bcrypt spped extremely decreasing during first minute

[blaster666]

While running 3200 mode attack, I meet following issue:

The progress starts:

hashcat.exe -m 3200 -w 4 -a 0 example0.hash ignis-10M.txt
hashcat (v6.2.6) starting

* Device #1: WARNING! Kernel exec timeout is not disabled.
            This may cause "CL_OUT_OF_RESOURCES" or related errors.
            To disable the timeout, see: https://hashcat.net/q/timeoutpatch
* Device #2: WARNING! Kernel exec timeout is not disabled.
            This may cause "CL_OUT_OF_RESOURCES" or related errors.
            To disable the timeout, see: https://hashcat.net/q/timeoutpatch
CUDA API (CUDA 12.2)
====================
* Device #1: NVIDIA GeForce RTX 4090, 23008/24563 MB, 128MCU

OpenCL API (OpenCL 3.0 CUDA 12.2.135) - Platform #1 [NVIDIA Corporation]
========================================================================
* Device #2: NVIDIA GeForce RTX 4090, skipped

Minimum password length supported by kernel: 0
Maximum password length supported by kernel: 72
Hashes: 325 digests; 286 unique digests, 286 unique salts
Bitmaps: 16 bits, 65536 entries, 0x0000ffff mask, 262144 bytes, 5/13 rotates
Rules: 1

Optimizers applied:
* Zero-Byte

Watchdog: Temperature abort trigger set to 90c

INFO: Removed 9 hashes found as potfile entries.

Host memory required for this attack: 843 MB

Dictionary cache built:
* Filename..: ignis-10M.txt
* Passwords.: 10000000
* Bytes.....: 94461698
* Keyspace..: 10000000
* Runtime...: 1 sec

For the first minute I have such a speed:

Session..........: hashcat
Status...........: Running
Hash.Mode........: 3200 (bcrypt $2*$, Blowfish (Unix))
Hash.Target......: example0.hash
Time.Started.....: Sat Aug 19 01:25:14 2023 (27 secs)
Time.Estimated...: Sat Aug 19 13:42:13 2023 (12 hours, 16 mins)
Kernel.Feature...: Pure Kernel
Guess.Base.......: File (ignis-10M.txt)
Guess.Queue......: 1/1 (100.00%)
Speed.#1.........:    62648 H/s (387.18ms) @ Accel:16 Loops:64 Thr:24 Vec:1
Recovered........: 9/286 (3.15%) Digests (total), 0/286 (0.00%) Digests (new), 9/286 (3.15%) Salts
Progress.........: 1474560/2860000000 (0.05%)
Rejected.........: 0/1474560 (0.00%)
Restore.Point....: 0/10000000 (0.00%)
Restore.Sub.#1...: Salt:30 Amplifier:0-1 Iteration:768-832
Candidate.Engine.: Device Generator
Candidates.#1....: 123456 -> 24091986
Hardware.Mon.#1..: Temp: 47c Fan: 31% Util:100% Core:2820MHz Mem:10251MHz Bus:16

After like 30 seconds, speed decreases in few times, and progress string like freezes for seconds on " 1474560/2860000000"

And the next status log is :

Session..........: hashcat
Status...........: Running
Hash.Mode........: 3200 (bcrypt $2*$, Blowfish (Unix))
Hash.Target......: example0.hash
Time.Started.....: Sat Aug 19 01:25:14 2023 (34 secs)
Time.Estimated...: Sat Aug 19 16:32:20 2023 (15 hours, 6 mins)
Kernel.Feature...: Pure Kernel
Guess.Base.......: File (ignis-10M.txt)
Guess.Queue......: 1/1 (100.00%)
Speed.#1.........:    50899 H/s (387.21ms) @ Accel:16 Loops:64 Thr:24 Vec:1
Recovered........: 9/286 (3.15%) Digests (total), 0/286 (0.00%) Digests (new), 9/286 (3.15%) Salts
Progress.........: 1523712/2860000000 (0.05%)
Rejected.........: 0/1523712 (0.00%)
Restore.Point....: 0/10000000 (0.00%)
Restore.Sub.#1...: Salt:31 Amplifier:0-1 Iteration:896-960
Candidate.Engine.: Device Generator
Candidates.#1....: 123456 -> 24091986
Hardware.Mon.#1..: Temp: 47c Fan: 31% Util:100% Core:2820MHz Mem:10251MHz Bus:16

And after this

Session..........: hashcat
Status...........: Running
Hash.Mode........: 3200 (bcrypt $2*$, Blowfish (Unix))
Hash.Target......: example0.hash
Time.Started.....: Sat Aug 19 01:25:14 2023 (1 min, 2 secs)
Time.Estimated...: Sun Aug 20 04:19:50 2023 (1 day, 2 hours)
Kernel.Feature...: Pure Kernel
Guess.Base.......: File (ignis-10M.txt)
Guess.Queue......: 1/1 (100.00%)
Speed.#1.........:    28594 H/s (388.22ms) @ Accel:16 Loops:64 Thr:24 Vec:1
Recovered........: 9/286 (3.15%) Digests (total), 0/286 (0.00%) Digests (new), 9/286 (3.15%) Salts
Progress.........: 1769472/2860000000 (0.06%)
Rejected.........: 0/1769472 (0.00%)
Restore.Point....: 0/10000000 (0.00%)
Restore.Sub.#1...: Salt:36 Amplifier:0-1 Iteration:320-384
Candidate.Engine.: Device Generator
Candidates.#1....: 123456 -> 24091986
Hardware.Mon.#1..: Temp: 47c Fan: 31% Util:100% Core:2820MHz Mem:10251MHz Bus:16

You could see progress is continuing, but not as fast as on first 30 sec~

And in few minutes ,I have this speed:

Session..........: hashcat
Status...........: Running
Hash.Mode........: 3200 (bcrypt $2*$, Blowfish (Unix))
Hash.Target......: example0.hash
Time.Started.....: Sat Aug 19 01:25:14 2023 (6 mins, 20 secs)
Time.Estimated...: Mon Aug 21 22:17:36 2023 (2 days, 20 hours)
Kernel.Feature...: Pure Kernel
Guess.Base.......: File (ignis-10M.txt)
Guess.Queue......: 1/1 (100.00%)
Speed.#1.........:    11172 H/s (390.30ms) @ Accel:16 Loops:64 Thr:24 Vec:1
Recovered........: 9/286 (3.15%) Digests (total), 0/286 (0.00%) Digests (new), 9/286 (3.15%) Salts
Progress.........: 4276224/2860000000 (0.15%)
Rejected.........: 0/4276224 (0.00%)
Restore.Point....: 0/10000000 (0.00%)
Restore.Sub.#1...: Salt:87 Amplifier:0-1 Iteration:960-1024
Candidate.Engine.: Device Generator
Candidates.#1....: 123456 -> 24091986
Hardware.Mon.#1..: Temp: 48c Fan: 31% Util:100% Core:2820MHz Mem:10251MHz Bus:16

Is it how hashcat naturally working with this algorithm , or I have some troubles to be fixed to have constant 60kH\s speed. 
I would appreciate your assistance so much, as I am testing new setup , and curios a lot

[blaster666]

One parameter to the story I've posted above:
Wordlist of 300 hashes, consists of : 75% $10$, 25% $7$, so maybe it shows speed for $7$ at first , but they are not at the start of the txt file.

[Chick3nman]

It's what you thought, the low cost bcrypt are being tried first and the average speed is falling as the 10 cost spend more time being processed.