Question Crucial MX500 500GB sata ssd Remaining Life decreasing fast despite few bytes being written

Page 6 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

Lucretia19

Great
Feb 5, 2020
101
5
95
2
Here's some log data regarding the beneficial effect of Power Cycling on ssd Write Amplification. The data strongly indicates that the more days have elapsed since the ssd was power-cycled off then on, the worse the amplification.

The obvious remedy -- a manual labor nuisance -- is to shutdown the computer more often. It's the reason why I chose to shutdown my pc last night, 21 days since the previous shutdown. (Although a nuisance, shutdowns may have additional benefits, since Windows tends to behave worse the longer it's been running. Windows 10 not so bad as Windows Vista, my previous OS.)

In recent weeks, beginning around the end of August, I've focused more on efforts to reduce the sum of SMART attributes F7 & F8 (247 & 248 in base 10) and stopped focusing on the write amplification F8/F7, since I believe the sum F7+F8 more directly affects ssd Remaining Lifetime. I reduced the rate of host pc writing to the ssd (which reduces the rate at which F7 increases) and this analysis doesn't attempt to control for that change. Of course, the ssd selftests regime has been running throughout to try to keep F8 in check.

The first of the two charts here doesn't separate F7 and F8; it's only about the sum F7+F8. Nevertheless, the beneficial effect of power cycling on amplification is suggested by looking at the rightmost two columns. The rightmost column is the increase of F7+F8 since the most recent power cycle, divided by the number of days since the most recent power cycle. There's a strong pattern: At each power cycle, the number in that column plummets, and then tends to increase each day until the next power cycle.

The log data is for the period from 6/13/2020 (Power Cycle Count reached 108 in the daily log) through today 10/21/2020 (Power Cycle Count reached 118 in the daily log).

After writing the above and pasting the first table (below), I realized I should have analyzed whether the effect on F7+F8 might be mainly an effect on F7, rather than on F8 amplification. I assume an effect on F7 would be due to Windows behaving worse the longer it's been running since a restart. (Note: Not all restarts involve power cycling.) So I created a second table that separates F7 and F8. I see no way to paste more data into the first table, so I pasted the second table below the first. It covers the same range of dates but doesn't include the Date column or some other columns of the first table. It repeats the F7+F8 column and adds a column for F7 and a column for F8. The F7 numbers are the increase of F7 since the most recent power cycle, divided by the days since the most recent power cycle. The F8 numbers are the increase of F8 since the most recent power cycle, divided by the days since the most recent power cycle.

Looking at the second table, it appears that "days since power cycle" does sometimes have a bad effect on F7, but the bad effect on F8 appears stronger and more consistent. I considered deleting the first table but its raw attribute data might be of use to someone.

