What do you think the specs will be on the next high performance 2.5" HDD?

[cbn]

Some categories to think about:

Spindle speed: 5400 rpm, 7200 rpm, 10,000 rpm

2.5" platter capacity: 500GB, 750GB, etc

Number of platters: 2, 3, 4 (or more)

buffer size: 64MB, 128MB, 256MB

Form factor: 7mm, 9.5mm, 15mm

Helium: Yes or No

SSHD : Yes or No (If yes, how much NAND?)

 

[Essence_of_War]

Is it given that the "next" one will even exist?

 

[cbn]

Is it given that the "next" one will even exist?

For Data Centers I'm sure the 2.5" will still exist for a while longer, for laptop I'm thinking 2.5" HDD days are getting very short though.

But maybe WD (or the other OEMs) surprises us with some interesting anyway (That might also be useful for data center).

 

[Dahak]

From a consumer point, nothing new likely

From an enterpise point, most of what you listed is already here, except helium drives
no point for SSHDs in enterprise, or if there in 95% of the usage it will not make sense

 

[Insert_Nickname]

Some categories to think about:

Spindle speed: 5400 rpm, 7200 rpm, 10,000 rpm

2.5" platter capacity: 500GB, 750GB, etc

Number of platters: 2, 3, 4 (or more)

buffer size: 64MB, 128MB, 256MB

Form factor: 7mm, 9.5mm, 15mm

Helium: Yes or No

SSHD : Yes or No (If yes, how much NAND?)

"High performance" and "2.5" HDD" in the same sentence? :biggrin:

Joking aside, I don't think any major change is in store. Lowest-end will still be crappy 5400RPM drives, with a few 7200RPM drives thrown in. SSD and eMMC will take over the rest.

However, I'd like to see some 2.5" 5400RPM drives with more cache (>32MB). 32MB really should be the bare minimum offered. 8MB drives are horrible, 16MB drives are only slightly less painful. Its 2016 already, not 2006.

 

[hojnikb]

I wish every consumer HDDs sold would have some sort of ssd caching. It helps a little and it shouldnt be that much more expensive to implement.
Nand is cheap and nand controller could be simply integrated to the main hdd controller (instead of a separate controller thats used nowdays).

 

[fuzzymath10]

No laptop or desktop owner should need to bear the pain that is spinning hard drives as a primary drive. The only acceptable moving parts in laptops today should be semi-passive cooling and maybe optical drives, and those too need to disappear eventually.

 

[Cerb]

Just platter density improvements, from here on out, I think. Maybe some better SSHDs will come out.

 

[cbn]

Just platter density improvements, from here on out, I think. Maybe some better SSHDs will come out.

Density improvements are nice because even when the track is near the spindle then sequential transfer could actually still be useable.

As a baseline take the SAMSUNG Spinpoint M9T ST2000LM003 (2.5" 2TB 5400 RPM 32MB laptop Hard Drive) which according to the HD platter data base here has three 667GB platters.

Looking at the HD Tune Results in this review we can see despite the 5400 rpm it is still above 60 MB/s at the end of the run :

This, in contrast, to my old Seagate ST380815AS 3.5" 7200rpm drive (factory short stroked to 80GB from a 160GB platter) where I am only getting 35MB/s by the end of a HD tune benchmark run.

 

[Jembo]

two and five eigths

 

[cbn]

"High performance" and "2.5" HDD" in the same sentence? :biggrin:

The VelociRaptor (fourth, fifth and sixth generation Raptor) was a 2.5" drive.

P.S. Something I find very interesting is that even at 10,000 rpm a 2.5" platter has less angular velocity (ie, velocity at the edge of the platter) than a 3.5" platter (actual diameter 3.74") at 7200rpm. So in cases where platter density is the same the 3.5" drive should have higher sequential transfer rate for the outermost tracks. However, at any point equidistant to the spindle the 10,000 rpm drive wins.

