I was in the same situation, I didn't have very good experiences with ATi in the past and really had my mind set on nVidia for my new build, but after reading the HD5870 reviews I couldn't resist. I'm happy to say I love my 5870 and it performs very well, exceeds my expectations and now I'm back in the ATi crowd. I'm using it with an Intel Core i7 930 CPU, there is absolutely no reason that an ATi card wouldn't support Intel CPU's, PCI Express is PCI Express and is supported across the manufacturer spectrum. The only issue would be multi-card support but Intel has designed their chipsets to support both ATi's CrossFire X and nVidia's SLi multi-GPU technologies. AMD on the other hand sometimes lacks SLi support, making Intel+nVidia a somewhat better setup than AMD+nVidia if you plan on using more than one card. Then again, nVidia did make their own chipsets for AMD CPU's to solve this problem.
As far as I can tell, the ATi HD58xx cards are better than the nVidia GTX4xx cards due to a more efficient power design and a better price point. The GTX480 does win as the fastest single GPU, but it is $200 more than the 5870, the second fastest single GPU. To be honest, the two are very close in performance but the 480 is slightly ahead, but the 5870 uses less power, makes less heat, costs less, and supports Eyefinity.
I didn't think Eyefinity would be a huge important thing, the demos looked neat but I finally decided to go for it and bought 3 21.5" 1080p monitors and it is nothing short of amazing.
The only real downside to ATi cards is that they don't have CUDA or Physx support. CUDA is nVidia's proprietary GPU general-purpose computing library that enables heavy duty processing applications to use the GPU as a highly parallelized processing device to run complex computations. It is commonly used for intense processing and has been incorporated into the distributed computing system Folding@Home which many users on this forum participate in. PhysX is a gaming technology originally developed by Ageia but later acquired by nVidia. It originally allowed a dedicated "physics processing unit" on an expansion card to take the physics calculation load off the CPU to improve performance in a select few games that incorporated PhysX technology, but nVidia modified it to use CUDA and now any recent nVidia GPU can act as a Physics processor.
ATi responded this generation with OpenCL and DirectCompute, both of which will also be incorporated into nVidia's products. This should level the playing field for most GPU computing as developers will most likely switch to a more standardized, cross-platform system like OpenCL. However, with nVidia still the owner of PhysX, it is unlikely we will see that supported on ATi hardware in the near future, even if it is technically capable.