TABLE 1 of 2:
DateF7 = total NAND pages written by host pcF8 = total NAND pages written by ssd's FTL controllerPower Cycle CountDays since last Power CycleAverage Daily Increase of F7+F8 since last Power Cycle
06/13/2020
262,686,703
1,473,283,999
108
0​
06/14/2020
262,910,985
1,473,494,607
108
1​
434,890​
06/15/2020
263,158,511
1,473,915,964
108
2​
551,887​
06/16/2020
263,409,107
1,474,189,617
108
3​
542,674​
06/17/2020
263,707,360
1,474,463,037
108
4​
549,924​
06/18/2020
264,027,678
1,474,962,512
108
5​
603,898​
06/19/2020
264,393,018
1,475,285,550
108
6​
617,978​
06/20/2020
264,637,602
1,475,570,003
108
7​
605,272​
06/21/2020
264,982,101
1,475,855,429
108
8​
608,354​
06/22/2020
265,731,603
1,476,265,234
108
9​
669,571​
06/23/2020
266,093,539
1,477,233,559
108
10​
735,640​
06/24/2020
266,432,523
1,477,876,771
108
11​
758,054​
06/25/2020
266,736,954
1,478,364,036
108
12​
760,857​
06/26/2020
267,037,173
1,479,090,682
108
13​
781,319​
06/27/2020
267,452,250
1,479,926,339
109
0​
06/28/2020
267,670,029
1,480,075,115
110
0​
06/29/2020
267,831,467
1,480,373,747
110
1​
460,070​
06/30/2020
268,055,495
1,480,843,566
110
2​
576,959​
07/01/2020
268,326,984
1,481,210,089
110
3​
597,310​
07/02/2020
268,508,455
1,481,530,249
110
4​
573,390​
07/03/2020
268,775,459
1,481,904,177
110
5​
586,898​
07/04/2020
269,097,697
1,482,223,281
110
6​
595,972​
07/05/2020
269,547,282
1,482,570,922
110
7​
624,723​
07/06/2020
269,865,012
1,483,031,681
110
8​
643,944​
07/07/2020
270,135,115
1,483,464,707
110
9​
650,520​
07/08/2020
270,363,723
1,484,133,984
110
10​
675,256​
07/09/2020
270,771,455
1,484,669,872
110
11​
699,653​
07/10/2020
271,198,165
1,485,203,829
110
12​
721,404​
07/11/2020
271,431,890
1,485,821,233
110
13​
731,383​
07/12/2020
271,707,066
1,486,451,312
110
14​
743,802​
07/13/2020
271,997,906
1,486,834,809
110
15​
739,171​
07/14/2020
272,324,624
1,487,297,232
110
16​
742,295​
07/15/2020
273,153,092
1,488,206,919
110
17​
800,875​
07/16/2020
273,416,549
1,488,853,892
110
18​
806,961​
07/17/2020
273,817,659
1,489,272,204
110
19​
807,617​
07/18/2020
274,050,476
1,489,652,733
110
20​
797,903​
07/19/2020
274,283,713
1,490,015,309
110
21​
788,280​
07/20/2020
274,530,619
1,490,354,959
110
22​
779,111​
07/21/2020
275,430,443
1,490,818,821
110
23​
804,527​
07/22/2020
275,682,753
1,491,351,610
110
24​
803,717​
07/23/2020
275,912,458
1,491,814,714
110
25​
799,281​
07/24/2020
276,163,396
1,492,417,946
110
26​
801,392​
07/25/2020
276,449,163
1,493,190,585
111
0​
07/26/2020
276,681,393
1,493,390,752
111
1​
432,397​
07/27/2020
276,913,201
1,493,709,105
111
2​
491,279​
07/28/2020
277,166,220
1,494,108,564
111
3​
545,012​
07/29/2020
277,403,763
1,494,384,251
111
4​
537,067​
07/30/2020
277,643,532
1,494,779,839
111
5​
556,725​
07/31/2020
277,941,002
1,495,192,802
111
6​
582,343​
08/01/2020
278,191,969
1,495,544,822
111
7​
585,292​
08/02/2020
278,567,870
1,495,949,954
111
8​
609,760​
08/03/2020
278,742,032
1,496,598,242
111
9​
633,392​
08/04/2020
278,908,331
1,497,170,767
111
10​
643,935​
08/05/2020
279,068,589
1,497,806,637
111
11​
657,771​
08/06/2020
279,231,233
1,498,262,587
111
12​
654,506​
08/07/2020
279,606,834
1,498,876,719
111
13​
680,293​
08/08/2020
279,848,556
1,499,294,138
111
14​
678,782​
08/09/2020
280,146,248
1,499,766,181
111
15​
684,845​
08/10/2020
280,361,128
1,500,068,175
111
16​
674,347​
08/11/2020
280,588,594
1,500,447,345
111
17​
670,364​
08/12/2020
281,072,906
1,500,913,119
111
18​
685,904​
08/13/2020
281,440,082
1,501,327,447
111
19​
690,936​
08/14/2020
281,833,461
1,501,764,591
111
20​
697,915​
08/15/2020
282,062,167
