In this day and age, SSDs are becoming quite ubiquitous for their performance benefits and many systems come with them pre-installed. But not too many years back, SSDs were considered performance items for enthusiasts, commanding high price-tags. Hobbyists like myself even experimented with CompactFlash cards to (at times) underwhelming results depending on the controller’s I/O behaviour and whether they supported hard drive operation rather than removable disk operation, and various UDMA modes rather than the performance-cripping PIO modes.
Given this fact, I found it highly improbable that I would salvage an SSD from a discarded system for at least another few years. Even the oldest serious SSDs perform pretty decently, as compared to hard drives. Those who couldn’t afford them at the time, like myself, instead worked with Western Digital Raptors (36Gb and 150Gb models, which I owned and benched a long time back).
So, when I picked up what appeared to be a water damaged laptop from next to a rubbish bin and started taking it apart, I joked to a friend that “I bet I won’t find an SSD in this bay,” and as I ripped it out, I proceeded to see the Intel logo. Intel doesn’t make hard drives … and so immediately, I laughed in disbelief, having literally contradicted myself in a matter of seconds. Merry Christmas to me!
Intel X25-M (Postville)
In the earlier days of SATA SSDs, we actually have a lot to thank Intel for. Their X25 and later lower-cost X25-M SSDs paved the way for making SSDs affordable to the enthusiast while offering the performance and reliability that people expect. Of course, Intel did have a little egg on their face, being an SSD manufacturer with a widely publicised firmware issues such as the X25-M update brick bug and the 320-series 8Mb brick bug. This was just the beginning of a series of confidence-shaking bugs like the OCZ Vertex 3/Corsair Force GT/Sandforce BSOD issue, Crucial m4 5,000 hour BSOD bug, Intel/Sandforce encryption bug, Samsung 840 EVO throughput bug and various TRIM implementation inconsistencies.
Anyway, all of that diversion aside, this was an Intel X25-M 34nm MLC drive with 160Gb capacity, a SATA II interface, and an HP OEM part number of SSDSA2M160G2HP.
The drive is 7mm in height and is closed together with regular Philips head screws.
The front and back are pretty plain, but the date code on the inside of the lid suggests that the unit is about 6 years old. Time flies so quickly …
The top side of the PCB has a few components, but no flash populated.
The underside has ten flash chips in a ten channel architecture – something not even attempted by most modern controllers due to the sheer number of connections. This is understandable, since flash interfaces had more limited throughput at the time. All of the flash is Intel branded, along with their own controller and an ISSI cache RAM and ST EEPROM.
As received, the drive was partitioned with a boot and data partition – the fact the boot partition is so full probably caused its owner a lot of grief, and part of the reason it may have been given the boot a lot earlier than it should have.
Condition As Received
The SMART data suggests the unit only had 806 hours of use, which is virtually “brand new”. It did see about 5.7Tb of writes, which is a fair amount for the time. The SSD only gives limited information in regards to its state, but the MWI suggests 98% of its life remaining. The firmware was 2CV102HA, which was not the latest, so it probably should be updated to avoid any future trouble and maintain best performance and compatibility.
For most Intel SSDs, this means that you need to use the Intel SSD Toolbox. While I was in there, I decided to do a quick read test after clearing the drive which succeeded just fine …
Unfortunately, this is where the OEM part number makes our life difficult.
Intel SSD Toolbox does not allow us to update the firmware nor secure erase the drive. This is quite an annoyance since I know the latest firmware is 2CV102M3 – I just needed to get it somehow.
As a result, I did temporarily give up and test the drive with its existing firmware, but after that, I went and tried my best to do the update and re-bench the drive, allowing us to compare the firmwares.
The site now works, but at the time I was trying to do the update, DNS resolution for that server name failed entirely.
I ultimately had to trudge through their FTP and found a link to the download there which succeeded. Trying to follow the update procedure having installed the drive as a secondary device resulted in this annoying message – thanks HP:
The model was indeed supported, so I had to invoke an administrative command prompt, unpack the SoftPaq files, and then invoke the updater manually which worked just fine.
This should have been a lot easier … if it were not for the silly OEM support arrangement limitations.
In the following tests, the results obtained with the older 2CV102HA firmware will be posted on the left, with the results from the newer 2CV102M3 firmware posted on the right.
The change in firmware did not produce any new SMART attributes for monitoring, but appears to have altered the MWI behaviour which has increased to 99 from a value of 98 despite having 1Gb more written to the drive. This may be a refinement in the life estimation of the firmware, as has happened also with my OCZ Vertex 3 in the past.
HD Tune Pro
Sequential read performance averaged 257.4MB/s under the old firmware and 259.6MB/s under the new firmware. This is about the performance you might expect from a pretty saturated SATA II link.
Sequential write performance averaged 101.9MB/s under the old firmware and 99.5MB/s under the new firmware. This isn’t close to saturating the SATA II link, but is right on the 100MB/s write speed specification. A fairly decent result, considering the age of the unit and the limitations of the ONFi 1.0 NAND interface and controller.
Random access read and write IOPS appeared to decline slightly under the new firmware.
A similar result was seen under the File Benchmark mode.
Extra tests were also performed, although their significance is limited.
ATTO shows that performance consistency on the read seems to be better on the new firmware, with some write variations coming through which weren’t so obvious in the older firmware. A slight boost at 4kB read performance is seen, although the numbers for <4kB accesses seem to be a bit rocky under the old firmware, and is more in line with expectations under the new firmware.
CDM seems to suggest that the performance in sequential and 512kB accesses has not changed markedly between firmwares, but the 4kB writes and queued performance has changed to favour reads over the writes, with a fairly sizable fall in 4kB QD32 write performance.
AS SSD Benchmark
In AS SSD, the drive improved its read score, at the expense of the write score with sequential writes seemingly most impacted. The read performance under 4k threaded conditions improved, which offset this fall, resulting in a total score that is better after the update.
The Anvil score fell with the newer firmware slightly, mainly due to slightly less write performance, although the read performance did improve slightly which offset the fall somewhat. On the overall score, the drive had a higher score prior to the update.
No data integrity errors occurred, and the results are so close, with a slight edge in both reads and writes to the newer firmware.
I didn’t think I’d ever salvage an SSD in this particular day and age, but as it goes to show, almost anything could and does happen. This particular 160Gb SSD is no longer the “fastest” kid on the block, but it’s no major slouch either, and is likely to fare pretty similarly as the lower end “value” line SSD today. It’s definitely one of the best salvages I’ve come across to date. The firmware update did seem to change the performance slightly – it seems that better read performance is favoured over writes, but only very slightly. To me, this SSD is an important part of the history of moving from HDD-only based machines to ones including SSDs, offering speed which really makes the Western Digital 10,000rpm Raptor drives that came 4-7 years before it (36Gb, 2003 and 150Gb, 2006) look quite slow by comparison.