When do you think we will see Hard drive manufacturers move to 5.25" form factor?

[cbn]

Thinking more about the effect of Helium described back in post #35 reducing disk flutter and thus allowing thinner (and thus potentially less rigid?) platters to be used.

I just have to wonder if this allows the regular (thicker/more rigid?) 3.5" (3.74") platters to make a comeback at 10,000 (or greater) rpm?

Certainly helium would lower power requirements (at any given rpm) compared to air due to having a lower density (and therefore less resistance). But the main point I am wonder about is if thicker platters at 10,000 rpm (or some high level of high rpm) have the same amount of flutter as thinner platters at a lower rpm (eg, 7200 rpm).

See post #42 for mention of the effect of rigidity on platters operating at high spindle speed.

Enhanced Rigidity: The rigidity of a platter refers to how stiff it is. Stiff platters are more resistant to shock and vibration, and are better-suited for being mated with higher-speed spindles and other high-performance hardware. Reducing the hard disk platter's diameter by a factor of two approximately quadruples its rigidity.

 

[cbn]

I bet nitrogen would work just as well. The reason air sucks vs pure anything else is because various elements behave differently as the temperatures increase or decrease. Sticking with one element makes things more predictable.

Helium is about 1/7th as dense as either Air or Nitrogen though:

http://www.engineeringtoolbox.com/gas-density-d_158.html

 

[Essence_of_War]

Helium is about 1/7th as dense as either Air or Nitrogen though:

http://www.engineeringtoolbox.com/gas-density-d_158.html

Helium is 1/7th the atomic mass as molecular nitrogen, but it isn't necessarily obvious to me that I'd fill a hard drive to the same pressure, or operate at the same temperature, if I was using one or the other. Without knowing more about the gas-fill/design parameters, I'm not sure if density is the right metric to use.

 

[StrangerGuy]

Probably because it's intert. Of the noble gases, Helium is probably the easiest to obtain.

Patently false. Argon is by far the easiest and cheapest, yet its still waaaaaay more expensive than nitrogen. Helium is so expensive that college labs will not use it unless its absolutely necessary like for NMR spectrometers.

 

[Essence_of_War]

Patently false. Argon is by far the easiest and cheapest, yet its still waaaaaay more expensive than nitrogen. Helium is so expensive that college labs will not use it unless its absolutely necessary like for NMR spectrometers.

I think you're talking about liquid Helium and Argon vs. liquid Nitrogen, aren't you?

It is easy and cheap to get helium gas for making helium balloons.

 

[cbn]

The thing is though, data is written to the edge of the platters first and the end of the drive is actually at the center. They do this because drive reading speeds are slowest towards the center of the disk.

Yes.

The faster-rpm drives have the added benefit of having lower latency for the entire disk.

You can also just short-stroke slower drives. That makes them about as fast as the next step up. But it's additional admin overhead, and wastes a bunch of disk space. (Unless you partition the rest as "slow" storage. But now when you request that data, you're hurting your high speed performance.)

That is interesting.

So this short-stroking is done simply be reducing partition size? (Then my disk only reads from the outer most part of the platter to somewhere in the middle? This depending on how large or small I make my partition?)

 

[cbn]

Smaller platters = shorter strokes = shorter access times = higher IOPS.

With Helium, I just wonder if they could use a different strategy:

One of which would be to use a larger diameter platter, but increase the size of the spindle. (Increasing the size of the spindle would effectively short stroke the platter. I believe the larger spindle would also increase the rigidity of the platter that is left remaining. The Helium then allows higher rpms at the same energy usage as a small platter).

So basically the effect would be to go from a small platter with short stroke to a large platter with short stroke, but higher rpms.

 

[Essence_of_War]

So this short-stroking is done simply be reducing partition size? (Then my disk only reads from the outer most part of the platter to somewhere in the middle? This depending on how large or small I make my partition?)

http://www.tomshardware.co.uk/short-stroking-hdd,review-31527.html

Short stroking aims to minimize performance-eating head repositioning delays by reducing the number of tracks used per hard drive. In a simple example, a terabyte hard drive (1,000 GB) may be based on three platters with 333 GB storage capacity each. If we were to use only 10% of the storage medium, starting with the outer sectors of the drive (which provide the best performance), the hard drive would have to deal with significantly fewer head movements.


