Discrepancy between Benchmark numbers and actual numbers

[XakEp]

Ok, so I've been doing a LOT of testing and have a mask attack at this speed now -

Speed.#1.........: 40525.5 MH/s (8.24ms) @ Accel:64 Loops:512 Thr:256 Vec:4
Speed.#2.........: 29474.6 MH/s (6.84ms) @ Accel:128 Loops:256 Thr:256 Vec:4
Speed.#3.........: 16398.9 MH/s (5.99ms) @ Accel:256 Loops:128 Thr:256 Vec:4
Speed.#4.........: 17355.7 MH/s (9.46ms) @ Accel:128 Loops:256 Thr:256 Vec:4
Speed.#*.........: 103.8 GH/s

The 2080RTX is very inconsistent. It speeds up and slows down and I can't tell why. I think at this point I need to start figuring out efficient ways to get GPU acceleration with rule based attacks. Nothing in the FAQ is getting me near the benchmarks. For example, I'm seeing this on NTLM benchmarks -

Hashmode: 1000 - NTLM

Speed.#1.........: 73142.2 MH/s (9.74ms) @ Accel:64 Loops:1024 Thr:256 Vec:2
Speed.#2.........: 57691.4 MH/s (15.03ms) @ Accel:128 Loops:1024 Thr:256 Vec:2
Speed.#3.........: 33680.1 MH/s (28.19ms) @ Accel:256 Loops:1024 Thr:256 Vec:2
Speed.#4.........: 30712.6 MH/s (23.18ms) @ Accel:128 Loops:1024 Thr:256 Vec:2
Speed.#*.........: 195.2 GH/s

But actual attacks are 60% slower than that. So more experimentation will be happening, maybe someone else can learn from my missteps. If anyone has suggestions for me I'm all ears, but I'll be plugging away at this.

Thanks for listening to this old fart. I'm not smart, I'm just too dumb to give up.