Wednesday, May 4, 2016

Mushkin ECO3 120GB Review (MKNSSDE3120GB)

The OS drive for my HTPC is ancient. It's an over 13 year old IDE Western Digital 80GB HDD (WD800JB to be exact). Connected to an AMD E-350 board (the one from this post) via an IDE to SATA bridge, it was a quick and dirty solution while I looked for a permanent drive. With the system usually resuming from sleep and 8GB of memory, files need by the OS were usually cached where they could be quickly retrieved.

I've definitely got my money's worth from that HDD purchase, and since the system is usually used for streaming, I haven't really needed anything faster or larger. The only real complaint is the noticeable sound of the older drive's motor and seeking heads.

Since cheap TLC based SSDs are now available for just over $30, I figured I'd take the plunge. Also since reviews for the lower end TLC drives are a lot less common, I'll try to correct that.

I settled on the Mushkin ECO3 120GB. The price was right and the specs were competitive for this segment. Yes I could have doubled the storage for $20 more, or moved to an MLC based drive, but price was king here. I'm replacing a working solution for an already low-power system.

The actual drive is of solid metal construction - it actually has some significant weight to it. Not more so then your average HDD, but definitely more then some of the plastic-enclosed SSDs with small PCBs. Looking at the screw locations on doing my best to peak through the mounting holes shows what appears to be a full size PCB. For warranty reasons I did not dissemble the drive (yet).

The SATA and power connector use the actual drive PCB for the connector, there's no soldered on connector. I’d imagine this was done to cut costs, and seems functional, but may cause issues with frequent connect/disconnect cycles. Likely not an issue.

I could not find any published flash memory endurance numbers (TBW).

From what I can find online, the details are:
Controller: Silicon Motion SM2256
DRAM Cache: 128MB
NAND Flash: TLC (Possibly SandDisk)

There is no dashboard or other software available from Mushkin.




Formatted capacity is 111 GB (120,031,539,200 bytes)



Crystal Disk Info shows a good number of SMART attributes. Firmware version is 01126A.

My drive reported a temp of 11C for the duration of all tests. With ambient being around 21C, either the temp sensor is faulty in my unit, or multiple software tools can't read the temp correctly.

Tests were performed on an AMD 970 chipset (SB950 southbridge with SATA3 6Gb/s ports). A newer platform could have marginally better performance figures.

First order of business is determining the size of the "SLC" cache. Most cheap TLC drives utilize a subset of their TLC flash and write to it at a rate of 1 bit/cell (versus the normal 3 bit/cell) for a performance speed-up. It's assumed that later when things are idle, the SSD will move this data to TLC memory. Writing a lot of data to an empty drive that has been idle for a bit is hopefully a decent method of determining this, but there's some things going on auto-magically behind the scenes - so it's still an estimate.


HD Tune shows the delta in write speed between this SLC cache and the regular TLC memory. The cache appears to be around 2.4 GB in size. The first 2.4 GB is written at what looks to be just over 300 MB/s, peaking at 377 MB/s. Once this is exhausted, write speed will vary between about 53 MB/s and 75 MB/s (averaging closer to 75, with dips to 53).


Text output from CDM w/ IOPS
Sequential Read (Q= 32,T= 1) :   553.452 MB/s
Sequential Write (Q= 32,T= 1) :   451.892 MB/s
Random Read 4KiB (Q= 32,T= 1) :   277.322 MB/s [ 67705.6 IOPS]
Random Write 4KiB (Q= 32,T= 1) :   241.179 MB/s [ 58881.6 IOPS]
Sequential Read (T= 1) :   471.884 MB/s
Sequential Write (T= 1) :   408.721 MB/s
Random Read 4KiB (Q= 1,T= 1) :    24.943 MB/s [  6089.6 IOPS]
Random Write 4KiB (Q= 1,T= 1) :    51.998 MB/s [ 12694.8 IOPS]




Overall I think this drive will meet my use case well. The only time I expect to overflow the SLC cache would be during OS install, some major software install, or the like. Even then, the source of this data has to be taken into consideration. Short of another SSD, this system won't ever see data throughput of 400+ MB/s outside of benchmarking. It also appears that the TLC write speed (after the SLC cache has been saturated) is still faster then the old HDD's best-case-scenario transfer rate. When you add in access time that's an order of magnitude faster, silent operation, and reduced power usage, it becomes clear that it was the right time to do this upgrade.