 

[Puffnstuff]

Unfortunately the veloraptor didn't fit in a laptop as it was too thick. For what they cost I'd opt for a ssd any day of the week.

 

[cbn]

Unfortunately the veloraptor didn't fit in a laptop as it was too thick.

That is right.

It was 15mm thick (laptops can only take up to 9.5mm thick) and it required 12V power (laptops only have 5v to the SATA).

But I wonder if they could fix this with a Helium VelociRaptor? (re: the Helium reduces the power requirements and/or allows thinner platters)

 

[cbn]

With that mentioned, I do wonder if a 9.5mm form factor allows any extra density over 15mm for a data center.

If not, then maybe a 15mm thick Helium VelociRaptor makes more sense? Perhaps even one that goes 15,000 rpm with lower (or same) power consumption as 10,000 rpm (non-Helium) VelociRaptor.

 

[cbn]

I wish every consumer HDDs sold would have some sort of ssd caching. It helps a little and it shouldnt be that much more expensive to implement.
Nand is cheap and nand controller could be simply integrated to the main hdd controller (instead of a separate controller thats used nowdays).

That is an interesting idea (NAND controller integrated with HDD controller) and perhaps the NAND controller could even share the HDD controller's DRAM. This rather than the situation today where the SSHDs I have seen have a SSD controller which is DRAM-less.

With that mentioned, there is always another possibility to consider:

Using a separate M.2 (or 2.5" SSD) for cache (as well as operating system and some applications).

Example: 256GB SSD (with 16GB to 64GB assigned to cache) with 2.5" HDD for bulk storage in a laptop.

 

[Insert_Nickname]

The VelociRaptor (fourth, fifth and sixth generation Raptor) was a 2.5" drive.

P.S. Something I find very interesting is that even at 10,000 rpm a 2.5" platter has less angular velocity (ie, velocity at the edge of the platter) than a 3.5" platter (actual diameter 3.74") at 7200rpm. So in cases where platter density is the same the 3.5" drive should have higher sequential transfer rate for the outermost tracks. However, at any point equidistant to the spindle the 10,000 rpm drive wins.

Unfortunately the veloraptor didn't fit in a laptop as it was too thick. For what they cost I'd opt for a ssd any day of the week.

Besides being too thick, there is noise, cooling, vibration and power consumption to worry about. That's not even mentioning reliability issues.

In short 10.000RPM drives would be a spectacularly bad fit for mobile devices. There is a reason nobody has put 10.000RPM drives in laptops. That goes double or triple today since any budget SSD is going to walk all over such a drive.

Besides, sure 10.000RPM helps a bit with seek times and transfer rate, but faster spindle speed is nowhere near the "magic bullet" you seem to think. Please feel free to correct me if I'm wrong.

BTW, this is coming from someone who used to RAID raptors. When I got my first SSD, I never once looked back.

With that mentioned, I do wonder if a 9.5mm form factor allows any extra density over 15mm for a data center.

If not, then maybe a 15mm thick Helium VelociRaptor makes more sense? Perhaps even one that goes 15,000 rpm with lower (or same) power consumption as 10,000 rpm (non-Helium) VelociRaptor.

If you think 10.000RPM disks are a bad fit for mobile, 15.000RPM drives would be even worse... D:

 

[cbn]

In short 10.000RPM drives would be a spectacularly bad fit for mobile devices. There is a reason nobody has put 10.000RPM drives in laptops.

7200rpm is fastest 2.5" drive for laptop I have seen.

But what if the 2.5" drive were filled with Helium? This to reduce resistance.

Maybe 10,000 rpm is possible without needing 12V.

 

[cbn]

Besides being too thick, there is noise, cooling, vibration and power consumption to worry about. That's not even mentioning reliability issues.

In short 10.000RPM drives would be a spectacularly bad fit for mobile devices. There is a reason nobody has put 10.000RPM drives in laptops. That goes double or triple today since any budget SSD is going to walk all over such a drive.

