Skip to content

Latest commit

 

History

History
50 lines (49 loc) · 1.35 KB

README.md

File metadata and controls

50 lines (49 loc) · 1.35 KB

Speed comparisons between Nvidia's cuPQC & Intel CPUs for Post Quantum & Classical Cryptography.

Specifications:

GPU:

GPU used: Nvidia's a4000X
CUDA version: 12.6
OS: Ubuntu 22.04 LTS (GNU/Linux 5.15.0)
Cryptographic Library: cuPQC, release 1

Intel CPU #0:

Processor: 12th Gen Intel(R) Core(TM) i9-12900
Architecture: x86_64
Cores : 24
OS: Ubuntu 22.04 LTS (GNU/Linux 5.15.0)
CPU Flags (used): AVX2, SSE2, RDSEED/RDRAND, SHA_NI, AES_N1, BMI2, POPCNT
Cryptographic Library: OpenSSL (3.4.0) + libOQS
LibOQS Version: v0.12.0-rc1
ML-KEM : FIPS-203
ML-DSA: FIPS-204

Results

For cuPQC:

  • MLKEM-512:
Key Generation Throughput: ~3170219.98 ops/sec
Encapsulation Throughput: ~3268917.41 ops/sec
Decapsulation Throughput: ~3144369.38 ops/sec
  • MLKEM-768:
Key Generation Throughput: ~2421352.50 ops/sec
Encapsulation Throughput: ~2475257.25 ops/sec
Decapsulation Throughput: ~1599282.74 ops/sec
  • MLDSA-44:
Key Generation Throughput: 2145190.54 ops/sec
Sign Throughput: 252759.62 ops/sec
Verification Throughput: 1152434.40 ops/sec
  • MLDSA-65:
Key Generation Throughput: 1324213.74 ops/sec
Sign Throughput: 202258.85 ops/sec
Verification Throughput: 819874.50 ops/sec

For OpenSSL

Results can be found in the OpenSSL dir