Tuesday, March 1, 2016

Dell "FOR DELL INTERNAL USE" Password Protected Software

I have a Dell Inspiron 3147 which has an interesting item listed in the "Downloads and Drivers" section of the Dell support page:

INSPIRON 3147/3148/3152/3153/3157/3158 G-SENSOR CALIBRATION TOOL(FOR DELL INTERNAL USE)
This package provides Dell G-sensor calibration internally and is supported on Inspiron 3147/3148 that is running the following Operating Systems: Windows 8.1(64bit).
Link: http://www.dell.com/support/home/us/en/19/Drivers/DriversDetails?driverId=1MGNJ

So what's actually interesting is that this software is provided publicly and is contained in a password protected zip file.

(Just want to take a moment and say - the average user of this system does not need this software, I have no idea what it actually does. It's likely there so support can have customers acquire and run it when their issue meets a specific set of criteria. Also, if you break something because you ran it, that's your fault not Dell's.)

Anyways, I dumped the hashes and fed them into John the Ripper (community enhanced "Jumbo" release has support for zip file).

$ ./john ~/encrypted/gsensor.hashes 
Using default input encoding: UTF-8
Loaded 1 password hash (PKZIP [32/64])
Will run 4 OpenMP threads
Press 'q' or Ctrl-C to abort, almost any other key for status
0g 0:06:41:48  3/3 0g/s 18284Kp/s 18284Kc/s 18284KC/s hrrlek0e
0g 0:07:29:56  3/3 0g/s 18130Kp/s 18130Kc/s 18130KC/s sumspy752*..sumskres9a
0g 0:09:38:53  3/3 0g/s 17980Kp/s 17980Kc/s 17980KC/s 1823adors4..1823adysac
0g 0:12:22:09  3/3 0g/s 17700Kp/s 17700Kc/s 17700KC/s 10932sho1*..10932spyon
0g 0:12:40:15  3/3 0g/s 17654Kp/s 17654Kc/s 17654KC/s 03433750458..03434741048
0g 0:13:21:40  3/3 0g/s 17532Kp/s 17532Kc/s 17532KC/s tutiairayes..tutiaimurla
0g 0:14:12:37  3/3 0g/s 17421Kp/s 17421Kc/s 17421KC/s lemrodm118..lemras061a
0g 0:15:01:52  3/3 0g/s 17302Kp/s 17302Kc/s 17302KC/s julk4tu00..julk46mb3
0g 1:05:22:06  3/3 0g/s 18348Kp/s 18348Kc/s 18348KC/s hg,bsp14s..hg,bh kuz
0g 1:08:09:16  3/3 0g/s 18343Kp/s 18343Kc/s 18343KC/s syarcr2ab..syarf0r61
0g 1:08:40:48  3/3 0g/s 18324Kp/s 18324Kc/s 18324KC/s zzjy@sk..zz42OF5
0g 1:13:25:46  3/3 0g/s 18142Kp/s 18142Kc/s 18142KC/s lk2hv''92..lk2hv/rs5
breakfix         (gsensor.zip)
1g 3:08:27:17 DONE 3/3 (2016-02-28 11:35) 0.000003g/s 18723Kp/s 18723Kc/s 18723KC/s bree3xyh..breal5fs
Use the "--show" option to display all of the cracked passwords reliably
Session completed
$

I assumed it would be something easy/simple that support could give an end-user over the phone, but still let JTR run with the default charset just in case. Leaving it to run over the weekend - the password is:
breakfix
Kind of a let-down. My Google-fu couldn't find any other software that had similar strings in Dell's support portal. Please comment if you find others. Also interesting that I can't find any examples of this password anywhere.

Curious if this is a one-time deal, or a common password used across multiple packages to keep the average user from accidentally running something.

I will add that the zip file contains some screenshots of the software running as well as a PDF instructions:



The 2nd pic shows what appears to be a snazzy Sensors Self Test Utility (SST).