Concorde Rules said:That mem B/W is certainly a killer. I have a feeling ATI knew and released them just ahead of the competition so they didn't need to use 800+mhz mem speeds. Which 1.1ns will do.
Wonder what this GDDD-4 will clock to 2.2ghz? 2.4?![]()
I see 16k easily, 18k with water and *possibly* 19/20k when people get these on dry ice with 1000mhz core and 1.3ghz mem clocks etc.
I think the X19x0s best core:mem ratio is something like 5.5:8.
Starting at 500/600, 2005 is about 9.5k
500/700 its 10k yet 600/600 is still around 9.7/8k
500/800 is pushing 10.5k and 600/800 is pushing 11k
Complex beasts the X19x0s are![]()
I have noticed this.
Upping the memory clock by a tiny amount say 50mhz gives a huge boost in synthetic benhcmarks at least, whereas upping the core although increases the score it doesn't by the same degree as an increased memory clock.
I believe this is all down to how efficient the ring bus memory controller is. However it is only efficient when its coupled with lightning fast memory (could explain why the X1800XL wasn't as quick as most people had hoped)
