Problem
An internal NVMe drive fills its SLC cache during a large multicam render, causing sustained write speed to collapse after an initially fast benchmark.
Solution
Root Cause / Diagnostic:
Consumer TLC/QLC NVMe drives allocate a dynamic pseudo-SLC cache (often 10–15% of free space) to accelerate short write bursts. During continuous multicam rendering exceeding the SLC buffer size (e.g., a 200GB continuous export), the controller must simultaneously write new data and fold old SLC data into TLC blocks, collapsing write speeds to 250MB/s.
Actionable Fix:
1. Media Drive Segregation: Relocate render cache and scratch disk allocations to a dedicated high-endurance NVMe drive with high native direct-to-TLC write speeds (e.g., Samsung 990 Pro or WD Black SN850X).
2. Capacity Buffer Maintenance: Ensure at least 30% of the NVMe drive remains completely unallocated to maximize the size of the dynamic SLC cache buffer.
3. Sustained Write Profiling: Benchmark the drive with a continuous 150GB sequential write test using AJA System Test to determine the exact sustained post-cache write floor.
Pro Tip:
For long video rendering, disregard peak PCIe 4.0/5.0 burst speeds; look specifically at the drive's post-SLC-exhaustion sustained TLC write speed.
Consumer TLC/QLC NVMe drives allocate a dynamic pseudo-SLC cache (often 10–15% of free space) to accelerate short write bursts. During continuous multicam rendering exceeding the SLC buffer size (e.g., a 200GB continuous export), the controller must simultaneously write new data and fold old SLC data into TLC blocks, collapsing write speeds to 250MB/s.
Actionable Fix:
1. Media Drive Segregation: Relocate render cache and scratch disk allocations to a dedicated high-endurance NVMe drive with high native direct-to-TLC write speeds (e.g., Samsung 990 Pro or WD Black SN850X).
2. Capacity Buffer Maintenance: Ensure at least 30% of the NVMe drive remains completely unallocated to maximize the size of the dynamic SLC cache buffer.
3. Sustained Write Profiling: Benchmark the drive with a continuous 150GB sequential write test using AJA System Test to determine the exact sustained post-cache write floor.
Pro Tip:
For long video rendering, disregard peak PCIe 4.0/5.0 burst speeds; look specifically at the drive's post-SLC-exhaustion sustained TLC write speed.