http://www.tomshardware.com/news/intel-haswell-leaked-clock-speeds,19719.html

I'll wait till  the 14nm broadwell to upgrade but this is good for those who haven't had a upgrade in ages I guess . Whats your thoughts ?

First off, there's always the possibility that it's a fake.  It's far from guaranteed that even Intel knows the final clock speeds six months away from release, as the bin splits depend on yields.

Next, those look awfully similar to the Ivy Bridge clock speeds.  That's a reasonable guess if you're making things up.  But if it's accurate, then it means what we've long suspected:  Haswell will be a little faster than Ivy Bridge, but not much.

Haswell is a new architecture.  The main advantages that it will bring as compared to Ivy Bridge are greatly reducing idle power consumption, offering somewhat less embarrassing graphics performance, and having an SoC version that can go in tablets without putting out so much heat as to defeat the purpose of a tablet.

For desktop, use, it will probably be a little bit faster than Ivy Bridge, but not a lot.  There will be no point in upgrading from Sandy Bridge or later to it, and no point in waiting for it until we're very close to launch (if it were launching tomorrow, I'd advise people to wait, but not 5 or 6 months out), but once it's already out, if you can get Haswell or Ivy Bridge for the same price, you'd probably prefer Haswell.

Only thing I'm realtively interested in Haswell for is overclock capability. Nahalem was very good, and Sandy was really good. Ivy was a disappointment for the most part, especially coming off the heals of two stellar performers. The stock clocks, or any information in this article really, don't tell much about potential overclockability.

Then again, neither do the typical pre-release leaks that we see (like the Bulldozer that was going at nearly 8.5Ghz that had everyone all excited).

But with a major emphasis on power, I hope to either see a much more aggressive Turbo Boost (preferred, because then it is more or less guaranteed by the manufacturer), or at least some nice overclock headroom. It's also possible that the new 3-D 22nm node process just doesn't overclock well at all (which is what they claimed was the reason with Ivy, and Haswell will be built on the same process). But we won't know much on this until we see some production parts start to leak out, which usually isn't until about a few weeks before release.

What I would really like to see would be a combination of AMD's PowerTune and nVidia's GPU Boost become standard in all these processing parts: PowerTune keeps you from exceeding a TDP ceiling, but does a poor job at utilizing available headroom to increase performance. Boost does a good job of using headroom, but doesn't cap your TDP to prevent going over the edge and crashing (or in extreme cases causing damage). Really, you need a combination of both to really optimize the process.

Intel Turbo Boost (and AMD Turbo Core) both were out first, but seem antiquated by now: They both basically just look at how many cores are in use, and then use a more or less static upclock if certain conditions are met. They don't really monitor TDP, and only downclock on excessive temperature (at which point you've already screwed the pooch, whereas if you downclock on TDP you may never exceed a given thermal threshold, because you've taken a proactive step to prevent hitting a thermal limit).

That's all just pipedreaming though, and the vast majority of people don't need that kind of technology in a CPU - or even if it were present probably wouldn't utilize it. But it makes sense for more than just Desktops.