Performance Benchmark

Compare architectures, measure inference speed, and optimize for deployment

Benchmark Configuration

Model Comparison

Model Params Latency (ms) Throughput Memory (MB) Status
CNN-Small 1.2M 2.3 435 img/s 45 Complete
CNN-CBAM 2.5M 4.1 244 img/s 92 Complete
ResNet-18 11M 5.8 172 img/s 156 Running
ViT-Tiny 5.7M - - - Pending
Hybrid CNN-ViT 8.3M - - - Pending
60% Complete

Hardware Info

Device: NVIDIA RTX 3090
CUDA Cores: 10496
Memory: 24 GB
Driver: 515.65.01
4.2ms
Avg Latency
6.8ms
P99 Latency
87%
GPU Utilization
285W
Power Draw

Performance Analysis