SSD Write Endurance 25nm Vs 34nm

[Christopher]

I'm particularly disappointed by the Turbo. I had such... high hopes for it. 116TB is nothing.

I'm thinking that maybe the Turbo isn't refurbished. Perhaps it just used lower binned 51nm Samsung because all the good stuff was gone. Another possibility is the new features of the 1.7FW. Supposedly, it uses different wear leveling, and maybe it's just not as good as 1.6FW.


Here is a quick and dirty chart from the last 23 hours of the Turbo's short, miserable life.


Read failures go from 1 to 72 in ~22 hours. I would just have to assume that it's relevant. I tried flashing it again a while back... but it's still dead.

[Christopher]

MTRON 7000 PRO 2.5" 16GB SLC, Day 1
100/100/7640236

GiB: 4,534.91
TiB: 4.4286
Avg MB/s 53.29


I'm not even sure what the differences are between the MOBI and PRO are.. To my knowledge, they perform identically.


UPDATE
I believe that the MOBI drives use Samsung SLC, while the PRO series utilizes Micron SLC. Now that I don't have to worry about breaking it, I can take the dead MOBI apart.


WAIT.....................................
The MOBI uses Hynix SLC.

[Hynix Korea HY27UG088G5B x 16] Data sheet is here

Let me also say that the quality of the PCB seems to be a good deal higher than modern drives, though the MTRONs were not cheap 4 years ago. It's actually quite wonderful.

[Anvil]

Kingston SSDNow 40GB (X25-V)

705.20TB Host writes
Reallocated sectors : 05 22
Available Reserved Space : E8 99
POH 6351
MD5 OK

32.80MiB/s on avg (~362 hours)

It is on a retention test as of tonight and will be reconnected late Sunday. (~50 hours)

--

Christopher,

The dead MOBI had Hynix NAND?

I'll get the new one into the charts but the old one didn't really get started so I wasn't planning on including it.

[Anvil]

It looks like attachments are back

I just managed to post a few just moments ago, lets hope they are back for good!

[Christopher]

Yes, the dead one had Hynix. I had the 3.5" 7025 PRO taken apart, but I couldn't really get it into the test rig, so I had to get another one (the one currently running). The 3.5" used Micron SLC, so I assume the 2.5" 7025 PRO does too.

I was just curious, that's all. I'm not familiar with all that many drives utilizing Hynix SLC...


Anvil, don't bother putting the old MTRON in the charts or anything.



-------------------------------------------------------------------------
MTRON 7000 PRO 2.5" 16GB SLC, Day 2
100/100/14101603

GiB: 8667.90
TiB: 8.6448
Avg MB/s 53.54

[Anvil]

The SF12xx drive that Vapor used had Hynix and I'm guessing that one of my 60GB Vertex 2's are Hynix as well, it's noticeably slower (in benchmarks) than the others I've got.

[Christopher]

The SF12xx drive that Vapor used had Hynix and I'm guessing that one of my 60GB Vertex 2's are Hynix as well, it's noticeably slower (in benchmarks) than the others I've got.

Hynix MLC is terrible, as far as I know. The older SLC might be great for all anyone knows, but I'm sure not very many drives ever used those chips. I'm sure there were some faster CF/SD cards that used the Hynix SLC for higher class ratings. Given the price difference between the MOBI and PRO MTRON's, I would guess that the Hynix was much less expensive than the Micron in 2008.

[Anvil]

Is it using 3GB static data? (like the previous one)

The MTRON's were really great and known as quality devices, it's been quiet lately though...

edit:

Could you check whether the MTRON is a 1.5Gb/s device?

[Christopher]

Anvil,

The MTRON is 1.5gbps. It's using 3GB of static data.

It's not native SATA, but is an ultra DMA IDE with a SiliconImage SATA bridge.

[Anvil]

OK, could you also check if it is 14.9GiB?
(it should be)

[Christopher]

14.90GB, Static data is actually 3.12GB.

I promise not to kill this MTRON on accident.

[Anvil]

Endurance_cr_20120210.png

I'll update as soon as B.A.T shows his face

--

Christopher,

You better leave that drive alone
(it could last for years though...)

[Ao1]

Christopher, hope you’re up for a long run. That Hynix NAND spec sheet has it down for 100,000 Program/Erase cycles. 2K page/ 128K Block.

HY27UG088G5B

PAGE READ / PROGRAM
Random access : 25us (max.)
Sequential access : 25ns (min.)
Page program time : 200us (typ.)

Block erase time: 1.5ms (Typ)

I don’t have much info for Hynix NAND but as a comparison to lower density NAND from Hynix:

H27UAG8T2B

3,000 Program/Erase cycles 8K page/2MiB Block

Page Read / PROGRAM
Random Access: 200 ㎲ (Max.)
Sequential Access : 25 ㎱ (Min.)
Page program time : 1600us (typ.)

