[TOC]
# MT7915E Performance Comparison: Mesh vs. VLAN Mode

The MediaTek MT7915E chipset demonstrates dramatically different performance characteristics when operating in Mesh Point mode versus AP-UP mode. This comparison analyzes these differences based on empirical test data.

## Performance Metrics

| Metric | MT7915E Mesh Mode | MT7915E AP-UP Mode | Factor Difference |
|--------|-------------------|-------------------|-------------------|
| **Throughput** | 215-250 Mbps | 33-34 Mbps | ~7× higher in Mesh |
| **Packet Drops** | 0 (0%) | ~6,000 (2.3%) | Infinitely better in Mesh |
| **Retransmissions** | 0 | 0 | Same (excellent) |
| **PHY Rate (RX)** | 573.5 Mbps (HE-MCS 11) | 573.5 Mbps (HE-MCS 11) | Same |
| **PHY Rate (TX)** | 390-541.6 Mbps (HE-MCS 8-11) | 270-300 Mbps (HE-MCS 13-15) | ~1.5× higher in Mesh |
| **NSS (Spatial Streams)** | 2×2 | 2×2 | Same |
| **Channel Bandwidth** | 40 MHz | 40 MHz | Same |
| **Signal Strength** | -46 to -47 dBm | -44 to -46 dBm | Similar |
| **CPU Interrupts** | ~146,000 (after 10 iterations) | ~620,000 (after 10 iterations) | ~4× fewer in Mesh |

## Technical Implementation Differences

### 3. Data Path Efficiency

#### Mesh Mode:
- Data tx duration: 126-175 million μs for test period
- Data rx duration: 3-4 million μs for test period
- Bandwidth utilization ratio: ~40:1 (tx:rx) - efficient asymmetric use
- Target offset (Toffset): 4.3 billion μs - very stable timing

#### VLAN Mode:
- Higher tx duration relative to data transferred
- Less efficient bandwidth utilization ratio
- Higher processing latency
- Higher access contention
- Larger management frame overhead

### 4. Hardware Resource Utilization

#### Mesh Mode:
- ~146,000 interrupts on CPU0 after 10 iterations
- ERR count: 1,893 (relatively low)
- Efficient CPU resource utilization

#### VLAN Mode:
- ~620,000 interrupts on CPU0 after 10 iterations
- ERR count: 3,722 (higher than Mesh)
- ~4× higher CPU utilization

## Technical Root Causes for Performance Difference

2. **Frame Encapsulation Overhead**:
- VLAN tagging adds 4 bytes per frame (sometimes 8 bytes for QinQ)
- Additional frame header processing
- Tag Control Information (TCI) processing
- Priority Code Point (PCP) verification
- VLAN Identifier (VID) validation
- Driver's buffer management differs between modes (The 2.3% packet drop rate (rx_drop_misc) indicates buffer management problems)