The result of short stroking is always significantly reduced capacity. In this example, the terabyte drive would be limited to 33 GB per platter and hence only offer a total capacity of 100 GB. But the result should be noticeably shorter access times and much improved I/O performance, as the drive can operate with a minimum amount of physical activity.
l

 

[cbn]

http://www.tomshardware.co.uk/short-stroking-hdd,review-31527.html

That is a very good explanation, and thanks for posting that.

I know in some cases hard drives are short stroked from the factory (Eg, 500 GB hard drive using two 333 GB platters, etc).

What I am would like to do is make my own short stroke hard drive. This by reducing partition size?

 

[Essence_of_War]

That is a very good explanation, and thanks for posting that.

I know in some cases hard drives are short stroked from the factory (Eg, 500 GB hard drive using two 333 GB platters, etc).

What I am would like to do is make my own short stroke hard drive. This by reducing partition size?

Yeah, if you skim the whole link, they go over their set-up for the short-stroking also, but that is the gist of it.

If you are trying to play with it a little, I'd use HDTune to find where the performance starts to decline, and then use gparted or windows disk management to make a single smaller partiton near the "front" of the disk. To run benchmarks, a short-stroke partition that is 5-10% of the disk's total capacity should give you a noticeable (at least in benchmarks) performance bump.

 

[cbn]

And a 400+ IOPS HDD is... probably impossible. 5.25" or otherwise.

I would think a 5.25" short stroked should be able to handle that.

P.S. A Short stroked Hitachi Ultrastar 15K450, 450 GB, SAS/300 (from 2009) could break 500 IOPS:

http://www.tomshardware.co.uk/short-stroking-hdd,review-31527-7.html

To match a 2U enclosure of 24x 2TB 2.5" HDDs pulling 2-4w each, you'd have to have 6x 5.25" HDDs, which would have to have >4 times the capacity, <4x the power draw, <4x the per unit cost, and >4x the IOPS, to be in any way a convincing argument.

I am wondering if a larger (than normal) diameter hub motor in a 5.25" HDD (basically short stroking the 5.25" platter) is able to enhance performance per watt compared to four smaller HDD motors each with classic 2.5" platters.

So each 5.25" drive with a larger than normal diameter hub motor replaces four 2.5" Hard disk drives.

 

[ksec]

Rather then seeing 5.25" HDD return ( moving up ) it is likely we will see 3.5" disappeared ( Moving down ).

Within a few years time most PC / Laptop will not have a 2.5" HDD / SSD. It will all be M.2 SSD. The Server Industry are already moving to 2.5" for even Mass Storage due to its size / capacity / power usage. For Amazon, Apple, Google, Facebook and Microsoft as well as other Super Large Cloud Services Provider ( They have a different TCO compared to others ) they may continue to buy 3.5". But 3.5" as a consumer drives will likely fade.

So, I dont see any reason why 5.25" will come back.

 

[Denithor]

I wonder if anyone looked at running the disks in a vacuum instead of helium atmosphere. Would seem like under very slight vacuum there would be zero "air" resistance, unless I'm missing something here? In order to run a helium atmosphere they are already making the platter chamber a fully sealed environment, so why not just suck out the air instead of replacing with inert gas?

 

[cbn]

The Server Industry are already moving to 2.5" for even Mass Storage due to its size / capacity / power usage.

Totally agree with you on that.

And Dave the Nerd did a good explanation on why that is in post #16.

But since 2.5" drive are more expensive per GB than 3.5", surely the industry must be working on an alternative.

My guess is that 3.5" HDD with larger diameter hub would be the next step. This would short stroke the 3.5" platters so they would have greater IOPS.

