engineering trouble doing that. |
Making perfect sense. It is Gigabits per second.
And no. memory bus speed is not memory bus speed. Each chip has it's own bus... So think along those lines.
Memory transactions are a parallel task.
It is okay it can get confusing quickly.
Basically 16x memory chips.
But I'll try and dumb it down for you.
Each blue line is a 16-bit data path to the GPU... And 16x chips is an aggregate 256-bit bus.
Once you hit 1GB of used capacity on each chip, then you can only start filling up the 2GB chips... Which is when bandwidth is cut in half as memory transactions can only occur in half the chips. (The 2GB ones.)
We actually have had similar cases where this has happened on GPU's before... A few of nVidia GPU's infact... For example the Geforce 970 where the 3.5GB of Ram ran at full speed and the last 512MB ran at a reduced bandwidth.
So no... Memory bus isn't just memory bus.
As for mixed memory... There have been tons of systems with mixed memory setups. - Playstation 3 (XDR and GDDR3), Playstation 4 (GDDR5 and DDR3) for example, so to assume it's unheard of is a bit of a fallacy.
Either way, I am not asserting anything as factual at this stage anyway.
--::{PC Gaming Master Race}::--