TR Forums

I'm only aware of the HGST 5K1500, which is 1.5TB on presumably 3 platters.
Any idea why other than this model everyone produces, at 9.5mm, only 1TB PMR and 2TB SMR?

The above is a 2013 model. If not 3-platter drives, surely 750TB/platter should be possible nowadays, considering SMRs are 1TB/platter?

I don't have an answer to your question, but I suspect that 2.5" HDD development has come to a near standstill since their primary use case (laptops) is going away. The other significant use case (compact high-capacity external backup) isn't as performance sensitive, so SMR drives are taking over that niche.

I think those 2TB SMRs aren't necessarily external-oriented. And for that there's also 15mm drives.

Too bad that 5K1500 is hard to find.

Beware. Many of Seagate's 2½" 15mm drives are infected with shingles.

P.S.: You might scan the 36 pages of this discussion in the TiVo forums as folks search for non-SMR 2½" drives.

meerkt wrote:

Too bad that 5K1500 is hard to find.

Why is PMR so desirable? Other than better performance, is it because it has lower rates of failure? You could get the 2TB Firecuda SSHD. The NAND cache should extend the drive's life considerably by shielding it from the normal wear and tear of everyday repetitive computing as well as giving better performance than plain HDD.

Igor_Kavinski wrote:

meerkt wrote:

Too bad that 5K1500 is hard to find.

Why is PMR so desirable? Other than better performance, is it because it has lower rates of failure? You could get the 2TB Firecuda SSHD. The NAND cache should extend the drive's life considerably by shielding it from the normal wear and tear of everyday repetitive computing as well as giving better performance than plain HDD.

The problem isn't so much that they tend to perform worse in general; it's that there are certain corner-case workloads that can cause throughput to tank so badly that the drive appears to be bad. We're talking about I/O latencies on the order of 60 seconds or more. If you don't hit any of these bad corner cases, they'll perform adequately for non-performance-critical or read-dominated workloads.

About a year and a half ago I made the mistake of using some SMR drives (Seagate Barracuda Compute) in my new NAS build. The SMR drives ended up falling out of the RAID array due to I/O timeouts. They weren't actually defective; they were just running so slowly that the RAID software was deciding the drives had failed.

Also, a nit-pick... I realize that a lot of people refer to non-SMR drives as PMR, but this is technically incorrect. SMR drives still use PMR internally; the difference is that the tracks overlap, not that we've ditched PMR. IOW all current SMR drives are in fact also PMR. Whenever this comes up, I urge people to use the term "non-SMR" or "CMR" (where the "C" stands for "conventional") to distinguish between the drive types.

just brew it! wrote:

Also, a nit-pick... I realize that a lot of people refer to non-SMR drives as PMR, but this is technically incorrect. SMR drives still use PMR internally; the difference is that the tracks overlap, not that we've ditched PMR. IOW all current SMR drives are in fact also PMR. Whenever this comes up, I urge people to use the term "non-SMR" or "CMR" (where the "C" stands for "conventional") to distinguish between the drive types.

Good point. I didn't know that. Thanks!

I mistakenly bought a 2½" Seagate ST3000LM016 drive to upgrade a TiVo Bolt. This was one of the models where Seagate obfuscated the use of SMR. After a few hours or a few days of use, the crappy SMR drive would fail so badly that it would corrupt the system files and brick the TiVo.

Don't let your next hard-drive be infected with shingles.

JustAnEngineer wrote:

Many of Seagate's 2½" 15mm drives are infected with shingles.

Being >9.5mm is also a problem.

Igor_Kavinski wrote:

get the 2TB Firecuda SSHD

I'm conflicted about these hybrids. The theory sounds... sound. But besides the unknown performance, I'm mainly worried about long-term reliability.
What are the specs of the flash they use? What happens if the flash fails but the platters are still okay?

They seem to come with 5-year warranties, which perhaps bodes well.
But after the stupid end-of-life behavior of the SSD drives in the TR long-term torture study, I wouldn't be surprised if in SSHDs bad flash = dead drive.

just brew it! wrote:

I urge people to use the term "non-SMR" or "CMR"

I suppose that makes sense. I wasn't sure if CMR is an "official" term, so went with PMR.

