SSD Write Endurance 25nm Vs 34nm

[B.A.T]

Todays update:

m4
901.3566 TiB
3809 hours
Avg speed 74.39 MiB/s.
AD gone from 98 to 95.
P/E 15538.
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
626.27 TiB
23121hours
Reallocated sectors : 157 to 158
Available Reserved space: 90
MWI=124
MD5 =OK
42.19 MiB/s on avg


Intel X25-E 64GB
62.85 TiB
208-30=178 hours
Reallocated sectors : 0
Available Reserved space: 100
MWI=99
MD5 =OK
101.36 MiB/s on avg


It's like looking at grass grow but now it's only 2 weeks left before 1 PiB is broken

I'm also considering to throw in a C300 128GB since Vapor never returned here. Got a couple of them laying around and it could be interesting to compare with the m4.

[Anvil]

@canthearu

I forgot about that one

@B.A.T

Looking good, not much variation in speed on the E.

A C300 would be great!
Can you manage all the drives on one rig

I'll order something by Monday!

[B.A.T]

I got 8 SATA ports on that motherboard so no problem. Just need to manage some power and sata cables behind the hdd cage.

[Anvil]

Next time I'll remove the Torqx 2 from the chart.

Endurance_cr_20120519.png

--

Kingston SSDNow 40GB (X25-V)

966.06TB Host writes
Reallocated sectors : 05 43
Available Reserved Space : E8 99
POH 8659
MD5 OK

37.93MiB/s on avg (~6 hours)

[Anvil]

@canthearu

Is there anything you'd like to add or adjust on the Torqx 2 summary?
(the list is updated in post #1)

[canthearu]

@Anvil

You need to change model name to Torqx 2.

Also, the Torqx 2 was showing a lot of write failures in the windows log and extremely unresponsive at the end.

I'm running now 1155 Z77 board now if that makes a difference if you want to put it in.

[Anvil]

I'll fix it.

When did you make the change to Z77?

I think I'm going to skip IB, a bit surprised that it's hotter than SB, maybe they can get it fixed in some revision.

[canthearu]

About when I got the vertex 4 (probably 1 week after) I got the Z77 board along with an i3-2120

[Christopher]

IB Board: Maybe
IB CPU: Probably Not

[Anvil]

That is a good point, chipset USB3 and probably Thunderbolt on some boards.
(must check)

[Ao1]

A bit random, but somehow I think this thread will be running for a very long time.

The vast bulk of the NAND market is for mobile devices. Performance for mobile devices is more dependent on larger block performance, hence continuing increases in page sizes as NAND geometry shrinks. This of course is not helpful for SSD’s, so as page sizes increase and NAND geometry shrinks SSD controllers become more complex. Geometry shrinks get progressively harder, but the limit is still not known. The pace of geometry shrinks (and the resulting changes in NAND composition) is really the technology driver for SSD’s. The controllers have to keep up. In this context it’s easy to see that controllers have a very limited life span and consequently I think this is why the controller market has consolidated.

The very structure of NAND may change altogether.



NAND die shrinks





Bring on the 3D ReRAM

[Christopher]

Samsung 830 256GB Day 63

1,495,909.52GiB
1,640.85 TiB

292.17 MB/s Average
1415 Hours

6500 Wear Leveling Count
MWI 1

5/0 Erase/Program Fail
Used Reserved Block Count: 10/20480 sectors

[B.A.T]

Todays update:

m4
907.8599 TiB
3834 hours
Avg speed 74.36 MiB/s.
AD gone from 95 to 91.
P/E 15648.
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
630.06 TiB
23146 hours
Reallocated sectors : 158
Available Reserved space: 90
MWI=124
MD5 =OK
42.33 MiB/s on avg


Intel X25-E 64GB
71.77 TiB
233-30=203 hours
Reallocated sectors : 0
Available Reserved space: 100
MWI=99
MD5 =OK
101.31 MiB/s on avg


With this speed on the E it'll write 3.6 PiB the next year

[Christopher]

With this speed on the E it'll write 3.6 PiB the next year

I'm not 100% convinced that you can actually wear out the E's 50nm SLC unless you're doing full span high QD random writes... who knows?