1,502,242,593
111
21​
698,334​
08/16/2020
282,303,454
1,503,058,695
111
22​
714,655​
08/17/2020
282,433,106
1,503,615,182
111
23​
713,415​
08/18/2020
282,577,538
1,504,068,079
111
24​
708,578​
08/19/2020
283,112,821
1,504,520,466
111
25​
719,742​
08/20/2020
283,293,751
1,505,210,866
111
26​
725,572​
08/21/2020
283,519,047
1,505,901,288
111
27​
732,614​
08/22/2020
283,744,498
1,506,447,632
111
28​
734,014​
08/23/2020
283,953,592
1,506,963,532
111
29​
733,703​
08/24/2020
284,140,881
1,507,622,138
111
30​
737,442​
08/25/2020
284,375,335
1,508,317,499
111
31​
743,648​
08/26/2020
284,616,616
1,508,900,885
111
32​
746,180​
08/27/2020
284,870,529
1,509,449,189
111
33​
747,878​
08/28/2020
285,127,137
1,510,497,763
111
34​
764,269​
08/29/2020
285,419,643
1,511,513,502
112
0​
08/30/2020
285,602,557
1,511,772,947
113
0​
08/31/2020
285,755,692
1,512,116,302
113
1​
496,490​
09/01/2020
285,912,183
1,512,479,611
113
2​
508,145​
09/02/2020
286,116,449
1,512,697,903
113
3​
479,616​
09/03/2020
286,241,394
1,512,978,477
113
4​
461,092​
09/04/2020
286,405,230
1,513,353,713
115
0​
09/05/2020
286,507,048
1,513,622,384
115
1​
370,489​
09/06/2020
286,613,571
1,513,899,954
115
2​
377,291​
09/07/2020
286,775,593
1,514,091,439
115
3​
369,363​
09/08/2020
286,962,315
1,514,532,938
115
4​
434,078​
09/09/2020
287,094,637
1,514,739,354
115
5​
415,010​
09/10/2020
287,176,773
1,515,458,638
115
6​
479,411​
09/11/2020
287,307,151
1,516,474,060
115
7​
574,610​
09/12/2020
287,377,354
1,516,949,786
115
8​
571,025​
09/13/2020
287,473,179
1,517,354,319
115
9​
563,173​
09/14/2020
287,557,148
1,517,635,436
115
10​
543,364​
09/15/2020
287,675,321
1,518,207,113
115
11​
556,681​
09/16/2020
288,306,718
1,518,672,448
115
12​
601,685​
09/17/2020
289,202,742
1,518,971,066
115
13​
647,297​
09/18/2020
289,271,628
1,519,596,805
115
14​
650,678​
09/19/2020
289,348,814
1,520,074,910
115
15​
644,319​
09/20/2020
289,420,533
1,520,720,060
115
16​
648,853​
09/21/2020
289,573,467
1,521,770,541
115
17​
681,474​
09/22/2020
289,651,773
1,523,259,467
115
18​
730,683​
09/23/2020
289,751,249
1,524,816,490
115
19​
779,410​
09/24/2020
289,828,172
1,525,663,097
115
20​
786,616​
09/25/2020
289,921,600
1,526,606,212
115
21​
798,518​
09/26/2020
290,006,153
1,527,946,681
116
0​
09/27/2020
290,313,118
1,528,168,257
116
1​
528,541​
09/28/2020
290,387,065
1,528,563,122
116
2​
498,677​
09/29/2020
290,511,492
1,528,894,695
117
0​
09/30/2020
290,572,303
1,529,066,586
117
1​
232,702​
10/01/2020
290,641,215
1,529,353,854
117
2​
294,441​
10/02/2020
290,700,175
1,529,731,168
117
3​
341,719​
10/03/2020
290,816,455
1,530,217,438
117
4​
406,927​
10/04/2020
290,882,910
1,532,024,503
117
5​
700,245​
10/05/2020
290,946,886
1,532,492,621
117
6​
672,220​
10/06/2020
291,041,878
1,532,824,408
117
7​
637,157​
10/07/2020
291,089,075
1,533,310,821
117
8​
624,214​
10/08/2020
291,228,154
1,533,731,083
117
9​
617,006​
10/09/2020
291,329,136
1,534,273,830
117
10​
619,678​
10/10/2020
291,490,268
1,535,545,502
117
11​
693,598​
10/11/2020
291,635,420
1,536,339,745
117
12​
714,082​
10/12/2020
291,750,877
1,537,376,552
117
13​
747,788​
10/13/2020
291,821,329
1,538,213,497
117
14​
759,189​
10/14/2020
291,875,046
1,538,802,787
117
15​
751,443​
10/15/2020
291,937,972
1,539,852,616
117
16​
774,025​
10/16/2020
292,013,400
1,541,264,724
117
17​
815,996​
10/17/2020
293,032,409
1,542,192,439
117
18​
878,815​
10/18/2020
293,101,812
1,543,189,726
117
19​
888,703​
10/19/2020
293,158,338
1,544,286,934
117
20​
901,954​
10/20/2020
293,218,796
1,545,957,965
117
21​
941,456​
10/21/2020
293,341,664
1,546,632,948
118
0​