Block erase time: 2.5ms (Typ)

As a comparison to Intel NAND

Intel JS29F64G08AAME1/ JS29F16B08CAME1/ JS29F32B08JAME1

5,000 Program/Erase cycles 8K page/2MiB Block

Page Read / PROGRAM
Random Access: 50 ㎲ (Max.)
Sequential Access : 20 ㎱ (Min.)
Page program time : 1200us (typ.)

Block erase time: 3.0ms (Typ)

[Christopher]

Ao1,

You're right, but it's actually worse that it would appear. The Hynix SLC was in the MOBI 3025 that I shorted out (and took apart today). The PRO 7025 currently running uses Micron SLC...

I would guess that they're pretty similar, but if I had to guess which was better, I'd say it's the Micron stuff.

I'm pretty deranged to even attempt such a thing.



A couple of weeks ago I took apart the 3.5" 7025 PRO, and I posted which Micron SLC it had (I was sure it was Samsung).

Just to be clear:
2.5" MTRON MOBI 3025 16GB <------I killed this one, took apart today
2.5" MTRON MOBI 3025 32GB <------System drive for the endurance rig
2.5" MTRON PRO 7025 16GB <------Currently testing
3.5" MTRON PRO 7025 16GB <------Took this apart, wouldn't fit in the testing rig


I wonder if Hynix and Samsung had some joint fabs in Korea (in the past)?
Maybe Hynix:Samsung::Micron:Intel



Fortunately, after 50 or so hours, MWI on the MTRON is now down to 99.
So let's see... 50 hours *100 = 5000 hours, or 208 days. If it died at MWI 0, that would be impressive, but since no drive dies at MWI 0, I'd say I'm in for a couple of years.

[Christopher]

MTRON PRO 7025 16GB SLC, Day 3


Yesterday, the MTRON seemed to jump back from 99/99/ to 100/99
I don't know why.

[Anvil]

I've just reconnected the Kingston (X25-V) and all is OK

[Christopher]

Anvil,

It was off for ~72 hours? That's pretty amazing, all things considered.

Now that I think about it, I wonder if Vapor's C300 is still alive (if it's just been sitting in a box since last fall, it would be very instructive if it was still working).

[Anvil]

Christopher,

It was ~50 hours, Friday night - late Sunday night.

It would be interesting to know what happened to the C300, if he doesn't show up within reasonable time I'll consider using one of my C300 64GB, I've got a few and they have been used for benching only, a couple of them should be as good as new. (I'll check)

[Christopher]

The MTRON PRO seems to be using Micron 29F8G08DAA SLC. I couldn't really find much on it, but I'm guessing it's 50nm, 100K PE rated.



------------------------------------------------------------------------------------------------------------
I would guess that Vapor's drive would/could have lasted 5 months unpowerered.

[B.A.T]

Endurance_cr_20120210.png
I'll update as soon as B.A.T shows his face

And I'm back. I've been busy with some family arrangement the last week.

Todays update.
Kingston V+100
359.7678 TiB
2228 hours
Avg speed 22.23 MiB/s
AD still 1.
168= 1 (SATA PHY Error Count)
P/E?
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
336,04 TiB
20928 hours
Reallocated sectors : 00
MWI=114 to 108
MD5 =OK
45.66 MiB/s on avg


m4
410.9854 TiB
1590 hours
Avg speed 64.66 MiB/s.
AD gone from 140 to 115.
P/E 7248.
MD5 OK.
Reallocated sectors : 00


The Kingston dropped out again and it looks like it happens after 24h during a random write. It has done that at least the last 3-4 times.
Also the m4 has dropped in write speed by quite a lot of MiB/s. I don't know why but I'll watch it closely and see if it stabilize or drop further.

[Christopher]

Welcome back, B.A.T. Do you think the M4's speed decrease is related to the M4 shuffling around the static data? I don't remember the first M4 dropping like that, though.

[Anvil]

Kingston SSDNow 40GB (X25-V)

708.08TB Host writes
Reallocated sectors : 05 22
Available Reserved Space : E8 99
POH 6376
MD5 OK

34.36MiB/s on avg (~24 hours)

--

Looks like the m4 is about to overtake the Samsung

Endurance_cr_20120213.png

[Christopher]

Looks like the m4 is about to overtake the Samsung

Yeah, and it only took the Samsung being on hiatus for two weeks, 5 weeks of writing at half speed, and probably some unholy goat sacrifices for the M4 to achieve that victory.



Luckily, I have some sacrificial goats and chickens left over from dark days of SandForce 2281 BSoD "remedies".

----------------------------------------------------------------------
MTRON 7000 PRO 16GB SLC, Day 5
187 98/98/36187185

TiB
21.7186

GiB:
22240.51 GiB

Avg MB/s:
53.67

[Anvil]

The drop in speed on B.A.T's m4 was unexpected.

[Christopher]

The Average MB/s on the MTRON is over 600MB/s.....

I don't think that's right.....