Seagate s/ATA disks

[DiTBho]

so I bought qty=16 discs, 8 old gen, 8 new gen, and the video tells exactly the surprise I got

the old gen are faster than the new-gen 

[alm]

Is it about SMR vs CMR drives? SMR drives suck for intense workloads involving lots of non-sequential writes like typical server applications. Fortunately these days it's pretty well documented which drives are SMR and CMR, for example here for Seagate.

[DiTBho]

exactly the same disks as shown in the video 

[Kjelt]

The Barracuda is SMR. That has been introduces about 6 years ago or something like that.
So your old hdds are probably from 6 years back or NOS ?
So if you want CMR you need the Barracuda Pro.
Also the salesprice is decreased dramatically which means the BOM should also drop significantly. They do that with SMR (less platters per capacity) but also with less cache RAM.

Nowadays I only buy enterprise datacenter drives. Much cheaper in €/GB and much better as in high MTBF, reliability and built quality (better bearings etc). They make more noise and run on higher speeds but I dont care about that. My WD Golds have 60000+ hours and not one error or bad sector.

[SiliconWizard]

Haven't bought a HDD in a while now, so I don't know what is currently the best performing ones.
Was always wary of Seagate stuff in the past, had very bad experiences with them, but it was a long time ago. I know the problems they had 20 years ago are long gone, but I'm still wary for probably no good reason.

Are the IronWolf series a safer bet compared to Barracuda?

[Kjelt]

You can not say that in general this is better than, because situations vary.
You quickly end up in a flame war just as discussing Windows vs Mac vs Linux.

I sometimes look at backblaze statistics to get an impression

https://www.backblaze.com/blog/backblaze-drive-stats-for-2022/

It is a decent indicator since the population is large, but they only use datacenter disks.
Personally I prefer WD Golds and nowadays the cheaper WD Ultrastar DC (previously HGST)
For Seagate that would be the Exos series.
They make more noise though so not always suitable for desktop usage.

Most important is the hdds run cool , steady no shocks and preferrably 24/7 then they last the longest.
But since energy prices went through the roof I unfortunately also shutbdown three of my five NASses when I dont need them. And one of the other two now only has ssd's.

[SiliconWizard]

Yes what can really kill a HDD is the number of power cycles. In particular, the number of head parking, which is sometimes worse.

I had drives from the infamous WD Green series at some point in my NAS (that I've replaced since) and had 2 of them fail (that resulted in no data loss as it was in a RAID array though). One specificity of the Green series is that they were constantly parking their head after short timeouts (to get in ultra-low power state) and it wasn't configurable, it was hard-coded in the firmware, so unless the drives were accessed 24/7 no stop, that would be pretty bad. That had killed thousands of these drives in the world. WD stopped selling them after a while.

From the list you posted, it kinda looks like the Seagate have the worst failurer rates overall, although it's a bit biased as they have used more of them than the other brands.

[BradC]

I had drives from the infamous WD Green series at some point in my NAS (that I've replaced since) and had 2 of them fail (that resulted in no data loss as it was in a RAID array though). One specificity of the Green series is that they were constantly parking their head after short timeouts (to get in ultra-low power state) and it wasn't configurable, it was hard-coded in the firmware, so unless the drives were accessed 24/7 no stop, that would be pretty bad. That had killed thousands of these drives in the world.

The head parking could be configured or disabled completely using hdparm or wdidle3.exe. I still have 7 of them in a RAID here.

WD stopped selling them after a while.

Because they consolidated their product lines, not because they weren't fit for purpose for their intended use.

[Ed.Kloonk]

I owned a couple of the early WD greens with the auto spin down. They really were awful. And the only sub group of drives I can honestly say seemed determined to commit hare kari.

[DiTBho]

So your old hdds are probably from 6 years back or NOS ?

yes, New Old Stock

Nowadays I only buy enterprise datacenter drives.

Which model? SAS? 

[Monkeh]

I owned a couple of the early WD greens with the auto spin down. They really were awful. And the only sub group of drives I can honestly say seemed determined to commit hare kari.

I just retired an array of them after 12 years of runtime. Two failures.

Around the same time I bought a handful of Seagate drives. One lasted a week, total three failures per drive.

The old greens were nowhere near as bad as hearsay will have it. The same basic drives, complete with unreported head unloads just to shut people up (not that I like lying on diagnostic data) went on for many years after with no major issues. Stop writing file access times to your drives with 5s commit intervals.

[DiTBho]

The Barracuda is SMR.

Yup, the Barracuda ST8000DM004 is SMR, not CMR. Seagate did NOT disclose.

[DiTBho]

Seagate - 3.5"   ST2000DM005   DM-SMR
Seagate - 3.5"   ST2000DM008   DM-SMR
Seagate - 3.5"   ST3000DM007   DM-SMR
Seagate - 3.5"   ST4000DM004   DM-SMR

[DiTBho]

Seagate Ironwolf ST4000VN008 4TB CMR
Seagate Ironwolf ST4000VN006 4TB CMR
WD Red Plus WD20EFZX 2TB CMR
WD Red Plus WD60EFPX 6TB CMR

[Kjelt]

Which model? SAS? 

No the SATA interface and also without the powerdown/enable pin. Look in the datasheets for the exact types.

[DiTBho]

from this document
- Barracuda > 1Tbyre are SMR
- Barracuda Pro  are CMR
- Barracuda 512GB are CMR

correct? 

[DiTBho]

so, yesterday I returned all my Seagate Barracuda 2T disks 

[golden_labels]

My WDs (Caviar Black, 1002FAEX 00Y9A0) have a bug. It seems that in the self-test log the time field overflows after 65535 hours (2¹⁶). Should they refund me, maybe?

As for the original issue: could it be a matter of Seagate doing binning and consumer batches now being drawn from what was left after removing the best performing specimen? Or Covid pandemic disrupting supply chains and Seagate having to rely on worse components? The other explanations I can see is fixing some bug in software or having to decrease performance after it was set too high initially.

[Nominal Animal]

My WDs (Caviar Black, 1002FAEX 00Y9A0) have a bug. It seems that in the self-test log the time field overflows after 65535 hours (2¹⁶). Should they refund me, maybe?

I recall Samsung spinny-rust hard drives also used only 16 bits of NVRAM for power-on hours.  I think it is a common practice, considering it's seven and a half years of continuous use.