V100's AI Training Capabilities
The V100's capabilities in AI training have been well-proven. It serves as the computational core of the Summit supercomputer and was among the first GPUs used by OpenAI for large model training. GPT-3, which marked a new breakthrough in artificial intelligence, was trained using V100 clusters. Although it has been on the market for 8 years and many V100s are beginning to retire, it remains a powerful GPU with substantial computing power.
Password Cracking Scenario Testing
As we all know, password cracking requires brute force methods that demand significant computational power. How does the V100 perform in this scenario? Hashcat is widely recognized as the most efficient software for hash decryption. We tested V100's hash computing (password cracking) capabilities by running Hashcat benchmark on a server equipped with four V100 GPUs. This capability has been leveraged by MyPassRecovery in their online service, building a powerful computational infrastructure. This is where MyPassRecovery has an advantage over other password recovery software.
Test Environment
- GPU: 4 x Tesla V100-SXM2-16GB
- Memory: 16GB per card
- CUDA Version: 12.8
- Hashcat Version: v6.2.5
- Test Mode: Benchmark with optimized kernel
Benchmark Test Results
System Information
hashcat (v6.2.5) starting in benchmark mode
Benchmarking uses hand-optimized kernel code by default.
You can use it in your cracking session by setting the -O option.
Note: Using optimized kernel code limits the maximum supported password length.
To disable the optimized kernel code in benchmark mode, use the -w option.
* 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.
* Device #3: WARNING! Kernel exec timeout is not disabled.
* Device #4: WARNING! Kernel exec timeout is not disabled.
CUDA API (CUDA 12.8)
====================
* Device #1: Tesla V100-SXM2-16GB, 15830/16144 MB, 80MCU
* Device #2: Tesla V100-SXM2-16GB, 15830/16144 MB, 80MCU
* Device #3: Tesla V100-SXM2-16GB, 15830/16144 MB, 80MCU
* Device #4: Tesla V100-SXM2-16GB, 15830/16144 MB, 80MCU
Hash Algorithm Performance Results
MD5 (Hash-Mode 0)
Device #1: 55972.1 MH/s (47.17ms) @ Accel:256 Loops:1024 Thr:128 Vec:8
Device #2: 55967.9 MH/s (47.15ms) @ Accel:256 Loops:1024 Thr:128 Vec:8
Device #3: 55971.4 MH/s (47.15ms) @ Accel:256 Loops:1024 Thr:128 Vec:8
Device #4: 55775.6 MH/s (47.31ms) @ Accel:256 Loops:1024 Thr:128 Vec:8
Total Speed: 223.7 GH/s
SHA1 (Hash-Mode 100)
Device #1: 17705.3 MH/s (75.08ms) @ Accel:32 Loops:1024 Thr:512 Vec:1
Device #2: 17715.0 MH/s (75.06ms) @ Accel:32 Loops:1024 Thr:512 Vec:1
Device #3: 17687.4 MH/s (75.18ms) @ Accel:32 Loops:1024 Thr:512 Vec:1
Device #4: 17629.5 MH/s (75.42ms) @ Accel:32 Loops:1024 Thr:512 Vec:1
Total Speed: 70737.2 MH/s
SHA2-256 (Hash-Mode 1400)
Device #1: 7651.7 MH/s (86.93ms) @ Accel:32 Loops:512 Thr:512 Vec:1