Besides, sure 10.000RPM helps a bit with seek times and transfer rate, but faster spindle speed is nowhere near the "magic bullet" you seem to think. Please feel free to correct me if I'm wrong.

BTW, this is coming from someone who used to RAID raptors. When I got my first SSD, I never once looked back.

Which Raptors did you use?

1st, 2nd, 3rd generation? (These are the Raptors that come in the 3.5" casing)

4th, 5th, 6th generation? (These were called VelociRaptor and came in 2.5" enclosure with a 3.5" heatsink/adapter around them called the "Icepack").

 

[cbn]

With that mentioned, I do wonder if a 9.5mm form factor allows any extra density over 15mm for a data center.

If not, then maybe a 15mm thick Helium VelociRaptor makes more sense? Perhaps even one that goes 15,000 rpm with lower (or same) power consumption as 10,000 rpm (non-Helium) VelociRaptor.

If you think 10.000RPM disks are a bad fit for mobile, 15.000RPM drives would be even worse... D:

I was thinking a 15,000 rpm 15mm thick Helium Raptor would not go in a laptop (in the same way a 10,000 rpm 15mm thick air filled Raptor would not).

However, with that mentioned I do wonder what the sweet spot for IOPS, capacity and performance per watt for a 2.5" drive would be.

Would it better for a data center to use more platters for X given thickness? Or does increasing RPM yield a greater benefit?

 

[Insert_Nickname]

7200rpm is fastest 2.5" drive for laptop I have seen.

But what if the 2.5" drive were filled with Helium? This to reduce resistance.

Maybe 10,000 rpm is possible without needing 12V.

Might be possible with helium. They'd still get toasty (especially without airflow), and the vibration would make it, ehh... unpleasant..., in a laptop. And I'd still be concerned about reliability.

Which Raptors did you use?

1st, 2nd, 3rd generation? (These are the Raptors that come in the 3.5" casing)

4th, 5th, 6th generation? (These were called VelociRaptor and came in 2.5" enclosure with a 3.5" heatsink/adapter around them called the "Icepack").

Original 36GB (WD360GD), 74GB (WD740GD) and finally 300GB (WD3000GLFS).

I got my first SDD in 2009 (Intel X-25M G2 80GB), didn't bother with RAID raptors any more after that experience. When I upgraded my storage drives later, I replaced 7200RPM with 5400RPM drives. There was just no need for the additional performance of 7200RPM, and it cut down on power usage too.

I was thinking a 15,000 rpm 15mm thick Helium Raptor would not go in a laptop (in the same way a 10,000 rpm 15mm thick air filled Raptor would not).

However, with that mentioned I do wonder what the sweet spot for IOPS, capacity and performance per watt for a 2.5" drive would be.

I like to take the Crucial BX100/BX200 and Samsung 850EVO as the real sweet spot for performance/power usage/capacity(cost) right now.

Sure you can get better drives for each point, but you're compromising the two others. Any good engineer will tell you you can only get two out of three things...

Would it better for a data center to use more platters for X given thickness? Or does increasing RPM yield a greater benefit?

That's another discussion entirely. Because there are a lot more factors involved.

The really short cut-down version is that its entirely dependent on workload.

 

[cbn]

I was thinking a 15,000 rpm 15mm thick Helium Raptor would not go in a laptop (in the same way a 10,000 rpm 15mm thick air filled Raptor would not).

However, with that mentioned I do wonder what the sweet spot for IOPS, capacity and performance per watt for a 2.5" drive would be.

I like to take the Crucial BX100/BX200 and Samsung 850EVO as the real sweet spot for performance/power usage/capacity(cost) right now.

Sure you can get better drives for each point, but you're compromising the two others. Any good engineer will tell you you can only get two out of three things...

Sorry if I did not make the last part of that quote more clear.

When I mention sweet spot for IOPS, capacity and performance per watt I am referring to 2.5" HDD used in a data center.

If I am not mistaken adding too much capacity, but not increasing IOPs could lead to a drive unbalanced to the needs of multiple users.

