That's one clock cycle or about 1nSec for 1000Mhz clock. CAS latency happens once every column address strobe. So you saved 1nsec assuming same clock rate between 8-8-8-24 and 7-7-7-24 after the initial start. It is a very small savings in actuality. Google's "blink of an eye' response is 500mSec, or 500 million times longer.
For CL=8 that's 8nSec. Twenty years ago CAC (column address access) was about 15nSec. So things haven't improved that much on initial access. But once the memory page gets rolling, you get subsequent data much faster, something like twenty times faster.