SSD Write Endurance 25nm Vs 34nm

[Christopher]

This thread just hit 1,000,000 views! To commemorate the event, it's time for a new drive...

[Anvil]

Kingston SSDNow 40GB (X25-V)

804.42TB Host writes
Reallocated sectors : 05 30
Available Reserved Space : E8 99
POH 7227
MD5 OK

32.02MiB/s on avg (~205 hours)

--

> 1.000.000 views is rather nice

I'll decide on what SSD to test when I get back home. (in about 2 weeks)
(Corsair Force 3 120GB, Intel 320 80GB, Crucial C300 are amongst the candidates)

[canthearu]

Patriot Torqx-2 64GB - Day 33

Drive hours: 790
GiB written: 49,808.00 GiB (48.64 TiB)
Avg MB/s: 19.72 MB/s

Bad blocks: 2/0/83
Wear cycle counter: 0/3113/4710

Intel 520 60GB - Day 24

Drive hours: 591
Avg MB/s: 91.59 MB/s

Host GB written (F1): 178401.88 GiB (174.22 TiB, 5708860 raw)
NAND writes (F9): 126,346 GiB (123.38 TiB)

Reallocated sectors (05): 0
Failure count (AB, AC): 0 program, 0 erase
Media Wearout Indicator (E9): 61

[Anvil]

Endurance_cr_20120320.png

@canthearu

Did you get the counter working?

--

I'll check out where B.A.T is, haven't seen him posting lately.

[Christopher]

No, I mean, I have a new drive.

I'm waiting until the 24hr mark before I post the first update.

It was either this drive, or the Plextor M3P 128 I just got.

[Anvil]

I got that Christopher

I'll be adding one as well, when I get back home.

I've got an M2P but I don't consider using it for Endurance testing, it's currently a boot drive and will stay so for a while.
(it feels just like any other drive but I consider the SMART attributes to be inadequate for testing at this point)

[Christopher]

No, the M3P's are useless for smart data. But it's pretty damn fast all the same...

If the Marvell partners get some new FW in the coming months which rectifies some of those issues, I'll run it.

But if I wasn't familiar with the performance of the M3P or the Intel 510, and I was asked which one I felt was faster, I'd say the Intel 510. Somehow, the 510 is what you would get if Intel decided to make a drive just for my day-to-day tasks.

-----------
FYI --
I bought one more old-new-stock MTRON. A 7500 PRO, and it's QD1 4K reads are fantastic. It's sequential writes are much faster as well. 4KRW is not improved, but it's just more MTRON-y.

But I've lost all taste for those things, at least as far as endurance testing goes. I will take it apart at some point to see what's inside.

I know it's doing something different as the FW is the same.

[canthearu]

Anvil: Nope, double-clicking on the TiB written box does nothing. Clicking on the GiB written box brings up a message, but after restarting, the GiB written isn't fixed.

[Christopher]

New Drive Time!
Samsung 830 256GB Day 0

I can't believe I'm about to sentence another innocent drive to a life sentence of hard labor. What can I say? I'm out of control.

Here is a shot about 16 hours in to the process:



I've been studying the initial problem I had with the first drive. I think I have it figured out.
Basically, if it does not happen to this drive, I have it solved. If it does happen -- back to the drawing board.

I won't go nutty with the benches, so here's one:


From the perspective of speed vs. capacity, it's like having a 64GB drive doing 73MB/s (like the M4). It's just barely possible.

It's clocking about 293MB/s average -- and a TB per hour. It will take a couple days to get to the point where I can determine whether 177 is updating properly, but so far it look like it.

I was hoping for a little more speed though. It's averaging around 330MB/s thoughout the loop, not as close as the 64GB was to it's max write, but I don't know that it's a fair comparison.

[bulanula]

Just wanted to drop a line in here to take the time to thank everybody for making this possible.

Anvil for the software code and the volunteers for the SSDs.

Sad to be stuck in the peanut gallery but ATM I cannot change that.

Has anybody considered using Linux for these tests with a more automated reporting setup etc. ?

[Christopher]

Has anybody considered using Linux for these tests with a more automated reporting setup etc. ?

You mean, like a robot that will just post updates on XS for me? I'm looking in to it.

[canthearu]

Christopher:

Nice speed you are getting from the samsung.

My 240gig Sandisk extreme (got it yesterday) is only 160meg per sec in ASU endurance (should spend a little bit of time figuring out why speed is so low, could be due to sata 2 and microsoft drivers)

And I do think you are crazy for throwing a 256gig drive on the fire ... but I guess it was bound to happen sooner or later!

[Christopher]

Christopher:

Nice speed you are getting from the samsung.

My 240gig Sandisk extreme (got it yesterday) is only 160meg per sec in ASU endurance (should spend a little bit of time figuring out why speed is so low, could be due to sata 2 and microsoft drivers)

And I do think you are crazy for throwing a 256gig drive on the fire ... but I guess it was bound to happen sooner or later!

I'm probably going to get thrown under the jail for doing this to a 256GB drive. It's not quite 19 hours in and I already have 21 000GiB written. Just... awesome.

I almost bought a Extreme 120 (They were seriously cheap here for a day... ) but I decided on the Plextor M3P 128. I just figured the incompressible writes were not going to be stellar on the SDEx, no matter what.

My Mushkin Chronos Deluxe 120GB has a max seq. random data write of 240MB/s... but it can only manage 180MB/s with 47% data in ASU. SF writes are funny under the best of circumstances... a sawtooth-like wave keeping avg. speed low. It's the nature of the beast, but not a big deal under normal use. In the endurance test, sometimes you just need a bigger hammer.