Which makes me wonder what direction the 2.5" HDD will go in order to compete better? (ie, more platters for any given thickness vs. higher RPM).

 

[Insert_Nickname]

Sorry if I did not make the last part of that quote more clear.

When I mention sweet spot for IOPS, capacity and performance per watt I am referring to 2.5" HDD used in a data center.

If I am not mistaken adding too much capacity, but not increasing IOPs could lead to a drive unbalanced to the needs of multiple users.

Which makes me wonder what direction the 2.5" HDD will go in order to compete better? (ie, more platters for any given thickness vs. higher RPM).

2.5" HDDs are in kind of a tight spot in a data center. If you want capacity, you go with cheap(ish) 3.5" drives with as much capacity as manufactures can squeeze into them. Bulk storage doesn't require massive performance any way, and tend to be sequential rather then random R/W. On the other hand, for performance you go (PCIe) SSD.

If you want both capacity and performance, it makes much more sense to implement a two tier system, with SSD handling cache and in flight data, and 3.5" HDDs handling bulk storage.

 

[XavierMace]

I think you're making a bad assumption about the quantity of 2.5" HDD used in data centers.

 

[cbn]

2.5" HDDs are in kind of a tight spot in a data center. If you want capacity, you go with cheap(ish) 3.5" drives with as much capacity as manufactures can squeeze into them. Bulk storage doesn't require massive performance any way, and tend to be sequential rather then random R/W. On the other hand, for performance you go (PCIe) SSD.

If you want both capacity and performance, it makes much more sense to implement a two tier system, with SSD handling cache and in flight data, and 3.5" HDDs handling bulk storage.

Since nobody has responded yet to help the 3.5" vs. 2.5" HDD (for data center) understanding I will post some responses I got on the topic from my 5.25" Hard drive thread.

http://forums.anandtech.com/showpost.php?p=37635906&postcount=18

For a data center, I have to wonder how much of a move to 2.5" (from 3.5") is due to power reduction or increased density at the cost of reduced performance and higher price per GB.

It's due to reduced power consumption for a given performance level, and overall improved performance per U.

Two 2.5" give a bit more sequential performance than one 3.5" (30-50%, unless things have changed a lot recently), and around double the random performance. With a case that was made primarily for 2.5" drives, more than twice the number of 2.5" drives can fit as 3.5" could have.

Since even simple backup jobs can sometimes get stuck with random seeking that's not cached, 3.5" make less and less sense all the time, outside of uses like NAS. But, SSDs are still way too expensive for many, that still need their TBs of capacity. Meanwhile, power use will be the same or lower, depending on comparison drives (IE, a 5400 RPM NAS drive wouldn't be compared to 2x10k or 2x15k 2.5" SAS, but instead a 10k or 15k 3.5" SAS drive).

http://forums.anandtech.com/showpost.php?p=37635598&postcount=16

The reason data centers moved from 3.5" to 2.5" HDDs is because they offer higher storage density in a smaller volume or smaller amount of rackspace, and use less watts per TB in the bargain. (snip)

http://forums.anandtech.com/showpost.php?p=37653002&postcount=65

$/GB isn't the number we're concerned about, it's TCO.

Purchase price plus rackspace cost plus electrical and cooling costs, etc. From that standpoint, it's either a wash or in favor on 2.5" drives.

 

[Insert_Nickname]

Since nobody has responded yet to help the 3.5" vs. 2.5" HDD (for data center) understanding I will post some responses I got on the topic from my 5.25" Hard drive thread.

Its rather simple. When you can quite literally get twice as much performance out of a PCIe SSD as a whole rack of 2.5"/3.5" HDDs, what's going to be most cost effective and reliable?

For bulk storage 3.5" drives still make more sense then 2x 2.5", because of simple storage density. You can get 8TBs worth out of a 3.5" "slot", while 2.5" drives top out at 2TB, yielding 4TB in the same space.