Samsung 830
169 574 GiB in 167 hours
Just shy of 7 days
3.5x the Turbo speed
Had 271.7GB on it before testing started.
Octane
42 205 GiB in 73.64 hours
~ 3 days
Started out of box
Vertex Turbo
197 542 GiB in 671 hours
~27 days
Started out of box
-----------------------
Fantastic
It seems too awesome to be true.
Last edited by Christopher; 03-26-2012 at 08:58 PM.
Makes me want to throw my 520 under a bus.Originally Posted by Christopher
Has the Media Wear Indicator moved on this one yet?
Keeps ticking away. Just how far it can go is anyones guess given the nice error free operation of the NAND
The Samsung is super sexy.
Not a bit. But this drive is really starting to grow on me. I think the interesting parts are the ones that don't show up in benches.
It's not like they're making a lot of 51nm Samsung NAND anymore, and I'm almost sad that I'm murdering such a wonderful drive.Keeps ticking away. Just how far it can go is anyones guess given the nice error free operation of the NAND
Almost.
I would like to see a SF on 100% incompressible data. I can think of a few things that might happen off the top of my head.
But I probably have enough drives going for the time being. There are some older drives in desperate need of testing, and I for one would very much like to see a Indilinx on 34nm Intel (and the Incompressible data on a SF).
But I think 3 concurrent drives is about my limit, so it's going to have to be someone else for a while. Unless I change my mind.
Last edited by Christopher; 03-27-2012 at 03:19 AM.
No, the velox v30 would be for me to test. That way I'd be at 3 drives as well.
No, I get that. I was just ruminating on the possibilities. I have that 120GB Chronos Deluxe I've been threatening to test for some time. But I think the next SF I do is gonna be on incompressible data. I think that's where the interesting stuff happens.
But I suspect that it would get VERY slow after a few hours.
OCZ Octane 128 Day 4
47208 GiB
162.92 MB/s Average
82.43 hours
Samsung 830 256 Day 7
178439 GiB
288.10 MB/s Average
176 hours
7891 Wear Leveling Count
MWI 78
0/0 Erase/Program Fail
OCZ Vertex Turbo 64 Day 28
200138 GiB
84.22 MB/s Avg
679 hours
3586 Avg PE count
20 MWI
0/0/0 Read/Program/Erase Fail
------------------------------------------------------
I sure do heart the 830. And the Turbo.
The Octane is doing well I suppose -- it's average hasn't really moved since it started 683 loops ago.
Here is some funky temp data from the 830:
Last edited by Christopher; 03-27-2012 at 05:47 AM.
Christopher, I found a couple of posts over at OCZ that reported SMART values. It seems that only attribute 5, 12 & 233 update. You have had no movement in 233 at all? With ~47TB of writes it’s starting to look like the drive does not have LTT.
ModelNumber : OCZ-PETROL
Serial Number : OCZ-W693SQFYKX9O7N38
WWN : 5-e8-3a-97 e2a6e73c7
Revision: 18
Attributes List
1: SSD Raw Read Error Rate 7
3: SSD Spin Up Time 0
4: SSD Start Stop Count 0
5: SSD Reallocated Sector Count 169
9: SSD Power-On Hours 0
12: SSD Power Cycle Count 20
232: SSD Total Count of Write Sectors 160
233: SSD Remaining Life (%) 59
ModelNumber : OCZ-PETROL
Serial Number : OCZ-2819HISWE49RZ6VF
WWN : 5-e8-3a-97 e1c601ac3
Revision: 18
Attributes List
1: SSD Raw Read Error Rate 7
3: SSD Spin Up Time 0
4: SSD Start Stop Count 0
5: SSD Reallocated Sector Count 9
9: SSD Power-On Hours 0
12: SSD Power Cycle Count 43
232: SSD Total Count of Write Sectors 160
233: SSD Remaining Life (%) 97
Todays update:
Intel X25-M G1 80GB
443,0081 TiB
21893 hours
Reallocated sectors : 05 to 07
MWI=80 to 79
MD5 =OK
44.60 MiB/s on avg
m4
589.5449 TiB
2558 hours
Avg speed 73.02 MiB/s.
AD gone from 17 to 14.
P/E 10274.