Device #2: 7657.8 MH/s (86.90ms) @ Accel:32 Loops:512 Thr:512 Vec:1
Device #3: 7651.0 MH/s (86.95ms) @ Accel:32 Loops:512 Thr:512 Vec:1
Device #4: 7623.7 MH/s (87.27ms) @ Accel:32 Loops:512 Thr:512 Vec:1
Total Speed: 30584.3 MH/s
SHA2-512 (Hash-Mode 1700)
Device #1: 2405.6 MH/s (69.38ms) @ Accel:4 Loops:1024 Thr:512 Vec:1
Device #2: 2413.7 MH/s (69.15ms) @ Accel:4 Loops:1024 Thr:512 Vec:1
Device #3: 2413.7 MH/s (69.15ms) @ Accel:4 Loops:1024 Thr:512 Vec:1
Device #4: 2406.0 MH/s (69.37ms) @ Accel:4 Loops:1024 Thr:512 Vec:1
Total Speed: 9639.0 MH/s
NTLM (Hash-Mode 1000)
Device #1: 100.8 GH/s (25.64ms) @ Accel:128 Loops:1024 Thr:256 Vec:8
Device #2: 100.8 GH/s (25.66ms) @ Accel:128 Loops:1024 Thr:256 Vec:8
Device #3: 100.7 GH/s (25.64ms) @ Accel:128 Loops:1024 Thr:256 Vec:8
Device #4: 100.2 GH/s (25.78ms) @ Accel:128 Loops:1024 Thr:256 Vec:8
Total Speed: 402.5 GH/s
LM (Hash-Mode 3000)
Device #1: 49293.1 MH/s (26.70ms) @ Accel:128 Loops:1024 Thr:128 Vec:1
Device #2: 49349.2 MH/s (26.69ms) @ Accel:128 Loops:1024 Thr:128 Vec:1
Device #3: 49308.1 MH/s (26.70ms) @ Accel:128 Loops:1024 Thr:128 Vec:1
Device #4: 49151.6 MH/s (26.78ms) @ Accel:128 Loops:1024 Thr:128 Vec:1
Total Speed: 197.1 GH/s
Windows Authentication Performance
NetNTLMv1 / NetNTLMv1+ESS (Hash-Mode 5500)
Device #1: 54278.5 MH/s (48.05ms) @ Accel:512 Loops:512 Thr:128 Vec:2
Device #2: 54274.7 MH/s (48.06ms) @ Accel:512 Loops:512 Thr:128 Vec:2
Device #3: 54248.8 MH/s (48.05ms) @ Accel:512 Loops:512 Thr:128 Vec:2
Device #4: 54160.5 MH/s (48.14ms) @ Accel:512 Loops:512 Thr:128 Vec:2
Total Speed: 217.0 GH/s
NetNTLMv2 (Hash-Mode 5600)
Device #1: 3897.2 MH/s (85.69ms) @ Accel:16 Loops:1024 Thr:256 Vec:1
Device #2: 3897.9 MH/s (85.65ms) @ Accel:16 Loops:1024 Thr:256 Vec:1
Device #3: 3897.5 MH/s (85.67ms) @ Accel:16 Loops:1024 Thr:256 Vec:1
Device #4: 3888.0 MH/s (85.87ms) @ Accel:16 Loops:1024 Thr:256 Vec:1
Total Speed: 15580.5 MH/s
WiFi Security Performance
WPA-PBKDF2-PMKID+EAPOL (Hash-Mode 22000) [Iterations: 4095]
Device #1: 880.4 kH/s (91.30ms) @ Accel:16 Loops:1024 Thr:256 Vec:1
Device #2: 879.4 kH/s (91.17ms) @ Accel:16 Loops:1024 Thr:256 Vec:1
Device #3: 881.6 kH/s (91.19ms) @ Accel:16 Loops:1024 Thr:256 Vec:1
Device #4: 879.5 kH/s (91.40ms) @ Accel:16 Loops:1024 Thr:256 Vec:1
Total Speed: 3520.8 kH/s
Unix/Linux System Hashes
descrypt, DES (Unix), Traditional DES (Hash-Mode 1500)
Device #1: 1918.4 MH/s (87.00ms) @ Accel:16 Loops:1024 Thr:128 Vec:1
Device #2: 1922.3 MH/s (86.82ms) @ Accel:16 Loops:1024 Thr:128 Vec:1
Device #3: 1920.2 MH/s (86.93ms) @ Accel:16 Loops:1024 Thr:128 Vec:1
Device #4: 1914.7 MH/s (87.17ms) @ Accel:16 Loops:1024 Thr:128 Vec:1
Total Speed: 7675.5 MH/s
md5crypt, MD5 (Unix), Cisco-IOS $1$ (Hash-Mode 500) [Iterations: 1000]
Device #1: 17589.9 kH/s (52.58ms) @ Accel:128 Loops:250 Thr:512 Vec:1
Device #2: 17584.5 kH/s (52.57ms) @ Accel:128 Loops:250 Thr:512 Vec:1
