is CUDA SDK Toolkit really required?
I installed brand new Windows 11 and tried to run Hashact without installing any additional drivers.

And it looks like it is running faster than my old Windows on the same hardware. This time I don't have CUDA SDK Toolint installed, so I wonder if this is necessary or not. How could I increase my performance for brute force MD5?

hashcat (v6.2.6) starting in restore mode

Successfully initialized the NVIDIA main driver CUDA runtime library.

Failed to initialize NVIDIA RTC library.

* Device #1: CUDA SDK Toolkit not installed or incorrectly installed.
            CUDA SDK Toolkit required for proper device support and utilization.
            Falling back to OpenCL runtime.

nvmlDeviceGetFanSpeed(): Not Supported

OpenCL API (OpenCL 3.0 CUDA 11.6.134) - Platform #1 [NVIDIA Corporation]
* Device #1: NVIDIA GeForce RTX 2070, 7488/8191 MB (2047 MB allocatable), 36MCU

Minimum password length supported by kernel: 0
Maximum password length supported by kernel: 31

Hashes: 1 digests; 1 unique digests, 1 unique salts
Bitmaps: 16 bits, 65536 entries, 0x0000ffff mask, 262144 bytes, 5/13 rotates

Optimizers applied:
* Optimized-Kernel
* Early-Skip
* Not-Iterated
* Single-Hash
* Single-Salt
* Brute-Force

Watchdog: Temperature abort trigger set to 90c

Host memory required for this attack: 1471 MB

[s]tatus [p]ause [b]ypass [c]heckpoint [f]inish [q]uit =>

Session..........: hashcat
Status...........: Running
Hash.Mode........: 22500 (MultiBit Classic .key (MD5))
Time.Started.....: Mon Nov 13 08:10:50 2023 (3 secs)
Time.Estimated...: Sun Nov 26 22:44:41 2023 (13 days, 14 hours)
Kernel.Feature...: Optimized Kernel
Guess.Mask.......: ?1?1?1?1?1?1?1?1?1?1?1?1?1 [13]
Guess.Charset....: -1 C:\Users\pawsa\Downloads\has\list.hcchr, -2 Undefined, -3 Undefined, -4 Undefined
Guess.Queue......: 4/11 (36.36%)
Speed.#1.........:  518.3 MH/s (64.95ms) @ Accel:4 Loops:1024 Thr:256 Vec:1
Recovered........: 0/1 (0.00%) Digests (total), 0/1 (0.00%) Digests (new)
Progress.........: 184344473272320/793714773254144 (23.23%)
Rejected.........: 0/184344473272320 (0.00%)
Restore.Point....: 67180916736/289254654976 (23.23%)
Restore.Sub.#1...: Salt:0 Amplifier:1024-2048 Iteration:0-1024
Candidate.Engine.: Device Generator
Candidates.#1....: ap0w1d0rrr8ai -> 0a8Ki-aaip8ai
Hardware.Mon.#1..: Temp: 44c Util:100% Core:1952MHz Mem:5495MHz Bus:16

[s]tatus [p]ause [b]ypass [c]heckpoint [f]inish [q]uit =>
without cuda sdk hashcat uses opencl for your GPU, i think it all depends on hashtarget and possible optimisations in the kernelcode ( if there is a special cuda function for a specific hash with no pendant in opencl, cuda should be faster )

for simple hashes like md5 i dont think there will be any difference
Not necessary, hence it not blocking you and is just a warning. Due to OpenCL locking Hashcat to 1/4 VRAM, it can be helpful to give Hashcat access to all of your VRAM, hence - `7488/8191 MB (2047 MB allocatable)` . There are very few kernels that use CUDA-specific code and is usually margin of error between the two platforms