Furthermore, since the inside edge to outside edge of the platter is now shorter, the effect would be increased stiffness. Increased stiffness (As I crudely understand the situation) now means the platter can be spun faster. And alternative to spinning faster would be to keep RPM the same and make the platters thinner....with more platters per 3.5" drive.

 

[dave_the_nerd]

$/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.

 

[fuzzymath10]

I'm not convinced such a device can be made in an economically sound matter. It seems like most aspects of addressing the performance problem of spinners or the cost problem of SSDs will result in a product that has mediocre capacity compared to a mainstream spinner drive, still be way slower than an SSD in iops, while consuming more power because of high rpms or platter count, while weighing more, and being noisier. Not sure such a product would be successful at any price let alone the cost plus margin and indirect costs or fixed expenses.

 

[cbn]

I'm not convinced such a device can be made in an economically sound matter. It seems like most aspects of addressing the performance problem of spinners or the cost problem of SSDs will result in a product that has mediocre capacity compared to a mainstream spinner drive, still be way slower than an SSD in iops, while consuming more power because of high rpms or platter count, while weighing more, and being noisier. Not sure such a product would be successful at any price let alone the cost plus margin and indirect costs or fixed expenses.

Whatever drive gets produced doesn't have to compete with an SSD for IOPs, it just has to be better than an equivalent number of 2.5" drives in the various metrics already mentioned in this thread (which includes TCO).

With that mentioned, Helium filled 3.5" drives with 7 platters already exist and they are dense (even if they had PMR, capacity would still be 8.4TB)

The next step, I would imagine would be to strike a balance between electric hub motor diameter to the degree of short stroking needed on the 3.5" platter.

Greater diameter on the hub motor = greater amount short stroking on the 3.5" platter.

Also it might be that the performance per watt at higher rpms (using 3.5" platter) is better with a larger diameter hub motor compared to smaller diameter hub motor.

The downside, though, is reduced capacity depending on how large the hub motor is. (But just realize we are starting at a capacity of 8.4TB using PMR).

 

[VirtualLarry]

I wonder if anyone looked at running the disks in a vacuum instead of helium atmosphere. Would seem like under very slight vacuum there would be zero "air" resistance, unless I'm missing something here? In order to run a helium atmosphere they are already making the platter chamber a fully sealed environment, so why not just suck out the air instead of replacing with inert gas?

But drives "need" an atmosphere. The "air" is what provides a cushion for the heads to fly over the platter. If they didn't have that, there would be head crashes. A vacuum wouldn't work.

 

[Essence_of_War]

Totally agree with you on that.

And Dave the Nerd did a good explanation on why that is in post #16.

But since 2.5" drive are more expensive per GB than 3.5", surely the industry must be working on an alternative.

Whoa there, you've made a big leap and a big assumption: that the enterprise hardware market is cheifly concerned about $/GB. This is not the whole story. The enterprise market cares chiefly about, as dave_the_nerd pointed out, total cost of ownership which include other things like IOPS/Watt (every bit of heat has to be offset by air conditioning) and IOPS / space-unit (because you can't scale-out into space you don't have).

10k/15k 2.5'' drives use about half the power of similar 3.5'' drives, so even if I give up a bit of capacity, I can double my IOPS in the same power profile. Additionally, you can't get more than 4x 3.5'' drives in 1U of rack space, but SuperMicro will be happy to sell you 8x and 10x hot-swap chasis for 2.5'' drives. I'm doubling my IOPS in the same space/power profile by switching to 2.5'' drives even if I give up a bit of capacity. This actually gets even better if you have bigger rack units, you can get 24x 2.5'' drives in a 2U chasis, but you're basically stuck with 8x 3.5'' drives in the same space.

 

[cbn]

Whoa there, you've made a big leap and a big assumption: that the enterprise hardware market is cheifly concerned about $/GB. This is not the whole story.

Yeah, I can see how the TCO is very important.

I think the trick will be how to make the larger drives act more like (or even exceed the performance of) multiple small drives (IOPs, density etc), while being cheaper to make in the bargain.

