On the Usage of GPUs for Efficient Motion Estimation in Medical Image Sequences

Table 1

Details of systems used for evaluation.

Parameters	System 1	System 2	System 3
System name	C1060	GTX480	C2070
Host CPU	Xeon 5110 (Harpertown)	Intel Core i7	Xeon 5650 (Gulftown)
Host CPU speed	1.6 GHz	2.8 GHz	2.67 GHz
Host OS	Ubuntu 10.10 (64 bit)	Ubuntu 10.10 (32 bit)	Ubuntu 10.10 (64 bit)
Kernel	2.6.35	2.6.31	2.6.35
Host RAM	2 GB	4 GB	24 GB
Host L1-cache	64 KB	64 KB	64 KB
Host L2-cache	4 MB	8 MB	12 MB
GPU series	C1060	GTX480	C2070
Compute capability	1.3	2.0	2.0
Device memory	1 GB	4 GB	6 GB
Multiprocessors	24	16	14
Cores per MP	8	16	32
Total cores	192	512	498
GPU L1-cache	16 KB	64 KB	64 KB
(Shared memory)
GPU L2-cache	N/A	128 KB	128 KB
CUDA version	4.0	3.3	3.2
Compiler flags	-O3 –arch = sm_13	-O3 –arch = sm_2.0	-O3 –arch = sm_2.0