To all my readers – a Merry Christmas and Happy New Year to you all, or alternatively, Happy Holidays or Merry *mas … It’s that time of the year again where we’re counting down to the end and people are spending time away from work with their families, possibly experiencing the joys of exchanging gifts. That being said, this silly season is also the time for abandonment. Often when you hear about that, you might picture a pet which becomes quickly neglected after the initial joy wears off. But no, this post is all about tech abandonment.
What happens to the old hardware when someone gets a gift of something new, possibly equivalent in nature? Maybe it’s a lack of space, or the complete lack of attachment with the equipment itself, or no desire to keep a spare around, but inevitably, it’s likely that the old hardware will be discarded in some way. Which means ’tis the season to be salvaged. As they say, one man’s trash is another man’s treasure … and you can’t beat a gift that’s free! And they say Christmas is being overly commercialized …
Rather luckily for me, around this time was the time for a council clean-up. This is an event that is held approximately every three months where householders are invited to put bulky rubbish on the kerbside strip for collection on a nominated date, but because of potential cost-cutting, they skipped one event and this was the one right after it, meaning the yields were much better. Add to this an e-waste pickup happening around the same time at a place I once frequented fairly often, results in a rather interesting chance to stock up on unwanted tech to play with over the break.
The first posting of this series deals with hard drives – spinning platters of rust where people store their computer programs and documents; mechanical and prone to unreliability. These devices grow obsolete very quickly, as their storage capacities are quickly eclipsed and interfaces occasionally change.
However, while they are likely to be items that are frequently discarded, they are also items that are fairly rare to find owing to the potentially sensitive nature of the data stored upon them. Identity theft is a big issue – letting a hard drive with your own personal documents “sit on the kerb” is a highly unwise idea. As some people know of this, many of the drives are frequently destroyed instead, resulting in a glut of machines with no hard drives to run from.
Rather surprisingly, I did manage to find a few hard drives and was also given a few directly. They aren’t particularly capacious by today’s standards, but they’re not entirely worthless. Instead of taking the time to peer through the data on the drives (which would be evil), I decided to wipe the data and re-certify the drives for use, to see how well the hardware has stood the test of time and doing a few benchmarks as a way of documenting the progress in storage performance over time.
3.5″ SATA Drives
These are common drives today, and I managed to salvage two, both out of pre-built machines.
Seagate 320Gb Barracuda 7200.11
Seagate drives have a bit of a mixed reliability in my experience, so I wasn’t expecting any miracles from this drive which was an HP OEM drive. I’ve had a pair of their 250Gb models fail in little more than a year with various ailments, of a similar “slimline” build as well.
As expected, the drive wasn’t entirely “happy” but it wasn’t in major trouble either. Then again, it is still fairly young in terms of power-on hours, reaching only about one year of power-on time. The major warning seems to be with the high-fly-writes parameter, which has counted up to 121 events – I suspect this is an indication of problems with the head heater which is used to regulate the flying height of the head, where high fly writes may be problematic for reliability of written data or cause delays as the drive retries the write. Other than that, the other vital signs look good, with no reallocations.
As the drive had an OS restore partition, I didn’t want to wipe the drive entirely just yet, so I only did read-tests with HD Tune.
I deleted most of the data in the data partition and did an H2testW against that slice of the drive, which completed without errors. A look at the post-test vital signs seems to show that there was some high-fly writes that occurred during the testing.
This continues the tradition of Seagate drives of the vintage having some odd behaviour and potentially shaky reliability.
Samsung 500Gb HD502IJ
I had salvaged some of these before, but as it turns out, there are still a few hanging around. On the whole, these drives have a very good reliability reputation in my experience, so I wasn’t expecting anything different.
The drive had done a decent 23,544 hours, with no reallocations and otherwise healthy vital signs.
It averaged a 92Mb/s read and 90.4Mb/s write, fairly decent for a drive of that density and vintage. There aren’t any major hiccups in the graph that would imply anything shaky about the drive, which is nice.
Random access has never been a major strong point of mechanical hard drives, only maxing out at 134 IOPS. SSDs really dominate when it comes to these access patterns because they don’t have the mechanical limitations that hard drives do.
Some extra tests were done as well, and the effect of the cache buffer can be seen especially on the read test.
It’s a bit of a pity that Samsung was bought out – while they weren’t a fully integrated manufacturer and relied on sourcing components from various companies, they managed to make competitive drives which have been quite reliable. It’s not to say I haven’t experienced a failure of a Samsung HDD, but it’s more-so the fact that Samsung HDDs seem to fail very infrequently and manage to often live to even 40k+ hours without perfect vital signs.
2.5″ IDE Drives
These drives were extracted from old laptops by a friend of mine who had saved them for me in case I wanted them. I did some final cleaning up of the drives by removing them from their proprietary carriers, but some people might be wondering why I’d even bother. The most desirable 2.5″ IDE drives are typically the larger (80+Gb) units, and even then, they are increasingly fragile and better-substituted by large UDMA-capable CompactFlash cards. The ones I had in my hand were 30Gb or less …
We can already see from the labels, the majority of the fleet are IBM (Hitachi) Travelstar drives of 4200RPM and capacities of 18-40Gb. There is a Hitachi branded 30Gb drive of their own design and a Toshiba 20Gb drive. All are 4200rpm drives, rather than more performant 5400rpm or 7200rpm models and are OEM supplied drives included in laptops.
As it turns out, I wanted the drives because there might be a use for them in the future for reviving old laptops in an “orthodox” way. Drives less than 32Gb have a place where BIOS bugs, or where early Windows 98 copies are involved, as they have a support barrier there. It would also be interesting to see just how well the drives aged, and what their performance is like compared to modern storage technologies – the result of at most, 17 years of progress.