TABLE 2 of 2:
Days since last Power CycleAverage Daily Increase of F7+F8 since last Power CycleAverage Daily Increase of F7 since last Power CycleAverage Daily Increase of F8 since last Power Cycle
0​
1​
434,890​
224,282​
210,608​
2​
551,887​
235,904​
315,983​
3​
542,674​
240,801​
301,873​
4​
549,924​
255,164​
294,760​
5​
603,898​
268,195​
335,703​
6​
617,978​
284,386​
333,592​
7​
605,272​
278,700​
326,572​
8​
608,354​
286,925​
321,429​
9​
669,571​
338,322​
331,248​
10​
735,640​
340,684​
394,956​
11​
758,054​
340,529​
417,525​
12​
760,857​
337,521​
423,336​
13​
781,319​
334,652​
446,668​
0​
0​
1​
460,070​
161,438​
298,632​
2​
576,959​
192,733​
384,226​
3​
597,310​
218,985​
378,325​
4​
573,390​
209,607​
363,784​
5​
586,898​
221,086​
365,812​
6​
595,972​
237,945​
358,028​
7​
624,723​
268,179​
356,544​
8​
643,944​
274,373​
369,571​
9​
650,520​
273,898​
376,621​
10​
675,256​
269,369​
405,887​
11​
699,653​
281,948​
417,705​
12​
721,404​
294,011​
427,393​
13​
731,383​
289,374​
442,009​
14​
743,802​
288,360​
455,443​
15​
739,171​
288,525​
450,646​
16​
742,295​
290,912​
451,382​
17​
800,875​
322,533​
478,341​
18​
806,961​
319,251​
487,710​
19​
807,617​
323,559​
484,057​
20​
797,903​
319,022​
478,881​
21​
788,280​
314,937​
473,343​
22​
779,111​
311,845​
467,266​
23​
804,527​
337,409​
467,118​
24​
803,717​
333,864​
469,854​
25​
799,281​
329,697​
469,584​
26​
801,392​
326,668​
474,724​
0​
1​
432,397​
232,230​
200,167​
2​
491,279​
232,019​
259,260​
3​
545,012​
239,019​
305,993​
4​
537,067​
238,650​
298,417​
5​
556,725​
238,874​
317,851​
6​
582,343​
248,640​
333,703​
7​
585,292​
248,972​
336,320​
8​
609,760​
264,838​
344,921​
9​
633,392​
254,763​
378,629​
10​
643,935​
245,917​
398,018​
11​
657,771​
238,130​
419,641​
12​
654,506​
231,839​
422,667​
13​
680,293​
242,898​
437,395​
14​
678,782​
242,814​
435,968​
15​
684,845​
246,472​
438,373​
16​
674,347​
244,498​
429,849​
17​
670,364​
243,496​
426,868​
18​
685,904​
256,875​
429,030​
19​
690,936​
262,680​
428,256​
20​
697,915​
269,215​
428,700​
21​
698,334​
267,286​
431,048​
22​
714,655​
266,104​
448,550​
23​
713,415​
260,171​
453,243​
24​
708,578​
255,349​
453,229​
25​
719,742​
266,546​
453,195​
26​
725,572​
263,253​
462,319​
27​
732,614​
261,848​
470,767​
28​
734,014​
260,548​
473,466​
29​
733,703​
258,773​
474,929​
30​
737,442​
256,391​
481,052​
31​
743,648​
255,683​
487,965​
32​
746,180​
255,233​
490,947​
33​
747,878​
255,193​
492,685​
34​
764,269​
255,235​
509,035​
0​
0​
1​
496,490​
153,135​
343,355​
2​
508,145​
154,813​
353,332​
3​
479,616​
171,297​
308,319​
4​
461,092​
159,709​
301,383​
0​
1​
370,489​
101,818​
268,671​
2​
377,291​
104,171​
273,121​
3​
369,363​
123,454​
245,909​
4​
434,078​
139,271​
294,806​
5​
415,010​
137,881​
277,128​
6​
479,411​
128,591​
350,821​
7​
574,610​
128,846​
445,764​
8​
571,025​
121,516​
449,509​
9​
563,173​
118,661​
444,512​
10​
543,364​
115,192​
428,172​
11​
556,681​
115,463​
441,218​
12​
601,685​
158,457​
443,228​
13​
647,297​
215,193​
432,104​
14​
650,678​
204,743​
445,935​
15​
644,319​
196,239​
448,080​
16​
648,853​
188,456​
460,397​
17​
681,474​
186,367​
495,108​
18​
730,683​
180,364​
550,320​
19​
779,410​
176,106​
603,304​
20​
786,616​
171,147​
615,469​
21​
798,518​
167,446​
631,071​
0​
1​
528,541​
306,965​
221,576​
2​
498,677​
190,456​
308,221​
0​
1​
232,702​
60,811​
171,891​
2​
294,441​
64,862​
229,580​
3​
341,719​
62,894​
278,824​
4​
406,927​
76,241​
330,686​
5​
700,245​
74,284​
625,962​
6​
672,220​
72,566​
599,654​
7​
637,157​
75,769​
561,388​
8​
624,214​
72,198​
552,016​
9​
617,006​
79,629​
537,376​
10​
619,678​
81,764​
537,914​
11​
693,598​
88,980​
604,619​
12​
714,082​
93,661​
620,421​
13​
747,788​
95,337​
652,451​
14​
759,189​
93,560​
665,629​
15​
751,443​
90,904​
660,539​
16​
774,025​
89,155​
684,870​
17​
815,996​
88,348​
727,649​
18​
878,815​
140,051​
738,764​
19​
888,703​
136,333​
752,370​
20​
901,954​
132,342​
769,612​
21​
941,456​
128,919​
812,537​
0​
 
