oclHashcat v1.36 slow on Windows 10

[michelin]

No luck here. Benchmarks still crash.

cudaHashcat-1.37>cudaHashcat64.exe -m 5600 -b
cudaHashcat v1.37 starting in benchmark-mode...

Device #1: GeForce GTX 980 Ti, 6144MB, 1076Mhz, 22MCU
Device #2: GeForce GTX 980 Ti, 6144MB, 1076Mhz, 22MCU

Hashtype: NetNTLMv2
Workload: 1024 loops, 256 accel

ERROR: cuStreamSynchronize() 719

If I run -a 3 crack it does run but nowhere near the speed I would expect (570 MH/s max per card) on a single hash.