SSD Write Endurance 25nm Vs 34nm

[Christopher]

MTRON 7000 PRO 16GB SLC, Day 16

GiB : 70779.94
TiB : 69.1210
MBs : 56.66

187 93/93/114768131

[Christopher]

MTRON 7000 PRO 16GB SLC, Day 17

GiB : 74906.19
TiB : 73.1519
MBs : 56.77

187 93/92/121666447

[canthearu]

Torqx day 10

Drive hours: 238
GB written: 12334.66 (12.0457 TB)
Avg MB/s: 17.54

Bad blocks: 83
Wear cycle counter: 0/824/1251

Intel 520 - day 1

Drive hours: ?
ASU Count GB written: 4395.34 (4.2924TB)
Smart GB written: 8287.53 (8.09TB , 265201 raw)
Avg MG/s: 75.89 MB/s

Reallocated sectors (05): 0
Failure count (AB, AC): 0 program, 0 erase
Media Wearout Indicator (E9): 100 (0 raw)
NAND writes (F9): 5861GB

Notes

I had some problems with ASU losing count due to power outages.

[Ao1]

The 09 value reported in the Intel tool box gives you the power on hours, but the counter starts from 894800 so you have to deduct that amount to get the power on hours. If you want to double check try smartmon tools and the command I gave earlier. You will need to deduct 6 hours from the value reported by device statistics to make allowance for the factory test period.

[A||uSiOn]

canthearu ,

Thank you for making this thread even more interesting by adding the Intel 520 Fingers crossed it amazes us all!

[Christopher]

Maybe something interesting will happen to the 520! All I know is, running an SLC drive is worse than watching paint dry... I mean, it only takes 24 hours for paint to dry. I'm not sure it's possible to kill an SLC drive outside of firmware failures.

Really, when was the last time you heard an antecdote about someone wearing out an SLC drive?

[B.A.T]

Todays update:
Kingston V+100
And it dropped out again.....And because I'm not home until next friday I can't take a cold reboot.

Intel X25-M G1 80GB
374,4503 TiB
21226 hours
Reallocated sectors : 00
MWI=102 to 98
MD5 =OK
33.55 MiB/s on avg


m4
477.7978 TiB
1890 hours
Avg speed 70.73 MiB/s.
AD gone from 100 to 85.
P/E 8400.
MD5 OK.
Reallocated sectors : 00

[canthearu]

The 09 value reported in the Intel tool box gives you the power on hours, but the counter starts from 894800 so you have to deduct that amount to get the power on hours.

Cool, that makes it easy ... I'll report power on hours, and reorganise the stats a bit for my next report.

[Hopalong X]

Maybe something interesting will happen to the 520! All I know is, running an SLC drive is worse than watching paint dry... I mean, it only takes 24 hours for paint to dry. I'm not sure it's possible to kill an SLC drive outside of firmware failures.

Really, when was the last time you heard an antecdote about someone wearing out an SLC drive?

You could create another fireworks display?

[Christopher]

You could create another fireworks display?

That is almost guaranteed to happen. Especially since at some point I'm going to measure drive power consumption un endurance testing.

[Hopalong X]

Dragging out the old Tesla coil this time I hope.

[canthearu]

Maybe something interesting will happen to the 520!

An interesting drive to test would be my A-RAM 100gig sandforce drive with 34nm NAND.

Given the force 3 was able to sniff the 1PB limit with 128gig of 25nm NAND, my desktop drive would probably smash that with it's 34nm NAND.

But alas, I can't part with it right now

[Christopher]

Dragging out the old Tesla coil this time I hope.

What? No, I was gonna use a Taser... why? Is that bad??


@canthearu

I've never ever heard of A-Ram (where are you located in the world?) I've heard of another Hong Kong brand, ASA (that uses SF). And the Chronos Deluxe 60 obviously used 32nm Toshiba, but didn't fare all that well either.

[canthearu]

@canthearu

I've never ever heard of A-Ram (where are you located in the world?) I've heard of another Hong Kong brand, ASA (that uses SF). And the Chronos Deluxe 60 obviously used 32nm Toshiba, but didn't fare all that well either.

Australia. A-RAM isn't a big brand here either, but one of the online stores here had stock of the sandforce 1 drives and was a bit cheaper then the OCZ drive, so I went with it.

The Chronos Deluxe 60 might not have faired that well at the end, but hell, it was quick getting there ..... much quicker then my Intel 520 is proving to be!

[Christopher]

Australia. A-RAM isn't a big brand here either, but one of the online stores here had stock of the sandforce 1 drives and was a bit cheaper then the OCZ drive, so I went with it.

The Chronos Deluxe 60 might not have faired that well at the end, but hell, it was quick getting there ..... much quicker then my Intel 520 is proving to be!

Well that makes sense. The Samsung 830 64GB was ripping off 11,000GB a day for a while. It's on hiatus for a while, but it should have some good times still ahead.

The Chronos Deluxe 60 is a beast, but suffers from the SF problem -- it's write speeds are nowhere near peak (for a couple reasons, but if you've ever seen a graph of SF writes you'll understand why). My CD60 was putting in 10,000GB a day host writes, but only 7300GB NAND writes. Hopefully I'll have some drives that beat the doors off it shortly. I got a new Vertex Turbo on the way (which won't be that fast), but who knows what surprises tomorrow may hold? I've been sitting on a few candidates for a while, but I can only run so many drives at once, and the Vertex Turbo and the MTRON (and the 64GB Samsung 830 eventually) are already gonna take up some spots in the endurance rig.

