cuda and double-precision floating-point arithmetics

Question

this is what i found: " Cores perform only single-precision floating-point arithmetics. There is 1 double-precision floating-point unit. "

is this true for all compute capabilities (versions) ?

http://stackoverflow.com/questions/2816992/double-precision-floating-point-in-cuda — Roger Dahl, Mar 26 '12 at 17:21
in NVIDIA CUDA C Programming Guide 4.1 section F.4.1 p.144 is written "32 CUDA cores for integer and floating-point arithmetic operations". by "floating-point" they mean both single and double? — user1281071, Mar 26 '12 at 21:22

score 2 · Answer 1 · answered Mar 26 '12 at 17:33

2

Single and double precision floating point accuracy and performance has continuously evolved and is different for each of the compute capabilities.

answered Mar 26 '12 at 17:33

Roger Dahl

Start with Appendix F.1 to see if your GPU supports double. – Tom Mar 26 '12 at 17:49
right now i am looking for new GPU and i am mainly focused on double precision calculations. so i would like to know if there is GPU which has more then one double-precision floating-point unit per MP. – user1281071 Mar 26 '12 at 20:16

score 0 · Answer 2 · answered Mar 27 '12 at 14:17

0

section 5.4.1, table 5-1.

answered Mar 27 '12 at 14:17

user1281071

2 Answers2