Three drives are made in the Philippines, two in Thailand and one in China.
All drives except except the DARA-218000 are 9.5mm height. That particular drive is a taller 15mm height – like those of large external storage 2.5″ drives sold today.
As I didn’t want to go to the lengths of running my older machine with native IDE ports, I settled on daisy chaining a Sunplus-based IDE to SATA adapter with an 3.5″ desktop IDE to 2.5″ laptop IDE adapter to run the drives off a SATA port on a Z97-based motherboard. This will explain why the screenshots imply these elderly drives are being connected through SATA.
The drives were tested in no particular order …
The first drive tested was the thick 18Gb drive, made in October 1999. This drive had 15,813 hours on the clock, which is a fair amount for a laptop drive, and was quite noisy with a characteristic whine due to the ball bearing motor in use. The drive did have some SMART issues as received with 3 reallocation event counts, and 1 pending sector which was cleared with the surface testing. However, the fact that reallocation events occurred with no sectors actually reallocated suggests there are weak sectors on the surface which may cause data loss in the future.
The drive achieved a read speed averaging 10.4MB/s and a write speed averaging 8.8MB/s. Access times were quite erratic for the writes, and fairly high on the read. Maybe this drive is showing its age to some extent. That being said, the performance could easily be bested by a quality modern microSD card.
However, when it comes to ATTO, it’s clear the drive’s small buffer is well optimized, and the drive delivers full I/O performance by 8kB accesses, with fairly close to full performance at 4kB. Some flash drives won’t even be able to match this today.
This drive came healthy with a fairly low 4,497 hours of usage. One UDMA CRC Error was recorded, possibly a case of interference, bad cabling or connections, but the vital signs are all good.
However, for a drive with good vital signs, its write speed graph seems to show very erratic behaviour. I’m not sure if it’s a timing compatibility thing with the IDE to SATA bridge, or some crazy adaptive-density scheme gone wrong due to a weak head, but it varies quite strongly even though the read performance is much more stable. Read averaged 13.5MB/s, whereas write only averaged 9.7MB/s. This does however, represent performance progress as the density is higher being a thin drive with a larger capacity than the last. It does have a larger buffer as well.
ATTO did seem somewhat confused by the performance of the drive, especially the writes, but 4kB read performance seems to reach the peak performance of the drive.
This drive was even more healthy than the last – good vital signs and just 2111 hours of use. This is a 20Gb drive, offering as much storage as the last drive in the same form factor.
Despite the same spindle speed, capacity and form factor, its performance pushes new boundaries with an average 16.4MB/s read and 15.9MB/s write. Disappointing by today’s standards, and a key reason why laptops often felt sluggish. It also seems to employ a faster UDMA interface than the last with burst rates suggesting UltraATA100 on this drive rather than the 66 on the previous ones.
ATTO shows full performance is reached on reads at about 8Kb accesses, and writes at 64kB accesses, but the small block performance is not too bad.
This is a 40Gb model of the same family as the one before – healthy as well with 4,003 hours of use, making it young.
Despite the improved capacity, the performance remains virtually identical.
The ATTO results have a similar profile but the individual numbers differ slightly.
We met this drive earlier, but there was a second unit with only 1,135 hours on the clock. Unfortunately, this one had one pending sector and two reallocations event counts. After testing, the pending sector was cleared, but the event count increased by one, suggesting this drive has a weak sector.
This unit was slightly faster, but it wasn’t completely co-operative, having a read error during a read-back test at the weak sector which was only resolved after rewriting it again. Performance was again comparable across the board to the previous two drives.
This drive was fairly sick sounding – it clicked loudly, it make whining noises and buzzing noises as if it had issues finding sectors and the stats show it. A total of 16,334 hours operating, with a whopping 925 reallocated sectors and 833 reallocation events. It even had load-unload retries, and read error retry rate issues in the past. Its load-unload cycle count value suggests it has served too long in a condition that was conducive to the heads parking and unparking frequently – based on the volume label, I am led to believe this drive was used in a USB enclosure.
A drive with such shaky vital signs is almost never going to complete benchmarks without causing problems. This drive proved to be the exception. Clear dips are visible where the drive had to seek to replacement sectors, but no read errors or write errors occurred during testing, and an H2testW run (not shown) did not reveal any integrity issues. Average read speed was 18.5MB/s and average write was 18.2MB/s, although slightly higher rates would be expected from a healthy drive.
The results from ATTO are probably not that reliable, but full read performance is shown by 4kB accesses, and write at 16kB.
This drive has a modest 8,538 hours on the clock, but sadly, has one reallocated sector, so isn’t quite in “tip top” condition.
Quite possibly being the most recent drive in the batch, it achieved an average 18.4MB/s read and 18.2MB/s write.
ATTO seems to show some very strange results, namely a performance that’s virtually independent of I/O size. I wonder if this is a fancy cache trick that the firmware is playing with ATTO to game the benchmark … as this is not really the expected result.
In all, the IDE laptop hard drives were a mixed bunch for age and health. That being said, while we can see performance improvements over the manufacturing dates covered by the drives, the performance is an underwhelming <25MB/s which actually means that even a decent USB 2.0 connection could offer all the sequential performance available from the drives and I didn’t need to go to the lengths of a SATA-IDE bridge. Of course, later and larger drives performed even better, but it’s an interesting peek to see how far we’ve come.
Hard drives are often rare to find discarded, but I managed to come across a few and test them. None of them were totally dead, although a few weren’t really that healthy. The Samsung drives seem to hold up fairly well, whereas the Seagate and Hitachi/IBM drives are a bit of a mixed bunch. Performance wise, the SATA 3.5″ drives are still worthwhile even for cold store or booting an older system, whereas the IDE 2.5″ drives are mainly of historical interest.