Benchmark Hashrate 4000 MH/s -> Live barely hitting 1000 MH/s

[enaske]

Hey guys,

so I tried to figure out why Benchmark is so much faster, but I can't get it.
So when I Benchmark -m 16500 (JWT)
I get a Rate of roughly 4000 MH/s with 4090.
When I'm running the "live" Attack, it barely hits 1000 MH/s
The prompt is:

hashcat token.txt -m 16500 -a 3 -w 4 --increment -1 charset.hcchr ?1?1?1?1?1?1?1

Charset is:
abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789_.-!@*$?&%

(small/big letters + Numbers + common Symbols)


When I get to 6-8 Chars It gets to:

Session..........: hashcat
Status...........: Running
Hash.Mode........: 16500 (JWT (JSON Web Token))
Hash.Target......: eyJhbGciOiJIUzI1NiIsInR5cCIgOiAiSldUIiwia2lkIiA6ICJ...X1gF7I
Time.Started.....: Fri Apr 07 18:01:53 2023 (44 secs)
Time.Estimated...: Fri Apr 07 20:49:35 2023 (2 hours, 46 mins)
Kernel.Feature...: Pure Kernel
Guess.Mask.......: ?1?1?1?1?1?1?1 [7]
Guess.Charset....: -1 charset.hcchr, -2 Undefined, -3 Undefined, -4 Undefined
Guess.Queue......: 7/15 (46.67%)
Speed.#1.........:  996.8 MH/s (268.42ms) @ Accel:4 Loops:1024 Thr:512 Vec:1
Recovered........: 0/1 (0.00%) Digests (total), 0/1 (0.00%) Digests (new)
Progress.........: 43754979328/10030613004288 (0.44%)
Rejected.........: 0/43754979328 (0.00%)
Restore.Point....: 0/26873856 (0.00%)
Restore.Sub.#1...: Salt:0 Amplifier:41728-41984 Iteration:0-256
Candidate.Engine.: Device Generator
Candidates.#1....: q17eran -> phsLeND
Hardware.Mon.#1..: Temp: 73c Fan: 73% Util:100% Core:2610MHz Mem:10501MHz Bus:16

Benchmark says:

Speed.#1.........:  4354.5 MH/s (61.50ms) @ Accel:4 Loops:1024 Thr:512 Vec:1

CUDA API (CUDA 12.1)
====================
* Device #1: NVIDIA GeForce RTX 4090, 23008/24563 MB, 128MCU

OpenCL API (OpenCL 3.0 CUDA 12.1.98) - Platform #1 [NVIDIA Corporation]
=======================================================================
* Device #2: NVIDIA GeForce RTX 4090, skipped

Minimum password length supported by kernel: 0
Maximum password length supported by kernel: 256

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

Optimizers applied:
* Zero-Byte
* Not-Iterated
* Single-Hash
* Single-Salt
* Brute-Force

Watchdog: Temperature abort trigger set to 90c

Host memory required for this attack: 1475 MB

[Chick3nman]

There are several underlying algorithms and your hash may have a slower one than the one used for the benchmark. You are also not using a mask that produces an "ideal workload" and will see some speed drop from that as well. This is perfectly normal.