Best Memory Interface for editing?

i can't find better answer on "memory interface" video editing.......

for example have gtx 580 384bit memory interface

and there gtx 680 256bit memory interface more ram , cuda cores dual gpu ppro doesn't use 2 gpu's isn't it, if iam not mistaken 680 dual gpu......

need upgrade 680, dont need more displays or more ram or cuda cores iam happy gtx 580's?

so gtx 680 may boost profomance, or diffrence in rendering/exporting/playback?

or cards less memory interface or more ram 4gigs enough editing?

 

thanks alot in advance....

dual gpu cards use single gpu pr. 590/690 used 50%. 580/680 single gpu cards.

 

video card performance

with number of observations available (980+) see clear tendency more cuda cores performance, @ least older architecture. usual performance increase of hardware acceleration on software around 12 x, cards fewer cuda cores lagging behind , ones more cores pulling ahead. increase in performance on software shows indeed 9800 gt , gtx 260 lagging when compared gtx 480/580/680. newer architecture shows benefits.

 

on other hand, introduction of kepler cards around 3 times number of cuda cores on fermi cards, people expected triple performance , disappointed shown wrong in expectations.

 

with results have our ppbm5 benchmark sure looks memory bandwidth may decisive factor impact performance. no longer number of cuda cores. time tell.

 

video card

the touted maximus solution, @ least adobe, utter waste of money, because requires expensive quadro card plus tesla c2075 card, slower simple 2 generations old gtx 470, simple reason lacks cuda cores , memory bandwidth make faster. have several maximus solutions in current ppbm5 benchmark , despite prices € 6000 quadro 6000 plus tesla c2075, these solutions outperformed many gtx 470/480/570/580/670 , 680 cards, cost fraction. thing impressive maximus solution price. reason opt maximus solution, despite cost, if absolutely need 10 bit output expensive 10 bit monitors, , quality, not performance.

 

what determines video card performance pr?

based on information available, not number of shaders or cuda cores. if case, kepler cards leave fermi , older cards in dust 3 times number of cuda cores, not case. determining factor may memory bandwidth , explains why kepler range faster fermi , why kepler faster quadro cards.

 

do need upgrade 680, dont need more displays or more ram or cuda cores iam happy gtx 580's?

no.

More discussions in Hardware Forum

adobe