SSD Write Endurance 25nm Vs 34nm

[hot120]

Why not download them directly to the storage drive ?

Faster to extract files while on the SSD, then move to the storage drives.

[Anvil]

34.68TB Host writes
MWI 80
no other changes

34_68TB_host_writes.PNG

[Ao1]

Ao1, can I ask what program you use to generate your charts?

It's an Excel Radar Chart - available in Office 2007 and above. Insert Chart>Radar

[kache]

43TB. 78%. I started the new software which includes some random writes 1TB ago.

Very interesting.
Wondering if any test like this will be done with the C300 SSDs, which are the ones that have the best price/performance ratio and are bought by a lot of people...

[hot120]

It's an Excel Radar Chart - available in Office 2007 and above. Insert Chart>Radar

Thanks!

[Anvil]

37.34TB Host writes
MWI 78 (just minutes ago, so down 2 from last report)

no other changes

37_34_tb_hostwrites.PNG

[Ao1]

Anvil are you now running the more onerous version of the app?

(had to guess a few of the stats for one_hertz below).

[Anvil]

Thanks for the updated chart!

No, not yet, was planning to start using it at 35TB (20GB*365*5) but as One_Hertz didn't start using it till > 40TB I've decided to go on a little longer.

I might make the switch tonight, will be moving it to another computer where it will be running as a spare drive from then on.

[One_Hertz]

49TB. 75%. The random writes aren't speeding anything up it seems.

[Anvil]

Are you using the default settings for random writes, 1000ms per loop?

How much random writes have been generated so far?

edit:
updated the chart in post #1, that 320 is really pulling away from the G2, impressive.

[One_Hertz]

~3GB so far I think. I am thinking of increasing the amount of random writes...

By the way, I got a write error in the log yesterday... Not sure if it was a fluke or not.

[Anvil]

OK, you could try increasing random writes to 3000ms or maybe 5000ms, that would'nt be too excessive imho.

Never seen the error, shouldn't really happen, it just means that WriteFile reported false for 1 write op.
I'm not logging the error # but if it happens again I might change to catch the error#.

[Anvil]

40.31TB Host writes
MWI 77
No other changes.

Still running off the same computer, will try moving it tonight.

[bulanula]

I really appreciate your efforts BUT if your experiments show the drives lasting for quite a long time under so much stress THEN why are only SLC drives recommended for servers and intense write I/O applications ?

Maybe this is not random enough with the file placement ? I think the SF drive somehow converts random files into sequential after some time but now that it is out of the race it really does not matter anymore etc.

[Computurd]

enterprise usage would be to remain under loads such as these, 24/7, for years

[Anvil]

About to move the drive.

Current status is
41.15TB
MWI 76

41_15_TB_hostwrites.PNG

[Anvil]

Looks like the move was successful.
It's sitting in my only AMD system, ASUS CHIV, an SB850 based MB.

As a non-OS drive it is writing a bit more per hour, right now it looks like it's increasing from ~2.6TB to ~3.0-3.1TB per day.
I've switched to using the version that produces random writes and I've set it to write random IO for 2500ms per loop.

updated the chart as well.

edit:
Increased the random IO part to 3500ms per loop, might have to give it some more, will be monitoring.

[sergiu]

Does somebody knows what should happen with the drive when wear level hits 100%? Will it refuse to work or it will work until all spare cells are dead?

[zalbard]

Does somebody knows what should happen with the drive when wear level hits 100%? Will it refuse to work or it will work until all spare cells are dead?

Should theoretically refuse to write, but still read data.

[johnw]

Does somebody knows what should happen with the drive when wear level hits 100%? Will it refuse to work or it will work until all spare cells are dead?

Do you mean the MWI (media wear indicator) SMART attribute? Which decreases to 1 (not increases to 100).

If so, that is just a guideline. The SSD will likely continue to work for some time after MWI hits 1.

[bulanula]

Does somebody knows what should happen with the drive when wear level hits 100%? Will it refuse to work or it will work until all spare cells are dead?

This. Nobody knows yet!

[Computurd]

o you mean the MWI (media wear indicator) SMART attribute? Which decreases to 1 (not increases to 100).

If so, that is just a guideline. The SSD will likely continue to work for some time after MWI hits 1.

I agree. The ssd will definitely not stop writing at that point. should be after some time after that threshold is reached of course, but the requirements for it to get to that level are intentionally low. they have given themselves tons of play (for sure with the intels)

[Anvil]

A quick report

12hours generated ~1.48TB meaning that it was a bit optimistic when it started, still it's looking like I'll get 2-300GB extra per day.
In that same 12 hour period it generated ~1.86GB of random 4K writes.
I'm going to let it run for 24h and then I'll decide what to do with random writes.

42.95TB Host writes
MWI 75

[One_Hertz]

56TB. 71. I've put the random writes loop length to 5000ms.

[Anvil]

Updated the chart (post #1)

43.81TB Host writes
MWI 75

~4GB of random writes in ~19.5hours, I might end up with 5000ms as well.