I was wondering since RAMBUS did demo the XDR2 memory that's damn fast. I mean it's so fast that can't imaging anything worth it on box specially when CPU won't go that fast, but raises a question in my mind on GPU. Since on XP system the RAM is used as 384MB of Graphical where as Vista uses 1/2 of RAM -> Non-Local dedicated Graphical usage = for example 8GB system would have the GPU Local memory + 4096 RAM.
Correction: would have '4096MB RAM - Dedicated Video Memory'
http://www.rambus.com/us/news/press_...09/090623.html
Makes me wonder is there anything to syncronize there to get better forward to GDDR3/5? Or does this go like 'HDD > RAM > CPU > GPU' since on that it would use the L2 caches first. Anyway, if it's not going that way that would mean that matching GPU memory speed on 1:1 or 1:2 would increate the over all cache forwarding.
Anyone with any deeper ideas how this works? (Since I was just being logical on above text.)
Correction: would have '4096MB RAM - Dedicated Video Memory'
http://www.rambus.com/us/news/press_...09/090623.html
Makes me wonder is there anything to syncronize there to get better forward to GDDR3/5? Or does this go like 'HDD > RAM > CPU > GPU' since on that it would use the L2 caches first. Anyway, if it's not going that way that would mean that matching GPU memory speed on 1:1 or 1:2 would increate the over all cache forwarding.
Anyone with any deeper ideas how this works? (Since I was just being logical on above text.)