[canthearu]

My Mushkin Chronos Deluxe 120GB has a max seq. random data write of 240MB/s... but it can only manage 180MB/s with 47% data in ASU. SF writes are funny under the best of circumstances... a sawtooth-like wave keeping avg. speed low. It's the nature of the beast, but not a big deal under normal use. In the endurance test, sometimes you just need a bigger hammer.

Yeah, they are, but the Extreme 240gig cost $329 and I just couldn't pass that up. Besides, the reads are always very quick (both compressed and incompressed) and in real use, I have no data source that can approach the maximum write speed of this drive. The write performance is only the bottleneck in benchmarks and ASU

[bluestang]

@Christopher,
Yes, you are definitely crazy...but it a good way...thanks for the 256 tester!

[Christopher]

You are most welcome, Bluestang.

[Christopher]

Vertex Turbo 64 Day 20

155590.84GiB
83.92MB/s

2789 Average Erase Count

45 MWI
529 hours

0/0/0 Program/Erase/Read Failures

----------------------------------

The Samsung wrote 25TB in 24 hours(!).

[Anvil]

That's pretty awesome Christopher

25TB per day on the Samsung, should be about 42 days (-1 day) until 1PB is reached, if it can maintain the speed.

[Christopher]

I think it may be really proactive about swapping the static data around, because for a couple loops I'll get a straight 330MB/s. Then, for a couple loops, it just drops substantially for the last ~80GB.

Coincidence? Maybe. The 64GB did something similar. I've read that one of the cores does reads, another does writes, and the third does houskeeping. I think the drop in speed is because the controller is doing something else. And looking at the temp graph, it seems like it's using significantly more or less power throughout the loop.

It does seem to be updating attribute 177 correctly.



And then, out of the blue, I get one loop that happens at 360MB/s. I stopped the test for 8 minutes last night to run some quick benches to check for degradation, and I was actually getting 415MB/s seq. writes with CDM, or 17MB/s better than I could manage out of the box. Reads and randoms were fine.

[ZOMGVTEK]

I almost bought a Extreme 120 (They were seriously cheap here for a day... ) but I decided on the Plextor M3P 128. I just figured the incompressible writes were not going to be stellar on the SDEx, no matter what.

How much did the SD Extreme 120 cost you? They're $120 shipped on the egg right now, which isn't bad at all. I'd love to see one of those tested, I currently have two in RAID0 in my notebook. They do the job well, and hopefully should be reliable.

Do you plan on trying to get some more definite data on the MTRON's? I dropped a 7000 in a fileserver and it definitely improved OS performance dramatically over the 7200RPM 3TB, especially when the boot drive is nearly full. I'd hope an SLC SSD would be more reliable for 24/7 mostly read only use than a mech variant, but its only 16GB...

[Christopher]

How much did the SD Extreme 120 cost you? They're $120 shipped on the egg right now, which isn't bad at all. I'd love to see one of those tested, I currently have two in RAID0 in my notebook. They do the job well, and hopefully should be reliable.

Do you plan on trying to get some more definite data on the MTRON's? I dropped a 7000 in a fileserver and it definitely improved OS performance dramatically over the 7200RPM 3TB, especially when the boot drive is nearly full. I'd hope an SLC SSD would be more reliable for 24/7 mostly read only use than a mech variant, but its only 16GB...

Well, I didn't end up buying the SanDisk, but they were on a Shell Shocker deal at the 'Egg last week.

Did you get the FW updated on the MTRON? I tried endurance testing two MTRONs (MOBI 3000 and PRO 7000), which didn't work out so well (I shorted one, the PRO broke). I'm using a third (a 3.5" 7000 16GB) as a boot drive in my endurance rig.

I don't know what the difference between 17 and 18 FW is, but it's not doing anything for the abysmal 4K RW.

[ZOMGVTEK]

My MTRON says it has 0.19R1H2 on it. What version is the one you sent me?



I've been slacking at doing anything about it, since that box has no VGA/input devices/optical drive.

[canthearu]

Do you plan on trying to get some more definite data on the MTRON's? I dropped a 7000 in a fileserver and it definitely improved OS performance dramatically over the 7200RPM 3TB, especially when the boot drive is nearly full. I'd hope an SLC SSD would be more reliable for 24/7 mostly read only use than a mech variant, but its only 16GB...

Hmmm, normally for a mostly read only load, a MLC drive would work just as well as the SLC drives. Intel 320/520 would be my first choice for reliability, with the 40gig 320 being quite inexpensive now.

[Christopher]

My MTRON says it has 0.19R1H2 on it. What version is the one you sent me?

Version .18! Geez. That's annoying -- I was able to get in touch with Imation to get the FW. It took a while, but they found it.

I mean, it's not like MTRON is still around to hand this stuff out. It will work on non-Imation branded MTRONS I think.

Somehow, your drive has .19, a FW Imation was not able to provide.




@canthearu // I don't know about ZOMGVTEK, but I got mine really cheap. I put it in the endurance rig just because it's the only good use I could find for a 16GB SLC SSD. The E-rig doesn't do all that much and only takes up about 11.2GB. I happen to think they're interesting.

[ZOMGVTEK]

Ever since the existence of the Battleship MTRON, I wanted an MTRON drive. When I noticed them super cheap on eBay, I had to snag one. SLC and its age makes it interesting.

I just decided to actually put it to use as a boot in a machine thats been running 24/7 for 4 years. The Windows install started to get really questionable and the thing would BSOD daily. It's been running for a few days now without issues, looks like it should be fine for years to come.