Skip to main content
. 2013 Oct 21;13:25. doi: 10.1186/1472-6807-13-25

Figure 3.

Figure 3

The speed-up obtained by using multi-threading to calculate 14,630 2D single-precision NCCs of size 4096×4096 on (a) a dual quad-core workstation (Dirac), and (b) a four-GPU server (Mbiserv). See Table 1 for hardware details.