SSD Write Endurance 25nm Vs 34nm

As competition to Anvil's Intel 330, I present the Sandisk Extreme G25 - 120gig.

Sporting 128gig of 24nm Toggle NAND and the sandforce controller, Sandisk has created a little ripper of a drive at a very competitive price. Unfortunately, due to the nature of the sandforce controller, and the configuration of NAND in this drive, speed will be a little bit slower then the Intel 330. However, it is also a bit cheaper, and can be gotten in larger sizes then the intel 330.

Initial SMART values
Attachment 127367

Benchmarks
Attachment 127368
Attachment 127369

Running ASU endurance test
Attachment 127370

Notes: I does appear that the sandforce controller can manage bad blocks .... loading the initial data on the drive caused a couple of bad blocks to appear. This drive must have escaped the factory before it's bad blocks where fully found.

Edit: I am using same specs as other drives, but have 64gig of static data to match Anvil's 330.

The Sandisk Extreme should be very interesting vs the 330

I had expected the Extreme to be a bit faster than the Intel, having said that, I'm positively surprised by the performance of the 330.

Corsair Neutron could be the next one to test, scheduled to be available in July. (new controller)

Quote:

---

Originally Posted by Anvil

The Sandisk Extreme should be very interesting vs the 330

I had expected the Extreme to be a bit faster than the Intel, having said that, I'm positively surprised by the performance of the 330.

---

Well, the Intel 330 isn't using RAISE, so the sandisk extreme g25 has to work harder during writes ... that said, I'm getting about 120meg per sec speed, so it isn't a lot slower.

Intel 520 60GB - Day 100

Drive hours: 2,351
ASU GiB written: 745,570.30 GiB (728.09 TiB)
Avg MB/s: 93.37 MB/s
MD5: OK

Host GB written (F1): 750,789.25 GiB (733.19 TiB, 24025256 raw)
NAND writes (F9): 531,501 GiB (519.04 TiB)

Reallocated sectors (05): 0
Failure count (AB, AC): 0 program, 0 erase
Raw Error Rate (8B): 120 normalised
Avaliable Reserved Space (AA): 100 normalised
Media Wearout Indicator (E9): 1

Sandisk Extreme G25 - 120gb - Day 0

Drive hours: 2
ASU GiB written: 722.44 GiB (0.71 TiB)
Avg MB/s: 119.28 MB/s

Host GB written (F1): 820 GiB (0.80 TiB)
NAND writes (E9): 631 GiB (0.62 TiB)

Retired Block Count (05): 2
Failure count (AB, AC): 1 program, 0 erase
Raw Error Rate (01): 120 normalized
Media Wearout Indicator (E7): 100 normalized
Wear Range Delta (B1): 0 raw
Reported Uncorrectable Errors (BB): 0 raw

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

Plextor M3P 128GB Day 9
219,135.95 GiB
214.00 TiB

269.94 MB/s Avg, down from 273.49 MB/s in the other rig

225 Hours

Rellocated Event Count 0

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

Samsung 830 256GB Day 79

1,891,886.94 GiB
1,847.54 TiB

290.52 MB/s Average down from 299.72 MB/s in the other rig
1871 Hours

8214 Wear Leveling Count
MWI 1

7/0 Erase/Program Fail
Used Reserved Block Count: 14/28672 sectors

-----------------------
The 830 has 12,350 GiB more in LBA writes than ASU.


Out of the box reallocations are seldom positive. Is the 120GB Extreme using 64gbit dice?

Quote:

---

Originally Posted by canthearu

Well, the Intel 330 isn't using RAISE, so the sandisk extreme g25 has to work harder during writes ... that said, I'm getting about 120meg per sec speed, so it isn't a lot slower.

---

That's true, otoh the 330 is said to be using 6 channels. (6 out of 8)

I'll update post#1 with the new contestant.

Attachment 127378

Generic USB Drive – 4 GB
Writes: 619.6 GB
Average speed: 2.68 MB/s

The 330 is using all 8 channels on the 60/120s.

I suppose it's possible be that the 180GB is only using 6 channels with 4 way interleaving, but the 180GB 330 is also different in that it's using "redundancy" (ie Intel RAISE), not found on the other models.

All in all, the 330 is pretty excellent IMHO, I'm curious to see if the 120GB happens to last forever too.