Oct 19, 2020
4
0
10
0
I think you should bite the bullet and do a secure erase. Chances are your ssd will recover from the buggy state it's in and operate normally and you won't have to run ridiculous workarounds anymore.
 

Lucretia19

Great
Feb 5, 2020
101
5
95
2
I think you should bite the bullet and do a secure erase. Chances are your ssd will recover from the buggy state it's in and operate normally and you won't have to run ridiculous workarounds anymore.
I assume that where you wrote "chances are" you mean you think the probability exceeds 1/2. How have you calculated or estimated the probability? What sources of information or personal experiences lead you to believe a secure erase is likely to help? In my conversations with Crucial's tech support in January/February, they never suggested trying a secure erase.

Do you have any insight into the kind of problem that a secure erase could fix, yet is also tamed by selftests?

I'm not the only person with this excessive write amplification problem. On my ssd, the occasional write bursts by the ssd's FTL controller correlate perfectly with the well-known Crucial ssd "Bogus Current Pending Sectors" bug. Assuming the bursts correlate perfectly also on the ssds of the many people who suffer from the "Bogus Current Pending Sectors" bug, the excessive write amplification is a very wide-spread phenomenon, which makes me doubt whether a secure erase would solve it.

Because the write amplification got worse over time, my hunch is that a secure erase, if it does help, would only help for a few months. The ssd was new when installed at the end of July 2019. I didn't log SMART data in 2019 -- I was only alerted each time Remaining Life decreased -- so my records aren't conclusive, but I noticed the decreases of Remaining Life accelerated, and since my usage of the computer hadn't radically changed it's reasonable to assume the acceleration was due to increasing write amplification. I reached that conclusion retrospectively after I logged more and learned more. It's consistent with the increasing amplification that I logged beginning 1/15/2020; by then Overall WAF had reached 5.5 and daily WAF was probably an order of magnitude higher. (By "overall" I mean it includes the ssd's entire history: 1 + F8/F7. Not a 1 + ΔF8/ΔF7 snapshot where the deltas are measured over a short period of time, like daily.) The high daily WAF caused Overall WAF to keep climbing and it peaked at 7.26 on 2/23/2020, when I began experimenting with selftests. I think the triggering event that caused the amplification to skyrocket was in late December 2019 when I significantly reduced the writing by the pc (to about 80 kBytes/second average, by moving or symlinking some "hot" folders to my hard drive).