We tried using SSHDs at work a while back, as server boot drives. We ran into lots of trouble because they were unusually sensitive to vibration. Probably a design flaw with just that particular model (i.e. nothing inherent to SSHD tech in general), but my limited experience with them was not a good one.

just brew it! wrote:

We tried using SSHDs at work a while back, as server boot drives. We ran into lots of trouble because they were unusually sensitive to vibration.

How did you determine that vibration was the troublesome factor? Were the SSHD's the source of the vibration due to being in a RAID array? If so, how many per array? Were the drives later re-used for some other purpose? If so, were they satisfactory or did the vibration issue damage them permanently somehow? I know, too many questions!

meerkt wrote:

Igor_Kavinski wrote:

get the 2TB Firecuda SSHD

I'm conflicted about these hybrids. The theory sounds... sound. But besides the unknown performance, I'm mainly worried about long-term reliability.
What are the specs of the flash they use? What happens if the flash fails but the platters are still okay?

They seem to come with 5-year warranties, which perhaps bodes well.
But after the stupid end-of-life behavior of the SSD drives in the TR long-term torture study, I wouldn't be surprised if in SSHDs bad flash = dead drive.

https://www.techspot.com/community/topi ... ar.255495/

Guess all Seagate cares about is shipping large quantity of drives rather than engineering reliable ones. Makes sense. More failures equal more sales. Especially in cases where the laptop/PC is still under warranty. I doubt PC manufacturers blacklist Seagate and put in a WD drive as replacement to prevent future distress to their customer due to repeat failure.

Igor_Kavinski wrote:

Guess all Seagate cares about is shipping large quantity of drives rather than engineering reliable ones.

Their reliability in the past 5-10 years is no worse than any of the other manufacturers (within reason).

Igor_Kavinski wrote:

How did you determine that vibration was the troublesome factor? Were the SSHD's the source of the vibration due to being in a RAID array? If so, how many per array? Were the drives later re-used for some other purpose? If so, were they satisfactory or did the vibration issue damage them permanently somehow? I know, too many questions!

The SSHDs were used singly, or in RAID-1. They were boot devices not data storage drives, but the servers they were in did have large RAID arrays in them. The vibration issue was diagnosed via internal testing, and the drives did exhibit a tendency to fail prematurely when used in that environment.

We've also seen issues with other (normal, non-SSHD) drives, where particular makes/models will have a tendency to report huge numbers of "G shock" events in their SMART data. Presumably this will eventually hit the threshold for declaring a SMART failure based on that counter (though I have not personally seen one that did that). In this case it is possible that the issue is merely a mis-calibrated accelerometer and the drive is actually fine; if so, the drives will continue to function normally, but the sysadmin's management console will light up like a Christmas tree with alerts when the drives start to go over SMART threshold...

Waco wrote:

Igor_Kavinski wrote:

Guess all Seagate cares about is shipping large quantity of drives rather than engineering reliable ones.

Their reliability in the past 5-10 years is no worse than any of the other manufacturers (within reason).

Yeah, every vendor occasionally has a problematic model.

That said... and take this with a grain of salt... it does seem to me that over the past ~15 years, HGST has had a somewhat better track record compared to Seagate and WD. (But that came on the heels of the 75GXP "Deathstar" fiasco, which ranks up there with the Capacitor Plague as one of the biggest tech fails of the '00s. The Deskstar brand got such a bad reputation that I was rather surprised Hitachi opted to continue using the name after they acquired IBM's HDD division.)

The 2.5" Firecuda's warranty length does suggest Seagate expects it to be more reliable than average.

Every manufacturer has had a few bad models, but in recent years it seems Seagate has more of them.
There's also a difference in failure modes. I still kept on using a Dethstar, once I figured out what's going on. It involved occasional work, but for non-critical stuff it was better than nothing.

Any particular reason you wouldn't use a SSD with a decent ram cache for this?

