Looks OK but not better than the Q6600:(
Its better but and its a big BUT
Its requires software re-compiling to take advantage of the extra gains using AMD compilers there are a few things leading up to 128bit AMD chips.
I think one can say AMD would have a long way to go to get all software designers to create AMD optimised code though…
Each core can run at a different frequency and power setting and this time around AMD have published an equivlant of NVidia’s overclocking utility for it.
I’d doubt I would jump on that bandwaggon just yet but its a good chip from things i’ve seen certainly a datacenter must have for the power concerned folks.
However I’d like a spilt design chip where at least 1 core can run at silly 4ghz speed and the others per normal.
Take all current bechmarks with a pinch of salt 
I really thought, per clock, they would be much closer. So much for my cheap farming plans.
Would have been nice to see some per clock comparisons with X2s, even if they have to limit the number of cores. That would help put things in context of improvements.
Where I think they still have a potential advantage is in cases where all 4 cores need to talk to each other, although how many home apps do you know that need that?
Individual core clocking may be interesting to help squeeze out that bit more total power where having the same clock on the cores is not important. But this will have to wait until they get the clocks ramped up and power down. So I guess like many of their other generational changes, it might take a stepping or two before it’s worth looking at.
Not sure how many people realise is but certainly do not take Sandra Memory benchmarks as been right because it not.
Sandra does not support multi-threads yet on the K10 unlike K8 where on a dual you will see 2 threads running at the mo only 1 thread is seen running.
the K10 has 2 memory modes Ganged 1x128bit dual channel & Un-ganged 2x64 bit dual channel for multi-threaded enabled benchmarking it needs to be in un-ganged mode.
At the mo as said take everything with a pinch of salt things far far too early to make true decisions.
If you search hard enough you’ll probably find them, but for Barcelona vs Opteron Rev F.
It’s a shame this hasn’t transferred well to the desktop environment - server side the game is much closer and Barcelona has an enormous lead in virtualisation where 4 Opteron Rev F’s were only just beaten down by 4 quad core Xeons and and the virtualisation benchmarks show way in excess of 100% improvement moving to Barcelona.
Still, I’m waiting to see what happens in the New Year when all the cards are finally on the table…