Introduction
Lanner collaborates with Intel to build the next-generation network security appliances powered by Intel Xeon 6 processors with Performance cores (P-cores) and Efficiency cores (E-cores). These hybrid architectures provide distinct advantages for different workloads, particularly for next-generation firewall (NGFW) applications. This whitepaper explores the performance and efficiency trade-offs between P-cores and E-cores when deploying NGFW on the Lanner NCA-6250 network security appliance.
Difference Between Intel Xeon 6 Processors P-cores and E-cores
Intel Xeon 6 processors feature Performance cores (P-cores) and Efficiency cores (E-cores), each optimized for specific workloads:
P-cores (Performance cores):
- Designed for maximum performance per core
- Ideal for multi-threaded and data-parallel workloads
- Feature a wider and deeper microarchitecture with larger caches, higher bandwidth, and wider vectors
- Excel in AI, ML, analytics, and workloads requiring data acceleration and vectorized code execution
E-cores (Efficiency cores):
- Prioritize performance per watt and high core density
- Single-threaded, optimized for throughput-oriented workloads
- Scale efficiently to handle high-density computing environments
- Provide better integer performance per watt and per socket
Source: Intel
Both core types share the same architecture and software ecosystem, enabling scalable optimizations across diverse applications in cloud, edge, and enterprise computing.
Testing NGFW on Lanner NCA-6250 with P-core and E-core
Lanner collaborated with Intel to run Intel’s NGFW reference implementation, included within the NetSec reference software from Intel, on the NCA-6250 network security appliance. This reference implementation includes essential security features such as firewall and intrusion prevention.
Source: Intel
Test Results
The test results comparing the Lanner NCA-6250 using Intel® Xeon® 6780E processor (E-core) and Intel® Xeon® 6787P processor (P-core) are summarized in the table below:
Metric |
Intel® Xeon® 6780E |
Intel® Xeon® 6787P |
Total No. of Cores |
144 |
86 |
Base Freq (GHz) |
2.3 |
2.0 |
TDP (W) |
330 |
350 |
Cache Size (MB) L3 |
108 |
336 |
Core Type |
E-core |
P-core |
Instructions supported |
Intel®AVX2 |
Intel® AVX-512 |
Intel® Hyper Threading Technology |
No |
Yes |
Throughput (Gbps) |
165 |
179 |
Throughput per Snort Core (Gbps) |
1.29 |
2.32 |
Throughput per Core (Gbps) |
1.18 |
2.13 |
Throughput per Thread (Gbps) |
1.18 |
1.07 |
Power (W) |
244 |
294 |
Perf/Watt |
0.68 |
0.61 |
No. of VPP Cores |
12c12t |
7C14T |
No. of Snort Cores |
128c128t |
77c154T |
Analysis
The results highlight the performance difference between P-cores and E-cores in NGFW applications:
Efficiency-Oriented (E-core) Configuration (Intel Xeon 6780E processor):
- Features a higher number of VPP cores and Snort cores.
- Delivers strong performance per watt, making it ideal for power-constrained environments.
- Handles lower throughput (165 Gbps) but provides better energy efficiency.
Performance-Oriented (P-core) Configuration (Intel Xeon 6787P processor):
- Features fewer VPP cores and Snort cores but achieves higher overall throughput (179 Gbps).
- Provides better performance per core and per socket.
- Leverages Intel AVX-512 instructions and hyperthreading to optimize compute intensive workloads.
Conclusion
Choosing between P-cores and E-cores for next-gen firewall applications on the Lanner NCA-6250 depends on the specific requirements of the deployment:
- For power-efficient, high-density computing with balanced throughput, E-cores offer better performance per watt.
- For maximum performance, higher throughput, and computationally demanding workloads, P-cores provide superior per-core and per-socket performance and performance density.