Bsgs will never faster then kangaroo on big ranges.
Example puzzle #80, width 79bit
Kanagaroo need in average 2^40.5 op to solve key
2080ti can serach DPs with speed 1500Mkey/s = 2^30.48
So you need in average 2^10 second to find key.

Bsgscuda can search 2^61/s, so 2^79 / 2^61 = 2^18 second to check whole 79bit range. Ofcourse key can be close to the begining and you can found very fast but it is not 100%.
we can devide pub to 64 pubkey and try to find one of them in range 1..3ffffffffffffffffff fortunately, the key we need is the first one on the list but it was just lucky.
by the way privkey will be 3a869719b73046d6b46 = 2^73.87
so we need  2^73.87 / 2^61/s = 2^12.87 seconds to find key. It is 7 more timer then need for kangaroo