Device #3: 17416.0 kH/s (52.57ms) @ Accel:128 Loops:250 Thr:512 Vec:1
Device #4: 17477.6 kH/s (52.94ms) @ Accel:128 Loops:250 Thr:512 Vec:1
Total Speed: 70068.0 kH/s
bcrypt $2*$, Blowfish (Unix) (Hash-Mode 3200) [Iterations: 32]
Device #1: 62762 H/s (94.89ms) @ Accel:256 Loops:32 Thr:24 Vec:1
Device #2: 62791 H/s (94.88ms) @ Accel:256 Loops:32 Thr:24 Vec:1
Device #3: 62482 H/s (95.30ms) @ Accel:256 Loops:32 Thr:24 Vec:1
Device #4: 62535 H/s (95.23ms) @ Accel:256 Loops:32 Thr:24 Vec:1
Total Speed: 250.6 kH/s
sha512crypt $6$, SHA512 (Unix) (Hash-Mode 1800) [Iterations: 5000]
Device #1: 383.4 kH/s (67.12ms) @ Accel:8192 Loops:256 Thr:64 Vec:1
Device #2: 384.1 kH/s (66.97ms) @ Accel:8192 Loops:256 Thr:64 Vec:1
Device #3: 379.5 kH/s (67.78ms) @ Accel:8192 Loops:256 Thr:64 Vec:1
Device #4: 384.0 kH/s (67.01ms) @ Accel:8192 Loops:256 Thr:64 Vec:1
Total Speed: 1531.0 kH/s
Kerberos Authentication
Kerberos 5, etype 23, AS-REQ Pre-Auth (Hash-Mode 7500)
Device #1: 1104.5 MH/s (75.60ms) @ Accel:32 Loops:1024 Thr:32 Vec:1
Device #2: 1105.7 MH/s (75.54ms) @ Accel:32 Loops:1024 Thr:32 Vec:1
Device #3: 1104.9 MH/s (75.57ms) @ Accel:32 Loops:1024 Thr:32 Vec:1
Device #4: 1101.8 MH/s (75.78ms) @ Accel:32 Loops:1024 Thr:32 Vec:1
Total Speed: 4417.0 MH/s
Kerberos 5, etype 23, TGS-REP (Hash-Mode 13100)
Device #1: 1142.9 MH/s (72.92ms) @ Accel:256 Loops:128 Thr:32 Vec:1
Device #2: 1142.6 MH/s (72.95ms) @ Accel:256 Loops:128 Thr:32 Vec:1
Device #3: 1143.0 MH/s (72.91ms) @ Accel:256 Loops:128 Thr:32 Vec:1
Device #4: 1140.8 MH/s (73.03ms) @ Accel:256 Loops:128 Thr:32 Vec:1
Total Speed: 4569.2 MH/s
File Encryption Formats
7-Zip (Hash-Mode 11600) [Iterations: 16384]
Device #1: 889.2 kH/s (82.10ms) @ Accel:32 Loops:4096 Thr:128 Vec:1
Device #2: 892.8 kH/s (82.18ms) @ Accel:32 Loops:4096 Thr:128 Vec:1
Device #3: 852.7 kH/s (82.12ms) @ Accel:32 Loops:4096 Thr:128 Vec:1
Device #4: 867.6 kH/s (82.30ms) @ Accel:32 Loops:4096 Thr:128 Vec:1
Total Speed: 3502.3 kH/s
RAR3-hp (Hash-Mode 12500) [Iterations: 262144]
Device #1: 88508 H/s (39.94ms) @ Accel:4 Loops:16384 Thr:512 Vec:1
Device #2: 88711 H/s (39.93ms) @ Accel:4 Loops:16384 Thr:512 Vec:1
Device #3: 88750 H/s (39.91ms) @ Accel:4 Loops:16384 Thr:512 Vec:1
Device #4: 87961 H/s (40.07ms) @ Accel:4 Loops:16384 Thr:512 Vec:1
Total Speed: 353.9 kH/s
RAR5 (Hash-Mode 13000) [Iterations: 32799]
Device #1: 92663 H/s (54.66ms) @ Accel:16 Loops:128 Thr:1024 Vec:1
Device #2: 92666 H/s (54.66ms) @ Accel:16 Loops:128 Thr:1024 Vec:1
Device #3: 92665 H/s (54.65ms) @ Accel:16 Loops:128 Thr:1024 Vec:1
Device #4: 92376 H/s (54.78ms) @ Accel:16 Loops:128 Thr:1024 Vec:1
Total Speed: 370.4 kH/s
KeePass 1 (AES/Twofish) and KeePass 2 (AES) (Hash-Mode 13400) [Iterations: 24569]
Device #1: 112.3 kH/s (60.44ms) @ Accel:4 Loops:1024 Thr:512 Vec:1
Device #2: 112.3 kH/s (60.42ms) @ Accel:4 Loops:1024 Thr:512 Vec:1
Device #3: 112.3 kH/s (60.43ms) @ Accel:4 Loops:1024 Thr:512 Vec:1
