SSD Write Endurance 25nm Vs 34nm

[Christopher]

Mushkin Chronos Deluxe 60 Update, Day 27

05 2
Retired Block Count

B1 12
Wear Range Delta

F1 255227
Host Writes

E9 196853
NAND Writes

E6 100
Life Curve

E7 10
Life Left

Average 126.49MB/s Avg
RST drivers, Asus M4g-z

620 Hours Work (24hrs since the last update)
Time 25 days 20hours

12 GiB Minimum Free Space
The AnvilPro beta9 preview's 500ms/thousand files is working, and I've reduced the pause a few seconds.

If it runs overnight, I can declare victory on the 2281 TRIM/delete problem, but the additional loops per hour might be hurting speed slightly.

Thanks for the preview, Anvil. So far so good.

[Computurd]

but I sure do have an unreasonable fondness for my Indilinx units -- all of them.

couldnt agree more, excerpt the complaining part LOL. seriously LOVE my indys, still have them all. dont even need my 8 member array, but i keep it for sentimental reasons, seriously i love them.

i have 12 of them spread out around here....was going to endurance test one of them for the sake of science, just cant bring myself to do it

[One_Hertz]

It will be interesting to see what One_Hertz finds.

Soldered on the power connector today. I am afraid the SSD stays in the busy state permanently

[Christopher]

couldnt agree more, excerpt the complaining part LOL. seriously LOVE my indys, still have them all. dont even need my 8 member array, but i keep it for sentimental reasons, seriously i love them.

i have 12 of them spread out around here....was going to endurance test one of them for the sake of science, just cant bring myself to do it

It's not like I didn't know that random performance gets worse as capacity decreases... I didn't necessarily need an expert to testify on that subject, but Eric seems to be a nice guy, and I'm glad he's still sticking his head in the door over here.

I sure as hell didn't need the last two Indy drives I bought. I put my oldest Agility 60 in my Dell D630 laptop, just because I like the drive so much. It sure as hell isn't the fastest Sata II drive I have laying around, but using any of the others just doesn't feel right. Weird, isn't it? I surely don't feel that way about my GPUs. Like I said, it's irrational.

[johnw]

Soldered on the power connector today. I am afraid the SSD stays in the busy state permanently

You mean you cannot get the SSD to be recognized by the BIOS and to mount the filesystem in Windows?

That is what I expected (although I hoped otherwise). I guess that is what Samsung 470s do when their flash is not only worn out for writing, but also worn out for storing and readback. It makes sense, I suppose, since the index of LBA to flash address must be stored on flash, and if it cannot read the flash, it cannot load its index, so it cannot do anything.

Feel free to run any experiments you can think of, including destructive ones.

[bulanula]

So then what is the consensus ? Can we not read the Samsung because it genuinely failed or because of the connector damage ?

[Anvil]

Kingston SSDNow 40GB (X25-V)

405.20TB Host writes
Reallocated sectors : 12

MD5 OK

37.56MiB/s on avg (~10 hours)

--

Corsair Force 3 120GB

01 84/50 (Raw read error rate)
05 2 (Retired Block count)
B1 65 (Wear range delta)
E6 100 (Life curve status)
E7 49 (SSD Life left)
E9 198247 (Raw writes)
F1 264008 (Host writes)

106.25MiB/s on avg (~10 hours)

power on hours : 780

WR delta looks to just keep on growing, strange!

[One_Hertz]

You mean you cannot get the SSD to be recognized by the BIOS and to mount the filesystem in Windows?

That is what I expected (although I hoped otherwise). I guess that is what Samsung 470s do when their flash is not only worn out for writing, but also worn out for storing and readback. It makes sense, I suppose, since the index of LBA to flash address must be stored on flash, and if it cannot read the flash, it cannot load its index, so it cannot do anything.

Feel free to run any experiments you can think of, including destructive ones.

I have specialized hardware for these things so I don't use the BIOS, but yes the SSD never becomes ready and never gives access to any data.

So then what is the consensus ? Can we not read the Samsung because it genuinely failed or because of the connector damage ?

