They can use it. Unlike the GK110 disabling of double precision for all SKUs except for TITAN (a single GK110 SMX has: 192 single‐precision CUDA cores, 64 double‐precision units, 32 special function units (SFU), and 32 load/store units (LD/ST))...
The main reason is because GCN uses two 32-bit to combine to be 64-bit (double precision) , Nvidia's Kepler uses separate double precision units for 64-bit calcs.
From GCN whitepaper: "VGPRs – Every work-item has access to some number of VGPRs, up to a
maximum of 256. VGPRs are 32-bits wide and are used by the vector ALU and
vector memory systems. Double-precision operations use two adjacent VGPRs
to form a 64-bit value." http://developer.amd.com/wordpress/media/2012/12/AMD_Southern_Islands_Instruction_Set_Architecture.pdf
VGPR = Vector General-Purpose Registers
"Double precision and 32-bit integer instructions run at a reduced rate within a SIMD. The GCN Architecture is flexible and double precision performance varies
from 1/2 to 1/16 of single precision performance, increasing the latency accordingly. The double precision and 32-bit integer performance can be configured for
a specific GCN implementation, based on the target application." www.amd.com/us/Documents/GCN_Architecture_whitepaper.pdf
Just look at Milkyway@Home & Folding@Home.
R9 270 scored 2.8 on F@h explicit double precision , in line with the W7000
Another little known fact is Radeons possess 10-bit color. They just are software locked by Radeon drivers , it's not even a hardware limitation.
P.S. , OP look at Vegas render time : there's approximately ~13% difference between a Pitcairn (Hd7870/r9 270/Firepro W7000) and a Tahiti Pro (Hd7950) / Tahiti XT (HD7970GE / HD7970 / R9-280X / Firepro W9000)
http://www.anandtech.com/show/7503/the-amd-radeon-r9-270x-270-review-feat-asus-his/15Edited by AlphaC - 11/18/13 at 4:58pm