The SMART data from a few other Crucial users suggests write amplification is worse for users whose computers don't write heavily to the ssd. In other words, the lower the rate of writing by the pc, the more frequent and larger are the write bursts by the ssd's FTL controller. It's almost as if the ssd has firmware that was designed to make it fail soon after the warranty expires even if it isn't used much.

If my hunch that a secure erase would be at most a temporary fix is correct, then its manual labor, downtime and risk sounds more ridiculous than automated selftests.

On the other hand, the risk associated with nearly nonstop selftests is unknown. I speculate that the 30 seconds pause between selftests allows the FTL controller's low priority background processes enough runtime to maintain the health of the ssd, and it would be nice to learn whether this is true. It will remain unknown if no one is willing to continue the selftests regime for many years.

The selftests are either a "ridiculous workaround" as you say, or a sensible workaround of a ridiculous bug.
 
Oct 19, 2020
4
0
10
0
If my hunch that a secure erase would be at most a temporary fix is correct, then its manual labor, downtime and risk sounds more ridiculous than automated selftests.
I disagree. You'll have to copy 500GB of data back and forth once. It's not a hard job as long as you have enough empty space around.

Secure erase will be resetting FTL and all the other internal data structures firmware is using. If your SSD has somehow tainted internal data causing it to go berserk a secure erase has a good chance of fixing that. (preferably coupled with a firmware update).

If that broken internal data was created by an early firmware version, and if by any chance newer firmware versions won't create such invalid internal data the bug won't relapse again.

If secure erase fixes the issue it'll confirm the broken FTL theory.

If it reoccurs shortly after it'll prove it wasn't fixed in later firmware versions.

If it doesn't change anything at all it'll prove the problem is not there but somewhere else.

In any case you'll have a valuable information about probable causes and use cases causing it. You'll have more data points to provide to Crucial if they got interested in the issue.

I'm not the only person with this excessive write amplification problem. On my ssd, the occasional write bursts by the ssd's FTL controller correlate perfectly with the well-known Crucial ssd "Bogus Current Pending Sectors" bug.
I'm aware you're not the only person with WAF problem, but we're not seeing people screaming over rooftops about rapidly decreasing health on MX500 drives, either. So it must be still a rarely occuring problem which is why I'm suggesting a secure erase.

Also beware that correlation between write bursts and bogus current_pending_sector might be only one way, ie. bogus Current_Pending_Sector might be occuring on more drives - maybe even on all drives (ie. not necessarily only on the drives plagued with the WAF bug)
 
Last edited:

Lucretia19

Great
Feb 5, 2020
101
5
95
2
I disagree. You'll have to copy 500GB of data back and forth once. It's not a hard job as long as you have enough empty space around.
What I meant is that if the effect of a secure erase is only temporary, it would not be a one-time task; it would need to be repeated, perhaps every few months.

Secure erase will be resetting FTL and all the other internal data structures firmware is using. If your SSD has somehow tainted internal data causing it to go berserk, a secure erase has a good chance of fixing that. (preferably coupled with a firmware update). If that broken internal data was created by an early firmware version, and if by any chance newer firmware versions won't create such invalid internal data the bug won't relapse again.
The ssd had the latest firmware when it was new and no update was available in February. There may be a new version now, but Crucial's website says no update is available. It looks like they have a problem figuring out how to allow users to overwrite the drive's existing firmware. (Maybe another bug in the drive?) Here's a tomshardware thread about that issue: https://forums.tomshardware.com/threads/crucial-mx500-firmware-update-error.3648524/
Here's where Crucial says no update is available (without the website knowing what version I have!):
https://www.crucial.com/support/ssd-support/mx500-support

What kind of tainted internal structures would cause the write amplification to go berserk? You haven't yet tried to answer the questions I asked about your sources of information, and your answer about the possible nature of the problem is more vague than I prefer.

If secure erase fixes the issue it'll confirm the broken FTL theory.

If it reoccurs shortly after it'll prove it wasn't fixed in later firmware versions.

If it doesn't change anything at all it'll prove the problem is not there but somewhere else.

In any case you'll have a valuable information about probable causes and use cases causing it. You'll have more data points to provide to Crucial if they got interested in the issue.
A secure erase may be an experiment worth trying, and I'll consider it someday... but not until after Crucial solves their inability to update the firmware. I don't have time now for the "secure erase, restore from backup" ritual. However, if the short term result is that the excessive amplification appears to have gone away, I think that would be less conclusive than you do. Crucial's bug might not manifest at first, especially if it's intentional.

