It's all Maxwell's fault
Nvidia is about to fail us again. Not only has the 700 series been pushed back to May because TSMC is still a joke, but they are no longer going to use the GK110 Big Kepler chip in the GTX 780. Instead it'll be the GK114, the "Ti" version, just like they put in the 680 (GK104). They don't want to use the GK110 because they worry the performance increase would be so much it would make the release of 20nm Maxwell in 2014 not seem quite as special. Instead of a nice 50% increase, we only get ~15%, which we can already get by overclocking. Some GPUs can be bought with a 20% OC out of the box. Nvidia is also screwing over the professional GPU buyers. They were promised a 3x DP increase in the K20 using GK110 over the K10. But it turned out they could only do ~65% because of TSMC's continued wafer imperfections.
Of course AMD will also be late with their 8000 series since they also buy their wafers from TSMC. And they are also promising only a 15% performance increase. The difference is they haven't been waving the GK110 in front of us for the past 6 months+ only to let us down. Besides, AMD has the best DP for us anyway. By far.
BTW: AMD is laying off 1800-2400 ppl either tomorrow or next week.
Reply With Quote
Originally Posted by zombie67