Todays update:

m4
1004.1994 TiB
4222 hours
Avg speed 74.70 MiB/s.
AD gone from 38 to 35
P/E 17312.
CE 26
MD5 OK.
Reallocated sectors : 00
http://www.diskusjon.no/index.php?ap...tach_id=493274

Intel X25-M G1 80GB
686.87 TiB
23532 hours
Reallocated sectors : 123 to 127
Available Reserved space: 77 to 76
MWI= 82 to 76
MD5 =OK
39.04 MiB/s on avg
http://www.diskusjon.no/index.php?ap...tach_id=493273

Intel X25-E 64GB
204.58 TiB
620-30=590 hours
Reallocated sectors : 0
Available Reserved space: 100
MWI= 98
MD5 =OK
90.93 MiB/s on avg
http://www.diskusjon.no/index.php?ap...tach_id=493276

Mtron Pro 7025 32GB
ASU write 6.7407 TiB
Host write 12,1457 TiB
Total host write 59,6812 TiB
Day 2
MWI= 99
MD5 =OK
46.97 MiB/s on avg
http://www.diskusjon.no/index.php?ap...tach_id=493275

3 days left to 1 PiB :)
But after the Mtron started both my Intel drives have dropped their writespeed. Looks like my endurance rig is taking a beating at the moment :D

Quote:

---

Originally Posted by Christopher

The 330 is using all 8 channels on the 60/120s.

---

I've got a few unboxed 520's 120GB, I'll compare it to the 330, if they don't compare it could still be down to different firmwares.
Not sure where the number of channels discussion started or if it ever ended :), there also were speculations on 3K NAND which doesn't seem to fit either.

@B.A.T

3rd drive to get past 1PB, not bad :)

4 drives doesn't leave much room for other than testing (I expect it's doing nothing but testing), with that combination of drives it's probably what your rig can take.

7 months of writing and finally I'm at 1 PB. 1 PiB in another 3 days. This m4 has behaved differently than the first. It could either be the upgraded fw or quality of the nand but this one has less reported error than the first and the wa is also lower. The first had 1.06 and this is around 1,004

It's a dedicated rig placed in my basement so it doesn't need to do much else :)

4 test drives + a boot device is a lot for one rig to handle. If I start testing some more drives, I'll have to figure out an alternate hardware arrangement.

The M3P and 830 don't seem to be butting heads, but they are slightly slower in the dedicated enterprise rig.

Quote:

---

Originally Posted by Christopher

4 test drives + a boot device is a lot for one rig to handle. If I start testing some more drives, I'll have to figure out an alternate hardware arrangement.

The M3P and 830 don't seem to be butting heads, but they are slightly slower in the dedicated enterprise rig.

---

that's cuz they are A#1 Drives

Quote:

---

Originally Posted by AMD_Freak

that's cuz they are A#1 Drives

---

You know, they are badass units.

Quote:

---

Originally Posted by Christopher

4 test drives + a boot device is a lot for one rig to handle. If I start testing some more drives, I'll have to figure out an alternate hardware arrangement.

---

At the moment it's 5 ssd in that rig. I use an Intel X25-V 40GB as os-drive. The best thing about the endurance rig is the 6 SATA 6.0 Gb/s ports which makes it easy to test more than 2 at the same time with SATA 6.0 Gb/s

Quote:

---

Originally Posted by B.A.T

At the moment it's 5 ssd in that rig. I use an Intel X25-V 40GB as os-drive. The best thing about the endurance rig is the 6 SATA 6.0 Gb/s ports which makes it easy to test more than 2 at the same time with SATA 6.0 Gb/s

---

I have been contemplating swapping the mobo and CPU out for a newer uATX AMD. I have a couple old AMD Athlon IIs laying about, so I could just get a new board and trade out the H67. If I start testing another 6gbps drive, I'll have to do something and that might be more cost effective.

With a 3th 6.0 Gb/s ssd that might be a good idea. :up:

Not having more than 2x 6Gbs ports is a real pain.
My AMD rig (CHIV) developed some instability that I never was able to fix, I'll have to consider where to start making changes.

On the 330 vs 520 thing, I wasn't able to test the 520 on the same platform where the 330 was tested (it's now running W8) but from what I can see, performance is in favor of the 520, by a small margin.
I tried testing the 330 but wasn't able to get it back to FOB speed by normal means and I won't secure erase it during testing.
I might do an experiment at some point to see if performance can be restored to FOB state.