I am quite certain the connector is fine. The drive IS getting proper power and it is displaying the BUSY registry so the controller is at least somewhat starting up; it just doesn't work.

[B.A.T]

A little update from me:
I'll be joining in with 2 more ssd from monday. Those will be:
- Intel X25-M G1 80GB
- MTRON Pro SSD 32GB

What makes them interesting is 50nm mlc and G1 controller without trim for the intel and slc nand for the MTRON. I'll post all the usual data before the test is started.
I would like to thank GullLars for selling them to me.

[Hopalong X]

Poor Samsung 470

[Christopher]

A little update from me:
I'll be joining in with 2 more ssd from monday. Those will be:
- Intel X25-M G1 80GB
- MTRON Pro SSD 32GB

What makes them interesting is 50nm mlc and G1 controller without trim for the intel and slc nand for the MTRON. I'll post all the usual data before the test is started.
I would like to thank GullLars for selling them to me.

Doesn't the MTRON use the Indilinx controller?

Anvil, I don't think the 500ms helped as much as I thought. It crashed during file deletion again this morning.

[Anvil]

The MTRON is much older than the Indilinx stuff, iirc it's their own controller and technology.

Strange that it didn't help but I'm not surprised.
The new fw has a fix for TRIM so lets see what it brings, I expect most will make the new firmware available this week.

[bluestang]

M225->Vertex Turbo 64GB Update:

534.69 TiB (587.90 TB) total
1460.48 hours
10847 Raw Wear
118.41 MB/s avg for the last 16.33 hours (on W7 x64)
MD5 OK
C4-Erase Failure Block Count (Realloc Sectors) from 9 to 10.
(1=Bank 6/Block 2406; 2=Bank 3/Block 3925; 3=Bank 0/Block 1766; 4=Bank 0/Block 829; 5=Bank 4/Block 3191; 6=Bank 7/Block 937; 7=Bank 7/Block 1980; 8=Bank 7/Block 442; 9=Bank 7/Block 700; 10=Bank 2/Block 1066)

[Christopher]

Well, as long as the Mushkin only has about 12GB of static data, it doesn't matter. For some reason, even with the additional delay in beta9, I still have to compensate by lowering minFreeSpace [since I have much more static data now]. I'll be upgrading to the new FW, but Mushkin doesn't want to say that there is a 3.3.2 FW, much less that it's going to be along shortly.

If it's going to crash while deleting files, it will do it rather soonish. I'm back to about 10K+ files per loop, so that should solve the problem in any event, somehow.

[bluestang]

Anyone have an idea of what the H and L mean in the Bad Block List below where it shows for example (1760,H,E) or (3191,L,E).

[Christopher]

So the grey listings were initial bad blocks (marked with I), and the red were the new ones...

H and L could mean HIGH or LOW...
I can't figure out what any of the letters mean. Does P mean program failure, and E mean erase failure?

Bluestang, you posted the link for the viewer, but I couldn't figure out why it wouldn't work. I just assumed I had to use it from a dos environment, or without AHCI.

[Christopher]

Double Post

[Anvil]

Well, as long as the Mushkin only has about 12GB of static data, it doesn't matter. For some reason, even with the additional delay in beta9, I still have to compensate by lowering minFreeSpace [since I have much more static data now]. I'll be upgrading to the new FW, but Mushkin doesn't want to say that there is a 3.3.2 FW, much less that it's going to be along shortly.
...

Corsair has the same "policy", they won't tell, I think they should make an exception, this is a serious bug for those that are affected.

Patriot has already told that it will be made available this week. (I've got a WildFire 240GB)

[Christopher]

The big announcement that OCZ made makes me more confident that, contrary to what had been said in the past, that firmware alone can fix the problem. It won't require Intel to intervene with new oroms, or any of the other non-SF solutions.

When it releases for the Mushkin, its going in the SandForce torture chamber.

If the new FW is really the cure-all solution, you would think ever vendor would at least acknowledge it's existence.

UPDATE
I finally broke the 200TB NAND writes on the Mushkin, but WRD seems to be spiking again due to more static data and less minFreeSpace.

[bluestang]

So the grey listings were initial bad blocks (marked with I), and the red were the new ones...