[B.A.T]

First I need to find out how long it takes to mwi=1%. Then it's possible to do some predictions.

[Ao1]

You could reset the timed workload media workload and it would be possible to tell quite quickly how long it will take for the MWI to expire.

What I found with the 830 is that 4k RW did not wear out the drive any quicker because the write speeds were so slow in comparison to the normal ASU workload. I would suspect that 4K RW at QD 34 would help reduce wear whilst maintaining a higher write speed, but it would be interesting to experiment and monitor using the workload timer. Lol, I paid £600 each for 2 of the X25-E’s when they very first came out and now I see Christopher picks one up for peanuts. I’ll keep my eye out for one.

[B.A.T]

You can get the same Kingston as I have for £199 each at amazon Run for it!!

[canthearu]

Vertex 4 reallocated sectors at 15.

I'm seriously doubting this drive is going to last more then another 100-200TiB.

[canthearu]

Intel 520 60GB - Day 88

Drive hours: 2,058
ASU GiB written: 653,647.21 GiB (638.33 TiB)
Avg MB/s: 95.42 MB/s
MD5: OK

Host GB written (F1): 658,293.19 GiB (642.86 TiB, 21065382 raw)
NAND writes (F9): 466,026 GiB (455.10 TiB)

Reallocated sectors (05): 0
Failure count (AB, AC): 0 program, 0 erase
Raw Error Rate (8B): 112 normalised
Avaliable Reserved Space (AA): 100 normalised
Media Wearout Indicator (E9): 1

Vertex 4 128GB - Day 35

Drive hours: 844 (556 + 288)
ASU GiB written: 366,203.35 GiB (357.62 TiB)
Avg MB/s: 124.59 MB/s
MD5: OK

Sector GB written (E8): 369,363.89 GiB (360.71 TiB, 486156078177 + 288456133860 raw)
Reallocated sectors (05): 16
Raw Read Error Rate (01): 6
Remaining Life (E9): 39 (69 current - 30)

12 new bad blocks for the vertex 4. Upgrading ASU and restarting testing.

[Anvil]

What sort of over-provisioning is there by default on the Vertex 4?

[canthearu]

@Anvil

The normal 7% from the conversion of binary gigabytes (GiB) into decimal gigabytes. (GB)

[Anvil]

OK, so the norm then.

I normally over-provision ~10% on most drives, especially if they are in a raid.
(during my V4 128GB testing it has been set to 100GB, 19.5GB over-provisioning)

[kensiko]

Vertex 4 reallocated sectors at 15.

I'm seriously doubting this drive is going to last more then another 100-200TiB.

I've been a fan of OCZ for a few years now for putting to market many new interesting products.

The Octane and V4 are not. I'm still a fan of SandForce and this testing just put more weight to my opinion.

[B.A.T]

Todays update:

m4
913.7974 TiB
3857 hours
Avg speed 74.35 MiB/s.
AD gone from 91 to 88.
P/E 15749.
MD5 OK.
Reallocated sectors : 00


Intel X25-M G1 80GB
633.51 TiB
23169hours
Reallocated sectors : 158 to 159
Available Reserved space: 90
MWI=124
MD5 =OK
42.44 MiB/s on avg


Intel X25-E 64GB
79.86 TiB
257-30=227 hours
Reallocated sectors : 0
Available Reserved space: 100
MWI=99
MD5 =OK
101.23 MiB/s on avg

[stevecs]

I normally over-provision ~10% on most drives, especially if they are in a raid.
(during my V4 128GB testing it has been set to 100GB, 19.5GB over-provisioning)

Just out of curiosity are you guys doing this with MHDD or similar (cutting the drive/changing the MaxLBA) or is this something else with a custom firmware? The handful of SSD's here I've just cut the drives but haven't really done a lot of stress testing to see the efficacy of that.