A 2TB SSD would cost 3-4 times as much.

meerkt wrote:

A 2TB SSD would cost 3-4 times as much.

https://www.amazon.com/Seagate-FireCuda ... 07H2F3741/

Seeing how cheap it is and assuming you want to put it in your laptop, you could get two and put one in an external enclosure and keep it synced with your working one so in case of catastrophic failure, pull out the old one, put in the backup SSHD and you are back in business.

meerkt wrote:

Every manufacturer has had a few bad models, but in recent years it seems Seagate has more of them.

Recent being a decade ago or so?

The recent Seagates were in the Backblaze stats.

Igor: Too much uncertainty.

Waco wrote:

meerkt wrote:

Every manufacturer has had a few bad models, but in recent years it seems Seagate has more of them.

Recent being a decade ago or so?

I thought the 3TB issue they got sued over was more recent than that (as in, roughly 2013 timeframe)?

Edit: Ahh, I guess the drive in question (ST3000DM001) was in fact released in 2011, so yes nearly a decade ago. I guess the issue did not start to get a lot of publicity until a few years after that. Lawsuit came in 2016.

My ST3000DM001 drives were purchased 11/28/2013, 4/14/2013 and 11/23/2012. Oddly enough, all are still working. Since those, I've purchased the crappy SMR ST3000LM016 and five IronWolf NAS drives.

I guess I'm less sensitive to it too. I don't trust ANY storage so I take appropriate steps to make recovery easy if it comes to that.

JustAnEngineer wrote:

My ST3000DM001 drives were purchased 11/28/2013, 4/14/2013 and 11/23/2012. Oddly enough, all are still working. Since those, I've purchased the crappy SMR ST3000LM016 and five IronWolf NAS drives.

Maybe they got the issue sorted out eventually? Or are these very lightly used drives?

JustAnEngineer wrote:

My ST3000DM001 drives were purchased 11/28/2013, 4/14/2013 and 11/23/2012. Oddly enough, all are still working.

Try using the HDD health option in Speedfan. It will compare the SMART values of your drives to all identical drives in its database and provide you with interesting information about any values that are not within acceptable range.

Besides the HGST 5K1500 there's also the Samsung Spinpoint M9T.
2013-2014 vintage (Samsung-Seagate transition era), with capacities: 1.5TB, 2TB, and... 1.75TB.

https://www.anandtech.com/show/15442/en ... r-qlc-ssds

I'm gonna go with this if they can price the 2TB QLC SSD <= 2x the price of a 2TB HDD.

By the way, not bad for the price if you can live with half the capacity: https://www.amazon.com/dp/B07Y3YH9RM/

This 750GB SSHD has SLC cache so should be a lot more reliable: https://www.amazon.com/dp/B07YNKSFHF/

This 1TB SSD may be using 32-layer Intel NAND, as per one customer review: https://www.amazon.com/dp/B073QN8KL4/

Igor_Kavinski wrote:

https://www.anandtech.com/show/15442/enmotus-midrive-rethinking-slc-caching-for-qlc-ssds

I'm gonna go with this if they can price the 2TB QLC SSD <= 2x the price of a 2TB HDD.

Don't hold your breath, 2TB QLC drives are currently at around 4x. OTOH, the cost of all the mechanical bits does put a lower bound on how much a HDD can sell for, so we should get there eventually, at least at these lower (4TB and below) capacity points.

TBH I think HDDs below ~8TB capacity will become increasingly rare in the near future, as consumer use of HDDs asymptotically approaches zero.

The market's even worse now. Even at 1TB it's becoming hard to find CMR.

1+ TB non-SMR:

WD: Only Red.
Seagate: None.
Toshiba: None.

Most use cases that historically called for 2.5" drives would be better served by SSDs these days, so people aren't willing to pay much for 2.5" HDDs any more. Vendors are trying to eke out profits on what has become a low-margin item, and SMR is one way to help with that. I expect 2.5" HDDs to disappear entirely within the next few years.

All of the new development in HDD tech is for high-capacity 3.5" drives aimed at enterprise and Cloud customers, since that's where the demand is now.