I would also need to seriously consider whether to trust a firmware update. If Crucial intentionally designed the MX500 to die soon after the warranty expires (by arranging for FTL write bursts to compensate when the pc isn't writing much), they might choose to raise the priority of the background process that's responsible for the write bursts, when that process isn't getting as much runtime as they want it to. In other words, they could prevent selftests from mitigating the problem. If their list of what's fixed by a firmware update doesn't include excessive amplification (or something that seems intimately connected to amplification) then I'd be less inclined to trust the update.

I'm aware you're not the only person with WAF problem, but we're not seeing people screaming over rooftops about rapidly decreasing health on MX500 drives, either. So it must be still a rarely occuring problem which is why I'm suggesting a secure erase.

Also beware that correlation between write bursts and bogus current_pending_sector might be only one way, ie. bogus Current_Pending_Sector might be occuring on more drives - maybe even on all drives (ie. not necessarily only on the drives plagued with the WAF bug)
I think few people have their pc write to their ssd as little as my pc does. This would explain why we don't see more people complaining about excessive amplification. It's not necessarily a rarely occurring problem; it could simply be that most people aren't aware that their ssd has excess amplification, because for them the ratio of excess amplification to necessary amplification is smaller. They have much more of the necessary amplification. Why would it occur to them that some of their amplification is unnecessary?

On my ssd, the write bursts correlate perfectly with Bogus Current Pending Sectors. In particular, every time it changes briefly to 1, there's a corresponding write burst. Why do you think other people might experience a brief change to 1 without a corresponding write burst? If that's so, what might it imply about the nature of the two behaviors?

It would help if some of the people who've complained about the Current Pending Sectors bug would perform high frequency logging of both Current Pending Sectors and NAND Pages Written by the FTL Controller (F8), preferably in a format such as .csv that would have one row per pair and could be opened by spreadsheet or database software, so that this question could be settled.
 

Lucretia19

Great
Feb 5, 2020
101
5
95
2
Another reason why I'm skeptical about whether a secure erase would reduce the ssd's excess write amplification is my recent analysis (posted here a few days ago) about the beneficial effect of power cycling on write amplification. If excess amplification is caused by corrupt FTL data that a secure erase would clean up, why does it take many days after a power cycle before the corruption resumes causing high amplification?