Quote:

---

Originally Posted by Anvil

The Sandisk Extreme should be very interesting vs the 330

I had expected the Extreme to be a bit faster than the Intel, having said that, I'm positively surprised by the performance of the 330.

Corsair Neutron could be the next one to test, scheduled to be available in July. (new controller)

---

Just got this one today:
http://monimag.eu/upload/977/as-ssd ...k 480gb.th.jpg
This is after throwing 300GB of data on it.

No, it's not for testing here :D

Intel 330 120GB

137.95TB Host writes
Reallocated sectors : 05 0
Available Reserved Space : E8 100
MWI 85
[F1] Total LBAs Written 4520422
[F2] Total LBAs Read 30800 // not much use in this one
[F9] Total NAND Writes 99577GB
POH 359
MD5 OK

122.85MiB/s on avg (~15 hours)

--

Kingston SSDNow 40GB (X25-V)

1015.13TB Host writes
Reallocated sectors : 05 46
Available Reserved Space : E8 99
POH 9085
MD5 OK

33.93MiB/s on avg (~15 hours)

--

kensiko

I've been thinking of that exact drive as I need the space (and they are currently discounted)
What's up with 4K writes at QD1? (a re-run might show normal figures)
Sequentials are looking good.

Quote:

---

Originally Posted by Anvil

What's up with 4K writes at QD1? (a re-run might show normal figures)

---

No, didn't change a bit. 480GB SSD are slower from the spec. I don't really care, it's fast enough for me. This SSD is used for games and storage, sequential speed is the most important for me in this case.

Intel 520 60GB - Day 101

Drive hours: 2,373
ASU GiB written: 752,803.39 GiB (735.16 TiB)
Avg MB/s: 93.17 MB/s
MD5: OK

Host GB written (F1): 758,057.09 GiB (740.29 TiB, 24257827 raw)
NAND writes (F9): 536,645 GiB (524.07 TiB)

Reallocated sectors (05): 0
Failure count (AB, AC): 0 program, 0 erase
Raw Error Rate (8B): 120 normalised
Avaliable Reserved Space (AA): 100 normalised
Media Wearout Indicator (E9): 1

Sandisk Extreme G25 - 120gb - Day 1
Drive hours: 24
ASU GiB written: 10,041.75 GiB (9.81 TiB)
Avg MB/s: 120.13 MB/s
MD5: OK

Host GB written (F1): 10,185 GiB (9.95 TiB)
NAND writes (E9): 7,569 GiB (7.39 TiB)

Retired Block Count (05): 2
Failure count (AB, AC): 1 program, 0 erase
Raw Error Rate (01): 120 normalized
Media Wearout Indicator (E7): 100 normalized
Wear Range Delta (B1): 3 raw
Reported Uncorrectable Errors (BB): 0 raw

The Sandisk has not had any further reallocations from the last 24 hours. I wonder when the media wear indicator is going to start moving? Also, the Power-on hours normalized value went from 100 to 0 in 24 hours ... so I think write throttling won't be a problem on this drive, unlike the silicon power drive from before.

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

Plextor M3P 128GB Day 10
241,493.45 GiB
235.83 TiB

272.10 MB/s Avg, down from 273.49 MB/s in the other rig

248 Hours

Rellocated Event Count 0

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

Samsung 830 256GB Day 79

1,916,184.25 GiB
1,871.27 TiB

304.34 MB/s Average
1894 Hours

8319 Wear Leveling Count
MWI 1

7/0 Erase/Program Fail
Used Reserved Block Count: 14/28672 sectors

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

I wonder about the Samsung, it just looks to go on and on and...

I looked at the 64GB drive (which was not exactly dead but still out of the contest) and you are way past what one could expect by just scaling up the figures from the 64GB drive.

415iTB and "P/E" (Host writes / Capacity) ended at ~7130 is the latest recording I've got and that would translate to 1660TiB, you are already +210TiB .

Attachment 127401

Anvil,

The 64GB 830 was subject to crazy high WA due to 01FW. When I put 02FW on it, WA went back to normal. Since the 256GB is on 03FW, the numbers aren't really comparable.

The 256GB is using 1 avg PE cycle for every 230GB written, the 64GB was using 1 for every 16GB or so for the first two weeks of testing until I put 02FW on it. After that it was 1 for every 58GB or so. I'd have to go back and look.

Attachment 127403Attachment 127404

When I stopped testing of the 64GB it had over 16,000PE cycles on it.