Originally Posted by criminal
Wait. You complain about the price of the Titan and you own a GTX690 which is clearly overpriced as well. Considering what the Titan offers and the performance it has with one gpu, it is worth more than the GTX690 IMHO. At least the Titan can compute and you NEVER have SLI issues to worry with.
No competition. This has been discussed before. The 256-bit bus and the memory configurations available for the GTX670/GTX680 show that the card was at best a midrange level card, until it performed the way it did. Nvidia instantly decided to sell it as the flagship cards since GK110 was not ready.
The Titan is at least 60-80% (sometimes over 100%) faster than the GTX580. Nvidia probably new that and decided to go the route they did. No way a generation jump has ever offered that big of a performance jump.
The HD5870 was a significant jump above the HD4870
though i don't know how it compares to 580 < Titan.... probably still abit short i guess, then again, there is the price
so lets not go patting Nvidia on the back just yet But i'm not gonna get into that today....
I have a question for someone alot more knowledgeable in GPUs than i am, what happened with the Cuda core scaling? I mean ever since the 8800 days it seems Nvidias GPU power has scaled in an almost linear fashion when it came to adding more cores even during die shrinks. For example, going from a GTX285 (240cores) 648Mhz
to a GTX480 (480cores) 700Mhz
netted almost twice the performance.
Yet going from a GTX580 (512
cores) 772mhz to a Titan (2688
) 837mhz does not net 5 times the performance, at best Titan is about as powerful as 3 GTX580s yet it has 5 times as many cores.
is the ROP and TMU count to blame?
did kepler dilute the performance of the cores and compensate by clocking them higher and adding more of them? Like i said
this stuff is way to complicated for me so if someone could give the the "for dummies" guide version i'd be grateful Edited by th3illusiveman - 4/5/13 at 4:51am