Below is a table of ssd log data that's similar to the tables I posted a few days ago. (The differences are described in notes further below.) The pattern after the most recent power cycle (7pm Oct 20) continues as before: Daily WAF grows higher when the ssd goes a long time without being power cycled, and each power cycle reduces Daily WAF back to a lower value.
Date​
Power Cycle Count​
ΔF7
1 row​
ΔF8
1 row​
ΔF7+ΔF8,
1 row​
Daily WAF
= 1 +
ΔF8/ΔF7​
09/04/2020
115
163,836​
375,236​
539,072​
3.29
09/05/2020
115
101,818​
268,671​
370,489​
3.64
09/06/2020
115
106,523​
277,570​
384,093​
3.61
09/07/2020
115
162,022​
191,485​
353,507​
2.18
09/08/2020
115
186,722​
441,499​
628,221​
3.36
09/09/2020
115
132,322​
206,416​
338,738​
2.56
09/10/2020
115
82,136​
719,284​
801,420​
9.76
09/11/2020
115
130,378​
1,015,422​
1,145,800​
8.79
09/12/2020
115
70,203​
475,726​
545,929​
7.78
09/13/2020
115
95,825​
404,533​
500,358​
5.22
09/14/2020
115
83,969​
281,117​
365,086​
4.35
09/15/2020
115
118,173​
571,677​
689,850​
5.84
09/16/2020
115
631,397​
465,335​
1,096,732​
1.74
09/17/2020
115
896,024​
298,618​
1,194,642​
1.33
09/18/2020
115
68,886​
625,739​
694,625​
10.08
09/19/2020
115
77,186​
478,105​
555,291​
7.19
09/20/2020
115
71,719​
645,150​
716,869​
10.00
09/21/2020
115
152,934​
1,050,481​
1,203,415​
7.87
09/22/2020
115
78,306​
1,488,926​
1,567,232​
20.01
09/23/2020
115
99,476​
1,557,023​
1,656,499​
16.65
09/24/2020
115
76,923​
846,607​
923,530​
12.01
09/25/2020
115
93,428​
943,115​
1,036,543​
11.09
09/26/2020
116
84,553​
1,340,469​
1,425,022​
16.85
09/27/2020
116
306,965​
221,576​
528,541​
1.72
09/28/2020
116
73,947​
394,865​
468,812​
6.34
09/29/2020
117
124,427​
331,573​
456,000​
3.66
09/30/2020
117
60,811​
171,891​
232,702​
3.83
10/01/2020
117
68,912​
287,268​
356,180​
5.17
10/02/2020
117
58,960​
377,314​
436,274​
7.40
10/03/2020
117
116,280​
486,270​
602,550​
5.18
10/04/2020
117
66,455​
1,807,065​
1,873,520​
28.19
10/05/2020
117
63,976​
468,118​
532,094​
8.32
10/06/2020
117
94,992​
331,787​
426,779​
4.49
10/07/2020
117
47,197​
486,413​
533,610​
11.31
10/08/2020
117
139,079​
420,262​
559,341​
4.02
10/09/2020
117
100,982​
542,747​
643,729​
6.37
10/10/2020
117
161,132​
1,271,672​
1,432,804​
8.89
10/11/2020
117
145,152​
794,243​
939,395​
6.47
10/12/2020
117
115,457​
1,036,807​
1,152,264​
9.98
10/13/2020
117
70,452​
836,945​
907,397​
12.88
10/14/2020
117
53,717​
589,290​
643,007​
11.97
10/15/2020
117
62,926​
1,049,829​
1,112,755​
17.68
10/16/2020
117
75,428​
1,412,108​
1,487,536​
19.72
10/17/2020
117
1,019,009​
927,715​
1,946,724​
1.91
10/18/2020
117
69,403​
997,287​
1,066,690​
15.37
10/19/2020
117
56,526​
1,097,208​
1,153,734​
20.41
10/20/2020
117
60,458​
1,671,031​
1,731,489​
28.64
10/21/2020
118
122,868​
674,983​
797,851​
6.49
10/22/2020
118
103,384​
240,568​
343,952​
3.33
10/23/2020
118
532,575​
481,317​
1,013,892​
1.90
10/24/2020
118
161,794​
184,042​
345,836​
2.14
A few days (Sept 16/17, Oct 17, Oct 23) show high ΔF7 (a lot of writing by the host pc) because Windows updated itself. Daily WAF was lower on those days, not surprising.

The table columns differ from the tables posted a few days ago in the following ways:
  1. It goes back only to Sept 4th when Power Cycle Count reached 115.
  2. It goes up to today (Oct 24) so it includes data after the most recent (118th) power cycle late on Oct 20.
  3. It shows the Daily WAF, which gives a rough idea at a quick glance.
  4. It's closer to the raw data; it doesn't show the cumulative moving average of the deltas nor the number of days since the most recent power cycle.
 

Lucretia19

Great
Feb 5, 2020
101
5
95
2
My pc's Power Plan is set to "never sleep." Today I did a quick test of sleep mode because it ought to be more efficient to occasionally power cycle the ssd by a brief sleep than by a shutdown/restart. (In order to knock down the WAF that grows after many days of ssd uptime.) The result: Yes... putting the pc to sleep for a few seconds incremented the ssd's Power Cycle Count. Not a surprise. The next time that WAF grows high, I'll verify that brief sleep knocks down WAF.

An unwanted side effect: The brief sleep stopped the ssd selftest that had been running. That was a surprise, since I thought I'd configured the ssd so that after a power cycle it would automatically resume a selftest that had been in progress. I should doublecheck that setting. If I can't get the ssd to reliably resume a selftest that was in progress, I would create a task that launches a selftest when the computer wakes. (I don't think there's a general solution to that, since Windows Task Scheduler doesn't have a "trigger on wake" option. But I don't need a general solution... the sleeps that I'm planning would be automatically scheduled, so the wakes could be scheduled too, using the "wake the computer to run scheduled task" condition, where the scheduled task would simply launch a selftest.)
 

ASK THE COMMUNITY