This is I brought up the short stroking and the larger diameter hub motor in the rest of post #64 and post #67.

 

[Essence_of_War]

I think the trick will be how to make the larger drives act more like (or even exceed the performance of) multiple small drives (IOPs, density etc), while being cheaper to make in the bargain.

This is I brought up the short stroking and larger diameter hub motor in the rest of post #64 and post #67.

I don't think this is as easy as you seem to think, and it's not obvious to me that it's possible.

Short-stroking is fun for messing around with benchmarks, but it requires you throw use 1/10th of your drive's capacity out for a IOPS improvement of maybe 2x-3x, maybe 5x being very charitable. That pushes your cost/GB into SSD range, and your 500 IOPS are now competing with an SSD that delivers 50,000-100,000 IOPS, produces no vibrations, and uses less power. There is also a subtle issue, restricting usage to the outermost 10% means that 10% is getting significantly more IO than it would be if the whole physical disk were in use. I don't know what the numbers look like for drive lifetime after short stroking, but I wouldn't be surprised if mean lifetime were reduced, squeezing your TCO from another angle.

 

[dave_the_nerd]

This actually gets even better if you have bigger rack units, you can get 24x 2.5'' drives in a 2U chasis, but you're basically stuck with 8x 3.5'' drives in the same space.

Well, 12.

There are also those high-density enclosures around that store like 80+ 3.5" drives in a 5U, but they weigh a ton and are hard to hot-swap. They're also kinda intended cooling-wise for slower drives as mass/archival storage.

 

[cbn]

Short-stroking is fun for messing around with benchmarks, but it requires you throw use 1/10th of your drive's capacity out for a IOPS improvement of maybe 2x-3x, maybe 5x being very charitable. That pushes your cost/GB into SSD range, and your 500 IOPS are now competing with an SSD that delivers 50,000-100,000 IOPS, produces no vibrations, and uses less power. There is also a subtle issue, restricting usage to the outermost 10% means that 10% is getting significantly more IO than it would be if the whole physical disk were in use. I don't know what the numbers look like for drive lifetime after short stroking, but I wouldn't be surprised if mean lifetime were reduced, squeezing your TCO from another angle.

Remember though that this short stroking I am referring to is accompanied by a larger diameter hub motor. So that should mean the rpms can increase (re: stronger motor and stiffer platter) or perhaps the performance per watt (at the same rpm) could improve a bit.

With that mentioned, one very important thing to consider is the power consumption of one larger diameter platter at X rpm vs. multiple smaller diameter platters at X rpm.

If the larger diameter platter has 2x the circumference of the smaller diameter platter that means the edge of the larger diameter platter will be moving at 2x the velocity of the smaller platter for any given rpm. But 2x the velocity is not 2x the power consumption, it will be more than that.

However, with that mentioned, remember this larger diameter platter will be spinning in Helium so there will be power consumption offset (depending on how fast the platter is set to go). I guess the argument could be made that Helium could be filled in smaller drives too, but I wonder if this just more economical to do it with the larger drives (maybe I am wrong).

 

[dave_the_nerd]

Performance Art!

 

[Lorne]

Helium is better then nitrogen, Its due to the atomic mass difference.
Using Helium brings the heads closer and also with less mass between the heads and the platters there is less (How to say in laymans) magnetic splatter and can read a finer or denser cylinder. (Like holding a flashlight and moving close and father from its target the flare sharpens)

All lighter elements have advantages and disadvantages to use.

O2, Straight up condensed is denser then the open atmosphere which is the opposite of what is needed and of course extreme oxidation.

N, Great element to use and abundant, Also great for decreasing oxidation and moisture, But cost affectiveness and to product improvement is pretty much nill.
Good possibility product life could double.

C, Or CO, Though lighter has its own problems, Notorious moisture condensation and oxidizer, Major pressure changes temperature change.

LI, BE, B, Are solids.

That leaves H and HE.

H, Is abundant, Works great, Major safety issues and costly, Same reasons you don't see a lot of H powered cars at this time.