MD5 OK.
Reallocated sectors : 00
1: AMD FX-8150-Sabertooth 990FX-8GB Corsair XMS3-C300 256GB-Gainward GTX 570-HX-750
2: Phenom II X6 1100T-Asus M4A89TD Pro/usb3-8GB Corsair Dominator-Gainward GTX 460SE/-X25-V 40GB-(Crucial m4 64GB /Intel X25-M G1 80GB/X25-E 64GB/Mtron 7025/Vertex 1 donated to endurance testing)
3: Asus U31JG - X25-M G2 160GB
Ao1,
The Octane doesn't move anything but POH and Power Cycles on my drive. I thought there was something wrong at first, but I don't think there is.
I went back and looked at some posts involving lower MWI, and they were all non-Octanes. But that doesn't have anything to do with it, IMHO.
Like the first Vertex Turbo I ran, MWI is being governed by reallocations and not host writes/PE cycles. Why they've chosen that route is unclear, but I imagine it had to do with people asking questions about terrible MWI on the forums. Notice how the answer to lower MWI on the forums is never "Oh, you have lots of bad blocks."
But I swear it to be true.
Last edited by Christopher; 03-27-2012 at 02:17 PM.
Yeah, seems strange to have linked to the life of the drive, but not measuring how many cycles are used.
The SSD must store this information, otherwise it wouldn't be able to properly wear level.
yup. Its just not exposed in smart and has no bearing on mwi.
One user over @ OCZ reported that relocated sectors had been wiped clean after a f/w update, which would appear quite irregular, but regardless the ratio of relocated sectors reported from various sources is not consistent with the predicted Remaining Life and in any case is nothing to do with the remaining P/E cycle count.
The Octane reports a logical and physical sector size as 512 bytes, which is what most SSD’s report for some strange reason.
According to the SMART read out below there are 250069680 user addressable sectors. X by 512 = 128035676160 bytes or 119.24GiB.
As the minimum addressable size is 4K or 8k depending on NAND geometry the minimum that a relocated sector can represent at the NAND level is 4K or 8K.
For NAND with a 4K page size the approx. values below can be used: (EDIT: Values taken from a Micron product sheet)
• 1 block = 256 4K pages = 1048576 bytes (1MiB)
• 1 plane= 2,028 blocks = 2076672 bytes (1.98MiB)
• 1 LUN = 2 planes = 4,153,344 bytes (3.96 MIB)
1 Relocated sector could = 1 Page (4K), 1 block (1MiB), 1 plane (1.98MiB) or 1 LUN (3.96MiB) depending on how a bad sector is sectioned off.
So, if there are 169 relocated sectors that could represent 0.66 MiB, 169 MiB, 334.62 MiB or 669.24 MiB.
If there is 8 GiB of reserve NAND, 669.24 MiB would = ~ 8% reduction. If you doubled the page size = 16%. In the first example I posted (#3894) there are 169 sectors = a 41% reduction.
ATA device, with non-removable media
Model Number: OCZ-OCTANE
Serial Number: OCZ-I379591E684V26SL
Firmware Revision: E0882540
Standards: Supported: 8 7 6 5. Likely used: 8
Configuration:
Logical max current
cylinders 16383 16383
heads 16 16
sectors/track 63 63
CHS current addressable sectors: 16514064
LBA user addressable sectors: 250069680
LBA48 user addressable sectors: 250069680
Logical Sector size: 512 bytes
Physical Sector size: 512 bytes
device size with M = 1024*1024: 122104 MBytes
device size with M = 1000*1000: 128035 MBytes (128 GB)
cache/buffer size = unknown
ModelNumber : OCZ-PETROL
Serial Number : OCZ-N6V39742RQS8IJ7Z
WWN : 5-e8-3a-97 e47f1a2b8
Revision: 18
Attributes List
1: SSD Raw Read Error Rate 7
3: SSD Spin Up Time 0
4: SSD Start Stop Count 0
5: SSD Reallocated Sector Count 95
9: SSD Power-On Hours 0
12: SSD Power Cycle Count 26
232: SSD Total Count of Write Sectors 160
233: SSD Remaining Life (%) 91
Last edited by Ao1; 03-28-2012 at 02:20 AM.
