Testing by Bilibili channel Homolab reveals that its high-capacity QLC NAND can plummet to a sluggish 1.1 MB/s, trailing behind TLC-based variants due to physical storage density trade-offs.
The Architectural Compromise of QLC NAND in Flagship Handsets
When Apple slotted QLC (quad-level cell) NAND into the top-tier 1TB iPhone 18 Pro Max, engineers prioritized raw storage density over sustained input/output operations. While TLC (triple-level cell) architecture crams three bits of data into each memory cell, QLC forces four bits into that exact same physical footprint.
Homolab’s teardown and benchmark suite laid bare the real-world penalty of this architecture. When stacked against a 512GB TLC-based iPhone 18 Pro in a low-queue-depth mixed workload, the QLC model lagged by 38%, scoring 8,168 compared to the TLC variant’s 11,285. Even under heavier workloads where the performance delta squeezed down, the TLC storage model retained a definitive 12% advantage.
Dissecting the Triple-Tier Buffer Collapse
The iPhone 18 Pro Max implements a sophisticated pipeline, but heavy operations will quickly overwhelm it.
Initially, the device leans on a high-speed SLC (single-level cell) cache, driving write speeds up to a blistering 3,000 MB/s. Once that primary cache saturates, a secondary TLC-based buffer steps in, averaging a respectable 578 MB/s. But push the system past that safety net, and the hardware has no choice but to write directly to the underlying QLC NAND. At that exact juncture, sustained write performance craters to 79.4 MB/s, with raw telemetry capturing drops as low as 25.6 MB/s.
At 60% capacity, the available SLC cache shrinks drastically from 250GB down to a meager 58GB. The secondary buffer throttles back to 396 MB/s, while direct QLC writes plummet to an average of 45 MB/s—occasionally stalling out at a paltry 1.1 MB/s.
Real-World Impact Versus Synthetic Benchmarks
Context matters. These failure modes are unearthed by punishing synthetic workloads designed to stress the storage controller to its absolute limits.
Instead, the bottleneck rears its head exclusively during storage-heavy edge cases. Content creators capturing high-bitrate 4K or ProRes video files, moving massive uncompressed asset libraries, or executing a full local data backup restore will stare at progress bars for considerably longer than they would on a base-tier TLC model. Apple’s storage engineering trade-off is clear: buy the 1TB configuration for maximum capacity, but expect a stark performance tax when the pipeline runs dry.