Device #4: 111.9 kH/s (60.65ms) @ Accel:4 Loops:1024 Thr:512 Vec:1
Total Speed: 448.8 kH/s
TrueCrypt RIPEMD160 + XTS 512 bit (Hash-Mode 6211) [Iterations: 1999]
Device #1: 665.1 kH/s (55.66ms) @ Accel:64 Loops:64 Thr:256 Vec:1
Device #2: 668.4 kH/s (55.44ms) @ Accel:64 Loops:64 Thr:256 Vec:1
Device #3: 669.5 kH/s (55.44ms) @ Accel:64 Loops:64 Thr:256 Vec:1
Device #4: 668.0 kH/s (55.59ms) @ Accel:64 Loops:64 Thr:256 Vec:1
Total Speed: 2671.1 kH/s
Windows DPAPI
DPAPI masterkey file v1 (Hash-Mode 15300) [Iterations: 23999]
Device #1: 148.8 kH/s (93.43ms) @ Accel:16 Loops:512 Thr:512 Vec:1
Device #2: 148.8 kH/s (93.43ms) @ Accel:16 Loops:512 Thr:512 Vec:1
Device #3: 148.8 kH/s (93.43ms) @ Accel:16 Loops:512 Thr:512 Vec:1
Device #4: 148.3 kH/s (93.77ms) @ Accel:16 Loops:512 Thr:512 Vec:1
Total Speed: 594.8 kH/s
DPAPI masterkey file v2 (Hash-Mode 15900) [Iterations: 12899]
Device #1: 80312 H/s (77.98ms) @ Accel:4 Loops:512 Thr:512 Vec:1
Device #2: 80320 H/s (77.97ms) @ Accel:4 Loops:512 Thr:512 Vec:1
Device #3: 80327 H/s (77.98ms) @ Accel:4 Loops:512 Thr:512 Vec:1
Device #4: 79932 H/s (78.21ms) @ Accel:4 Loops:512 Thr:512 Vec:1
Total Speed: 320.9 kH/s
Other Encryption Systems
macOS v10.8+ (PBKDF2-SHA512) (Hash-Mode 7100) [Iterations: 1023]
Device #1: 988.1 kH/s (78.50ms) @ Accel:64 Loops:31 Thr:512 Vec:1
Device #2: 989.8 kH/s (78.46ms) @ Accel:64 Loops:31 Thr:512 Vec:1
Device #3: 989.2 kH/s (78.49ms) @ Accel:64 Loops:31 Thr:512 Vec:1
Device #4: 986.0 kH/s (78.71ms) @ Accel:64 Loops:31 Thr:512 Vec:1
Total Speed: 3953.1 kH/s
LastPass + LastPass sniffed (Hash-Mode 6800) [Iterations: 499]
Device #1: 5442.8 kH/s (42.88ms) @ Accel:16 Loops:124 Thr:1024 Vec:1
Device #2: 5442.3 kH/s (42.88ms) @ Accel:16 Loops:124 Thr:1024 Vec:1
Device #3: 5403.4 kH/s (42.91ms) @ Accel:16 Loops:124 Thr:1024 Vec:1
Device #4: 5420.1 kH/s (43.05ms) @ Accel:16 Loops:124 Thr:1024 Vec:1
Total Speed: 21708.6 kH/s
Bitcoin/Litecoin wallet.dat (Hash-Mode 11300) [Iterations: 200459]
Device #1: 8990 H/s (74.21ms) @ Accel:1024 Loops:1024 Thr:128 Vec:1
Device #2: 8946 H/s (74.59ms) @ Accel:1024 Loops:1024 Thr:128 Vec:1
Device #3: 8835 H/s (75.52ms) @ Accel:1024 Loops:1024 Thr:128 Vec:1
Device #4: 9011 H/s (74.09ms) @ Accel:1024 Loops:1024 Thr:128 Vec:1
Total Speed: 35782 H/s
Test Runtime Information
Started: Sun Aug 31 02:37:44 2025
Stopped: Sun Aug 31 02:47:47 2025
Total Runtime: 10 minutes 3 seconds
Performance Summary
Below are the performance data tables compiled from the benchmark results:
Primary Hash Algorithm Performance
| Hash Algorithm |
Single Card Speed |
Four Cards Total |
| MD5 (Mode 0) |
~56 GH/s |
223.7 GH/s |
| SHA1 (Mode 100) |
~17.7 GH/s |
70.7 GH/s |
| SHA2-256 (Mode 1400) |
~7.65 GH/s |
30.6 GH/s |
| SHA2-512 (Mode 1700) |
~2.41 GH/s |
9.64 GH/s |
| NTLM (Mode 1000) |
~100.8 GH/s |
402.5 GH/s |
| LM (Mode 3000) |
~49.3 GH/s |
197.1 GH/s |
