4Kn HDD finally arriving in 2012 according to IDEMA

[taltamir]

Every month or so since the horrid 512e (4k drives that emulate 512 drives) drives came out I have been looking for a native implementation of 4K sectors (4Kn). And I finally found mention of actual 4Kn drives:
http://www.idema.org/?page_id=2148
Well it's about damn time! I was almost ready to surrender and get some 512e drives but now that I can see the deadline I am rejuvenated with new resolve to wait it out.
I am amazed they are taking so long to bring such drives to market, I see no reason for such delays.

If anyone has more concrete dates then "2012" or additional sources please share.

 

[Idontcare]

talta, I know you are the right person for me to ask this question - what does it matter? 512e vs. 4Kn? Is 512e leaving performance on the table that 4Kn will deliver? (like SATA II vs. SATA III or USB2 vs. USB3)

When I was able to start buying terabytes of 7200rpm drives for less than $60/TB I pretty much gave up caring about following the technological intricacies of spindle-drives.

I'm sure some sexy stuff is afoot in the industry, it just doesn't rise to the front page on just about every review site out there nowadays if it isn't an SSD.

 

[jvroig]

I'd also like to know about this.

I thought as long as you used any recent Windows/Linux/Mac OS, all your 512e drives will be detected as 4K and work as 4K and not suffer any performance penalty at all?

Do you mean to say it isn't like that, and due to being 512e, they somehow still lose performance even when used in 4K-aware OSes?

 

[taltamir]

512n = Native 512 sector drives.
512e = 4K drives that emulate 512 sectors.
4Kn = 4K drives that present the OS 4K sectors.

Some drives that are 512e are completely indistinguishable than 512n. Some supposedly do somehow advertise they are really 4K drives in disguise according to some recent info I read. Identification is not reliable and AFAIK only BSD/linux can even make said unreliable identification.

It does cause performance loss. Alignment is an issue, even if you align the sectors correctly you still lose performance unless you emulate 4K sectors. I am not quite clear on whether that is due to read-modify-write cycles, OS trying to access two emulated sectors in the same real sector at once, or both issues.

Several tools can emulate 4k sectors in BSD / Linux (only, no windows or MacOS; AFAIK no solaris either) but those bring bugs, compatibility issues, and are hell to configure. Those tools also cause some very serious issues if used in conjunction with ZFS which is the only reliable FS out there right now.

Having layers that convert back and forth adds small but unnecessary delays due to conversion and also adds more places where you could have a bug or error, so the firmware reliability is a concern for such HDDs (there are more complex things for there to be an error in) as well as the reliability of the OS tools like the 4k re-emulation (which are in fact known to have various issues and bugs).

Finally, moving forwards we can expect more detection and automatic emulation to occur which brings compatibility issues with said emulation schemes even more severely into the spotlight. I am told that linux will now identify some drives and automatically emulate 4K sectors on 512e drives for UFS. This can cause serious issues down the line.

4Kn drives aren't going to be magical and will naturally have issues with software that assumes 512 sectors. But it would be really simple and direct to overcome. The whole 512e thing brings to mind the "smart" phone landline network vs the "dumb" (by design) internet. Or the old saying

 

[razel]

Don't we already have 4k native drives in use? Wasn't 2011 the year that all new HDDs manufactured must be native 4k. Did they mean 'should' instead of 'must'? With my real world experience, I'm getting drives in 2011 laptops that seems to be native 4k with 512e available via jumper. That's what was running through my mind just before I read the article. Then while reading, I can across 4k native drives in 'high performance server environments'.

 

[taltamir]

Don't we already have 4k native drives in use? Wasn't 2011 the year that all new HDDs manufactured must be native 4k. Did they mean 'should' instead of 'must'? With my real world experience, I'm getting drives in 2011 laptops that seems to be native 4k with 512e available via jumper. That's what was running through my mind just before I read the article. Then while reading, I can across 4k native drives in 'high performance server environments'.

no, every single drive out there right now is 512e. You cannot buy a 4Kn drive at the moment.

The jumper is a sector shift. If you have the jumper on then it adds a 1 to each sector number, so if the OS tries to access sector 331 it will get sector 332 instead. This mitigates one aspect of the performance loss on windows XP and other legacy OS. However, this is just one of several issues that cause performance loss with said drives.

 

[Mark R]

3TB USB drives have been 4kn (at least that's how they appear to the OS).

This is because 4kn drives have better XP compatability - they work in XP (but are not bootable and can't be formatted with XP's built in format tools), whereas 3 TB 512n/e drives are limited to 2 TB

 

[taltamir]

3TB USB drives have been 4kn (at least that's how they appear to the OS).

This is because 4kn drives have better XP compatability - they work in XP (but are not bootable and can't be formatted with XP's built in format tools), whereas 3 TB 512n/e drives are limited to 2 TB

I am pretty sure that is a 64bit 512B controller emulating a 32bit 4K controller. It is emulating 4K sectors for a 512e drive... so the drive is 4K internally, the drive emulates 512B, the controller then re-emulates 4K for it. That is not 4Kn.

512e drives actually need a 64bit controller to address more then 2TB. 32bit controller means 4,294,967,296 addressable locations. 4,294,967,296 * 512B = 2TB.
If it was a 4Kn drive it would be able to address 8 times that amount, 16TB, on a 32bit controller. To actually use a 64bit controller to address more then 32bits you need something more advanced then MBR though, GPT will do it but XP has no GPT support at all, so it can't really be a simple 64bit controller displaying 512b sectors.

If you physically remove the drive from the external enclosure you would find a standard internal drive. if you plug this in it would be 512e.

At least, AFAIK that is the case. I was unable to find hard evidence either way in regards to said external drives.
You can easily prove it one way or another. Take an external 3TB drive, take it apart, plug it to a 64bit controller. It will show up as either 512B (512e) or 4KB (4Kn)

 

[KingFatty]

As a user, how would the improvement be felt? For example, would it reduce boot-up time from 33 seconds to 32.85 seconds? Would photoshop take 7.99 seconds to load, instead of 8 seconds?

 

[Silenus]

taltamir already mentioned but for me it will finally make ZFS nas/san builds way easier to get properly aligned! That is too say we won't have to do anything at all, ZFS will properly align everything once is see a natively exposed 4k drive. That will be much nicer going into the future as you can then count on any future drives being added to your pools as all native 4k instead of worrying about weird mixtures of 512n and 512e drives like we have now. It will be simpler!