[canthearu]

but suffers from the SF problem -- it's write speeds are nowhere near peak. My CD60 was putting in 10,000GB a day host writes, but only 7300GB NAND writes.

To be fair, I think people get a bit too upset about compressible vs incompressible performance. All 60gig drives are designed fairly cheap, to use as few chips as possible without utterly crippling the controller, so you shouldn't be expecting 256gig drive performance out of a 64gig drive. It doesn't help that virtually drive makers post extremely over the top sequential write values in their specs ... For example, my torqx 64gig manual says 230meg per second sequential writes ... fat chance, it does 18meg or so a second sustained and that is it.

It isn't really fair to criticise a sandforce drive for slow incompressible writes when it is paired with 64gig of Async NAND, As the Chronos Deluxe show, pair it up with some nice Toggle Mode NAND, and the sandforce controller can do incompressible about as well as any other drive.

That isn't to say the sandforce drives can't go faster when doing incompressible data .... would be nice to see a bit of a boost in this area when the sandforce 3 comes out

[Christopher]

To be fair, I think people get a bit too upset about compressible vs incompressible performance. All 60gig drives are designed fairly cheap, to use as few chips as possible without utterly crippling the controller, so you shouldn't be expecting 256gig drive performance out of a 64gig drive. It doesn't help that virtually drive makers post extremely over the top sequential write values in their specs ... For example, my torqx 64gig manual says 230meg per second sequential writes ... fat chance, it does 18meg or so a second sustained and that is it.

It isn't really fair to criticise a sandforce drive for slow incompressible writes when it is paired with 64gig of Async NAND, As the Chronos Deluxe show, pair it up with some nice Toggle Mode NAND, and the sandforce controller can do incompressible about as well as any other drive.

That isn't to say the sandforce drives can't go faster when doing incompressible data .... would be nice to see a bit of a boost in this area when the sandforce 3 comes out

I just mean for the purposes of this testing.

[canthearu]

I just mean for the purposes of this testing.

Haha, I was just ranting in general!

[Christopher]

MTRON 7000 PRO 16GB SLC, Day 18

GiB : 80 002.81
TiB : 78.1277
MBs : 56.90

187 92/92/130169283



Snooze!

It keeps getting a little faster, but that's about it. In two more days MWI should hit 90, making 20 days to MWI 90, or right around 200 days to MWI 0.

[canthearu]

It keeps getting a little faster, but that's about it. In two more days MWI should hit 90, making 20 days to MWI 90, or right around 200 days to MWI 0.

And that is only what the MWI counts ... Who knows how long this SSD will last beyond it's specified rating.

[Christopher]

And that is only what the MWI counts ... Who knows how long this SSD will last beyond it's specified rating.

I would guess maybe 3 years --BUT-- I'm not sure it's even possible, given the conditions and speed of the writes. I don't think anyone knows for sure.

I'm seriously thinking that absent firmware c0ckup, the drive lasts indefinitely from an endurance perspective.

[RaZz!]

first of all, thanks for this thread. it's very interesting and informative!

...but just out of curiosity: do you guys really buy all these drives just to send them straight to hell? or do you get these for cheap from supporters/sponsors?

keep it up!

[canthearu]

first of all, thanks for this thread. it's very interesting and informative!

...but just out of curiosity: do you guys really buy all these drives just to send them straight to hell? or do you get these for cheap from supporters/sponsors?

keep it up!

Unfortunately I have no-one to sponsor me, so I have to buy the drives myself.

There certainly is an opening if you are interested.

Edit: I don't view it as sending the drives to hell, although the troqx belongs there. I view it as a sacrifice in the process of learning how SSDs live and die. How can I preach the longevity of SSDs if it is never tested.

[Christopher]

first of all, thanks for this thread. it's very interesting and informative!

...but just out of curiosity: do you guys really buy all these drives just to send them straight to hell? or do you get these for cheap from supporters/sponsors?

keep it up!

Yeah, I'd be happy to test anything a manufacturer wants to send me (but no one has taken me up on my offer), but like canthearu, I don't think of it as just sending a drive to Hell. You get to see all sorts of behavior you'd not otherwise ever get to see. And since it's usually easier to "get a result" on a less expensive, lower capacity drive, it lowers the barrier of entry. If you could only use 512GB drives there wouldn't be many SSDs in the test since A)they are expensive and B) it would take a long damn time. I was able to get some BNIB MTRONs pretty cheap for example, and if you shop around there are some really good deals.

It's not just an endurance test in the strictest sense, it's a variety of things all rolled into one.

EDIT
But yeah, canthearu, that Torqx suqx something fierce.


My Patriot with the 3016 suqx worse though:

[B.A.T]

first of all, thanks for this thread. it's very interesting and informative!

...but just out of curiosity: do you guys really buy all these drives just to send them straight to hell? or do you get these for cheap from supporters/sponsors?

keep it up!

No sponsors. But the smallest capacities is'n too expencive and some of them are used SSD.