On Saturday, September 22, 2018 at 12:06:04 PM UTC-4, Marcus Linsner wrote: > Actually I have to correct the record here. Basically I was wrong about the > cause of this 2MB/sec write speed being caused by that new Samsung firmware > update: > > The fact is, I don't know when that rewriting/refreshing takes place(it could > be even more often than I thought, if judging by the SSD temperature being > usually 44-47 Celsius, and only sometimes 32-35 Celsius)! > What I was seeing as 2MB/sec was in fact a quirk of my Lenovo Ideapad Z575 > laptop whereby after one(and any subsequent) Windows 7 Restart (ie. > Start->Restart) the random write speed would be capped(I don't know why) to > that 2MB/sec and only a Sleep or Shutdown would bring it back to normal! And > this only happens when the SSD is connected via a drive caddy in place of the > optical disk drive(ODD), alone with nothing else connected (on ESATA, or main > drive bay).
That seems to point to the bios (re-?)initializing the drive into PIO (polled IO) mode instead of DMA mode. Does the BIOS have the sata ports set to AHCI mode? Is the BIOS up to date? Is the drive tray from a 3rd party, not Lenovo? > The disk life being decreased due to refreshing is just my guess based on my > assumption(possibly wrong?) that there's no way to refresh without decreasing > cell lifespan. The other one with 1-2minutes of freeze is still true, it was > happening even when Samsung SSD was on main drive bay alone(they were just > rarer) Cell lifespan really is no longer an issue for the vast majority of use cases. Mid- to top-tier consumer SSDs (as well as professional/datacenter SSDs) generally have several layers of mitigations in place to deal with signal loss and cell wear at the flash physical layer: - Physical storage over-provisioning that is not exposed at the logical level (primarily data-integrity focused). - Data redundancy schemes, even to the point of data protection when losing multiple silicon layers in the current stacked chip designs. Effectively: matrixed, RAID-like schemes at the slice-of-silicon level. - Active monitoring and management of the health of blocks of cells as well as "bad block" management layers below the logical data layer exposed through the sata or pcie ports. The amount of insight/signal on this that is exposed via S.M.A.R.T. varies by manufacturer. - Logical storage over-provisioning that is modifiable by the end-user (primarily performance focused). - Performance-oriented data storage optimization to keep the pool of available data blocks high. Some of the data management/redundancy features have strong parallels to approaches used in enterprise-class SANs from the 2000s/2010s...and have now been shrunk down into credit-card sized boards to ensure your baby pictures don't self-immolate. There are some published ... adverpapers? ... from the manufacturers on these with just enough information to be intriguing, but not enough to easily clone, (plus there are the piles of patents behind them). Brendan -- You received this message because you are subscribed to the Google Groups "qubes-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/qubes-devel/e1b20d4f-fa5d-4be1-a469-04db573a5859%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
