GeForce GTX 970 memory allocation issue explained by Nvidia

After comments about the disappointment of GTX 960, I find this humorous. So which card does everyone want, the underachiever 960 or the underachiever 970? I'm still happy with my underachieving GTX 660.
 
After comments about the disappointment of GTX 960, I find this humorous. So which card does everyone want, the underachiever 960 or the underachiever 970?
Strangely enough, the 970 was lauded for it's performance, price, and efficiency at launch. Now the card is an underachiever because the specifications are amended but the performance, price, and efficiency remain unchanged.

I'm thinking this says more about the consumer than it does about the board.
 
The article at Anandtech is way more complex than I can grasp. I read someone mention that an easy way to understand it is "the gtx 970 is a 4Gb Vram card in which 3.5Gb is accessed with a 224-bit bandwidth and (the last) 0.5 Gb is accessed with a 32-bit bandwidth", is this correct?
 
"the gtx 970 is a 4Gb Vram card in which 3.5Gb is accessed with a 224-bit bandwidth and (the last) 0.5 Gb is accessed with a 32-bit bandwidth", is this correct?
If you divided 256 bits by 8, what would you have for each one individually? If only 7 are accessed at the same time, this minimizes the bit traffic from 256 to 224, which is divided equally among the 7. Each one is accessed at 32 bits, not just the 0.5GB segment.
 
I have 780tis superclocked in sli which I got just before these came out so I am glad I did, was peeved when these hit the shelves but with problems people are having, my EVGAs are awesome cards.
 
Back