Windows Authentication Hash Performance
| Hash Type |
Four Cards Total |
| NetNTLMv1 (Mode 5500) |
217.0 GH/s |
| NetNTLMv2 (Mode 5600) |
15.6 GH/s |
WiFi Password Cracking Performance
| Protocol Type |
Four Cards Total |
Iterations |
| WPA-PBKDF2-PMKID+EAPOL (Mode 22000) |
3,520.8 kH/s |
4095 |
Unix/Linux Password Hash Performance
| Hash Type |
Four Cards Total |
Iterations |
| descrypt (Mode 1500) |
7,675.5 MH/s |
- |
| md5crypt (Mode 500) |
70,068 kH/s |
1000 |
| bcrypt (Mode 3200) |
250.6 kH/s |
32 |
| sha512crypt (Mode 1800) |
1,531 kH/s |
5000 |
Kerberos Authentication Performance
| Protocol Type |
Four Cards Total |
| Kerberos 5 AS-REQ (Mode 7500) |
4,417 MH/s |
| Kerberos 5 TGS-REP (Mode 13100) |
4,569 MH/s |
File Encryption Format Performance
| File Format |
Four Cards Total |
Iterations |
| 7-Zip (Mode 11600) |
3,502.3 kH/s |
16384 |
| RAR3-hp (Mode 12500) |
353.9 kH/s |
262144 |
| RAR5 (Mode 13000) |
370.4 kH/s |
32799 |
| KeePass (Mode 13400) |
448.8 kH/s |
24569 |
| TrueCrypt RIPEMD160 (Mode 6211) |
2,671.1 kH/s |
1999 |
Windows DPAPI Performance
| Encryption Type |
Four Cards Total |
Iterations |
| DPAPI v1 (Mode 15300) |
594.8 kH/s |
23999 |
| DPAPI v2 (Mode 15900) |
320.9 kH/s |
12899 |
Other Encryption Formats
| Encryption Type |
Four Cards Total |
Iterations |
| macOS v10.8+ (Mode 7100) |
3,953.1 kH/s |
1023 |
| LastPass (Mode 6800) |
21,708.6 kH/s |
499 |
| Bitcoin wallet.dat (Mode 11300) |
35,782 H/s |
200459 |
Runtime Status Monitoring
The test ran for 10 minutes with moderate GPU load and good temperature control. Monitoring via nvidia-smi showed:
- Moderate GPU utilization
- Well-controlled temperatures
- Stable power consumption
Performance Analysis
The test results reveal several key insights:
-
Excellent Basic Hash Algorithm Performance: The V100 demonstrates outstanding performance on basic hash algorithms like MD5, SHA, and NTLM, with NTLM reaching an impressive 402.5 GH/s.
-
Reasonable Iterative Hash Performance: For algorithms requiring extensive iterations like bcrypt and scrypt, performance meets expectations. These algorithms are specifically designed to resist brute force attacks.
-
Cost-Performance Advantage: The V100's performance is comparable to the RTX 2080 Ti, but with better pricing in the second-hand market. For scenarios requiring large-scale password recovery or security testing, the V100 offers excellent value for money.
Conclusion
Despite being 8 years old, the V100 still demonstrates powerful computing capabilities in password cracking scenarios. For enterprise-level password recovery services, security auditing, and penetration testing applications, the V100 remains a worthy consideration. Particularly in the second-hand market, the V100's cost-performance advantage becomes even more apparent, providing an economical solution for building high-performance password cracking clusters.