Kingston SSDNow 40GB (X25-V)
825.06TB Host writes
Reallocated sectors : 05 36 // 4 up
Available Reserved Space : E8 99
POH 7407
MD5 OK
33.88MiB/s on avg (~49 hours)
-
Hardware:
I'm sorry, but I think I'm going to have to laugh for a minute.
Hey Ao1
Maybe I'm going about this wrong. Maybe the Octane needs a power cycle, like the first 830.
UPDATE
Oh my God. The Octane's MWI doesn't change when powered off, but the POH sure does.
Powering off the Octane resets Power On Hours. WTF??
Before
After
Notice the attribute in the first shot that has a raw value of 110? That's POH.
In the second shot it's back to 0.
WTF!
Last edited by Christopher; 03-28-2012 at 09:36 AM.
Wondering how long this Sandforce drive will last this time around ...
So I tried swapping the Octane into my SATA II ICH8M laptop... and it benched faster in seq writes. But after 40 minutes of endurance loops, speed dropped into the 20MB/s range. So something is going on. TRIM is enabled and working, but it almost seems like there is no TRIM.
And it was only doing 118MB/s max when I first started it anyway.
The Octane clearly hates my lappy.
Last edited by Christopher; 03-28-2012 at 02:00 PM.
Maybe if you rub off the OCZ writing on the controller, It will say Phillson ... like my Torqx 2, the Octane looks like it can't count hours or go faster then 20meg per second :P
This entertainment makes me feel a bit better about being ripped on the velox v30. BTW, the velox has the fixed firmware, sandforce version 3.3.2.
Back in the endurance rig, the Octane is doing half the average it was before. I put it in my laptop for an hour, and the speeds went down to 20MB/s. It's doing an average of 80MB/s at the moment, so it will either get better, or I'll have to SE it.
Had I just left the drive alone and not moved it to the laptop, it would still hum along as it was before (I believe).
And whatever the opinion of the Octane, it's surely better than a Phison. I liked it for the few hours I used it, but that's mostly because of the good feelings I have for the Intel 510. And the 510 and the Octane are just about the same.
Of course it is a better drive ... I was just having a little bit of fun.
In any case, the SMART values on the Octane are a write-off. I wouldn't expect any help from those figures in the future.
As for the intel 510, I'd love to give that drive a good endurance testing, but I'd find better uses for a 120gig Intel 510 before I could bring myself to sacrifice one.
I swill swear upon everything good and holy in this world that Intel designed the 510 for me because it seriously feels faster somehow. It feels faster than the Plextor M3P and faster than the 256GB 830. But if you just look at the numbers... It looks pedestrian. It should be seriously outclassed. But subjectively it feels blazing fast.
I don't take the SSD benchmarks too seriously. As long as it has good read/write latency and decent sustained sequential write performance, the SSD won't often be a bottleneck in a system. Even small 60gig SSDs with relatively poor sequential writes don't do too badly either most of the time, and only see any slowness when copying large quantities of data to them.
I'm not at all surprised the Intel 510 works as well as any other SSD despite poor benchmarks.
So I secure erased the Octane -- well, actually I just updated the FW again, which is just about the same thing. Speed went back to normal. It looks as though TRIM wasn't working on my laptop even though it was supposed to be, and that was the problem. When I put it back in the e-rig speed was slowly getting better, but still not good after 4 hours.
But I performed a little experiment.
I copied some data -- 18.6GB of my static data -- from the Samsung and Vertex Turbo while they were still doing loops. The Vertex Turbo's instantaneous speeds are ~94MB/s though the whole loop. While copying the 18.6GB of static data, instant speeds dropped into the mid-50mB/s range and I could only copy to the Octane at ~78MB/s.
The 830, on the other hand, saw no drop in instant speed when reading the 18.6GB folder to the drive. And it copied to the Octane at about 170MB/s.
Pretty awesome if you ask me.
Yeah, my Intel 520 sees very little drop in instant write performance during surface scans and data copying. (I throw the odd HDD Scan at the SSDs to keep them on their toes!)
Sad about the Octane's poor TRIMless performance though, it was supposed to have pretty decent garbage collection inbuilt.
Reply With Quote
Bookmarks