> > Unless there is something special / wrong with sdhci-sprd (unlikely? Its just a flag) or your eMMC (maybe more likely?) then Id expect this or similar performance degradation for any host controller and eMMC. > I can redo some measurement if you provide your workload. > But I'd say if you don’t want to pay the price of reliable write then make sure to not issue them, by not issuing FUA? > Maybe I'm misunderstanding, but why would the host controller driver control FUA? > Maybe one more point on that: Compare performance with fua disabled but cache off, should be comparable unless there is an actual reliable write problem. If reliable write is performing horribly on that eMMC maybe a quirk could be thought of. Anyway your 13MB/s random with fua, if a reliable write is basically always active for a given cache size, doesn't sound totally unreasonable (over 3000 IOPS, assuming 4k writes). Hyperstone GmbH | Reichenaustr. 39a | 78467 Konstanz Managing Director: Dr. Jan Peter Berns. Commercial register of local courts: Freiburg HRB381782