H and L could mean HIGH or LOW...
I can't figure out what any of the letters mean. Does P mean program failure, and E mean erase failure?

Bluestang, you posted the link for the viewer, but I couldn't figure out why it wouldn't work. I just assumed I had to use it from a dos environment, or without AHCI.

Yes, grey are intitial bad blocks (B8) andthe reds are the program (C3) and erase (C4) failures. I don't think it's as simple as High or Low.

And at the top under the white lettering "BAD BLOCK LIST" there is a legend of what the letters mean in grey, at least for the second part of the initials.

It works under windows. Just double click "SmartViewer_2_21.exe" and a DOS style window prompt appears with what to do.

[B.A.T]

m4
708.2702 TiB
2545 hours
Avg speed 89.16 MiB/s.
AD gone from 207 to 203.
P/E 12287.
Value 01 (raw read error rate) has changed from 4 to 5.
MD5 OK.
Still no reallocated sectors


I've updated to Betav9 of ASU.

Kingston V+100:
Will be online again from monday.

[deathman20]

Wow, 4x the life span. Now I wouldn't even know where to start guessing of when the M4 will start failing. 850-900GB sounds good for a start, maybe 1PB!

Good Luck with the other drives there BAT. G1 drive will probably suffer what I had with my G2's not being able to trim properly in Raid. That SLC drive though, hope you got a PC that will be running for the next few years to test that one

[Christopher]

Bluestang,

I just forgot that the Indilinx wiper and smartviewer don't work with RST on my ICH8m C2Duo Lappy.

I have 66 bad blocks on the older Agility, but they're all initial bad blocks.

EDIT

That Mtron might have really high WA/questionable wear leveling, so while it might last a long, long time, it might not be optimized in the same way an X25-E is.

That's why you have to do the test... if they lasted as long as the manufacturer said, their wouldn't really be a need to endurance test them.

Very curious to see how they work out. I know the G1 Intel drives aren't quite as good as the G2 (without trim), but should still be very interesting as well.




I've been thinking of another drive to test with, so I'm taking suggestions.

[B.A.T]

Wow, 4x the life span. Now I wouldn't even know where to start guessing of when the M4 will start failing. 850-900GB sounds good for a start, maybe 1PB!

Good Luck with the other drives there BAT. G1 drive will probably suffer what I had with my G2's not being able to trim properly in Raid. That SLC drive though, hope you got a PC that will be running for the next few years to test that one

I'll probably put the mtron in my foldingrig where it can work for as long as it takes.
I think the write speed will suffer, how it will affect the WA I don't know. We'll just have to wait and see.

Bluestang,

I just forgot that the Indilinx wiper and smartviewer don't work with RST on my ICH8m C2Duo Lappy.

I have 66 bad blocks on the older Agility, but they're all initial bad blocks.

EDIT

That Mtron might have really high WA/questionable wear leveling, so while it might last a long, long time, it might not be optimized in the same way an X25-E is.

That's why you have to do the test... if they lasted as long as the manufacturer said, their wouldn't really be a need to endurance test them.

Very curious to see how they work out. I know the G1 Intel drives aren't quite as good as the G2 (without trim), but should still be very interesting as well.

I've been thinking of another drive to test with, so I'm taking suggestions.

According to their spec sheet they claim 50GB a day for 140 years. I wonder what the avg write speed will be.

[Anvil]

The lack of TRIM will indeed make the G1 an interesting drive.
I expect it will settle at 40-50MiB/s.

I'm not sure about the Mtron, iirc it was generally pretty good at QD1.

@Christopher
What about the Vertex Plus?, the 120GB is cheap and the Indilinx was/is a pretty good sequential performer.
(not sure about the current firmware though)

or, you can pick up another Mushkin and put 10GB (or more) extra over-provisioning on it, it might do a lot of difference vs the standard one.
(it could be any drive of course but over-provisioning is interesting and none of the current drives are over-provisioned)

or, you can get one of the synchronous based drives and compare vs the async one I've got.

a lot of options, I'll be looking for another drive within a month or two, and there's always the new Intel