Here's the conclusion since some people here CLEARLY didn't read the article. Here is Brent's conclusion. I suggest reading it entirely. It covers all the incorrect statements that were mentioned by several posters in this thread.
If your card is not behaving like it is supposed to then there's something wrong with a PBCK or the card itself. Proven already with both the DCU and Windforce units having sub par cooling and/or defective press samples. This is the very same as getting one of those crappy custom Palit cards back from the Fermi era and shouting that the whole Fermi lineup was crap, overclocking, temp and noise wise.
If you go to any 290(X) Tri-X review with OC performance and temps figures you will see that all you're saying here is bogus. If you slap a good cooler on these cards they will perform as they're supposed to showing the scaling you'd expect with OC and helping you to figure out where the problem is with your OC. PCPer settled with a 1225 Mhz OC for the 290 Tri-X and it started to throttle while being power constrained. Anyone settling with a OC lower than being limited by the power throttle threshold doesn't know how these cards work. It's the only thing you can't overcome by now and it's only alleviated by lower temps that obviously can't be achieved with coolers like the DCU or the Windforce.
Every time anyone publishes something that fits your agenda it's the new holy grail and everything else is WRONNGGG. Every time you find something to bash these cards with you have the urge to let everyone know to be shot down sooner or later.
You're just looking too hard for something against these cards and that's worrisome.