keep in mind that if you plan on crunching MW@H, stick w/ a 79xx, 69xx, or 58xx series GPU. the double precision (FP64) performance of a 78xx series GPUis a dismal 1/12 of its single precision (FP32) performance...whereas the FP64 performance of a 79xx or 69xx series GPU is a much more palatable 1/4 of its FP32 performance, and the FP64 performance of a 58xx series GPU is marginally less efficient at 1/5 of its FP32 performance.
even though the new and improved GCN architecture of the 78xx series GPU is much more efficient than the VLIW4 and VLIW5 architectures of the dated 69xx and 58xx series GPUs, the fact that the 78xx seires GPU's FP64 performance is so crippled means it'll trail even the 69xx and 58xx series GPUs in anything that requires double precision calculations, such as MW@H. if you have a look at this
ATI GPU comparison, and focus specifically on section 5.12-5.14 (
Evergreen,
Northern Islands, and
Southern Islands), you can see how the double precision performance of these GPUs all stack up. just to give you one brief comparison, an HD 5870 is approx. 3.4 times as fast/efficient as an HD 7870 with respect to FP64 calculations. heck, even the lowly HD 5830 is 2.25 times faster than the HD 7870 in this regard...