So do the HDD manufacturers literally have no plan to keep offering relevant products? It just isn't plausible that there's a market for HAMR drives, the slightly lower cost-per-GB isn't worth the awful performance, density, heat, and power.
The electronics industries don't revolve around your computer. Datacenters need absurd amounts of immediately-accessible data. Facebook is not going to put every stupid post you made 10 years ago in an SSD.
The text of the post probably is... all of Wikipedia is around 150GB compressed. It's hard to comprehend, but computers could easily record and store every keystroke from the day you're born to the day you die.
@Kjella right, the big data hogs have moved from text to richer media, like photos and videos. Since a lot of people are including these in their posts, @Wereweeb's point still stands.
And, while text went through some evolution from plain ASCII to unicode + markup, it never ballooned in size the way that higher-resolution, higher-framerate, and higher-dynamic-range cameras are causing photos and videos to grow.
Yeah, but the images & other media files are already compressed and much, much bigger. According to that link you cite:
"The size of the media files in Wikimedia Commons, which includes the images, videos and other media used across all the language-specific Wikipedias was described as well over 23 TB near the end of 2014."
In the past 7 years, it's surely grown by orders of magnitude.
You have no idea what you're talking about, higher capacity to drive lower TCO-per-GB is exactly what the market (hyperscalers + large enterprise) wants.
OptiNAND is not a new name for the existing hybrid drive concept, so it's incorrect for you to assume they'll hit 30TB by simply aggregating 22TB of magnetic storage with 8TB of NAND. They can't afford to put 8TB of NAND on a single hard drive, and will actually be increasing the real areal density of the magnetic storage.
If you're talking about OptiNAND, that's not what it is.
"... they are stressing the fact that it is not a hybrid drive (SSHD). Unlike SSHDs, the OptiNAND drives do not store any user data at all during normal operation. Instead, the NAND is being used to store metadata from HDD operation in order to improve capacity, performance, and reliability."
I mean, they can say whatever they want, and it sounds like they are lying, because if the goal of OptiNAND is to increase capacity - whether through "metadata" or whatever - if its increasing capacity, user data is inherently stored there. My guess is they are using parity to cover data loss in the event of NAND block failure, damaged sectors, or power loss corruption. Who knows. They won't disclose exactly how it works.
At the end of the day what they HAVE disclosed if they are going to make 30TB drives using 22TB platters. That implies there is at least 8TB of NAND, and while that is extremely expensive, I doubt a 30TB storage product is going to cost less than $1500 when the 22TB drive is going to cost $800.
If they were lying, there would be a big board *full* of expensive NAND chips attached to the HDD, and it would be much more expensive. This isn't something you can fudge.
> if its increasing capacity, user data is inherently stored there.
Are you more interested in being stubborn or actually learning something. The article I linked is fairly clear about what sort of information it holds. I'm not going to spoon-feed you more quotes, when you can just go read it, yourself.
> My guess is they are using parity to cover data loss
Disks have already done that for *decades*!! There's no real advantage to be had from moving that to NAND storage.
> They won't disclose exactly how it works.
Did you try reading what they *did* disclose?
> what they HAVE disclosed if they are going to make 30TB drives using 22TB platters.
As flyingpants265 already mentioned, this is to be accomplished by "increasing areal density of its ePMR disks by about 36%". So, that will make them 3 TB platters.
Western Digital is scum. I just discovered they removed Helium from all HDDs lower than 12TB. Including WD Gold, I think this happened in 2020. Since all the drives I have are from 2019. I had a DOA from WD which was made in 2020 but had no Helium for a 10TB Red Pro.
I have an 8TB Red CMR from Easy Store which has Helium, damn HGST had even 4TB drives with Helium filling. Now if consumer wants Helium they have to pay a ton of cash. HAMR is only there for 16TB and 18TB drives in WD Gold Portfolio or Ultrastar DC since Red maxes out at 14TB. Note, WD Red has higher 65C max operating temp while Gold is at 60C max and higher sound levels.
Insane segregation, they are real trash and sadly due to HGST technology they are the reliable option until Seagate proves themselves. Also speaking of Seagate I think their Ironwolf recently got Helium treatment.
The Physical drives do not have breather holes. That's the best way to check if the drive is Helium filled or not. After HGST acquisition WD made all the WD Red with Helium only because that's how HGST technology existed. By 2019 and 2020 that got changed. Next with He you get lower power consumption, naturally lower power consumption = less heat and more longevity.
I agree with ksec, that maybe what changed is the number of platters. With fewer platters you have less surface area and thus less drag. That lessens the need/benefit of helium.
> I just discovered they removed Helium from all HDDs lower than 12TB
Helium is a non-renewable resource. If the drive can still hit its operational parameters without it, then it's better not to use it. Its price has also gotten increasingly volatile, in recent years.
> due to HGST technology they are the reliable option
Didn't HGST also bring 2nd gen SMR technology, ushering in the wave of SMR consumer drives?
But it's not being produced where we can easily get it. We've had a glut of helium supply because it's a byproduct of natural gas extraction, but that won't continue forever.
No it does not, there is an abundance in space and in the sun, but on earth not so, it's only created by radioactive decay and about a tenth of what we use yearly is created yearly, see an issue here?
Currently we are using old helium trapped together with natural gas in rock far beneath the surface but we will run out of that, so until we learn to harvest it from outside our atmosphere it is a limited resource.
If they can make a hard drive without helium, it is probably better in the long run. If helium leaked from one of these sealed drives, it would render it useless.
That's a false-equivalence. The filters/seals on HDDs containing normal atmospheric gas only need to keep out dust. In the case of helium, the molecules are very small and the difference in density creates constant osmotic pressure for it to leak out.
It would be cool if they called next-gen HAMR technology JackHAMR. Maybe they do that with the drives utilizing suspension control, in their PZT actuation.
I suppose there's some nice long paper somewhere, but a 25 words or less context filler here would be helpful: why is it easier to nuke a drive with microwaves than a bit of heat? seems counterintuitive from an energy generation and control point of view. corollary: why does heat make a 'better' drive than microwave? that 5G network isn't doing much with microwaves, either.
From wikipedia: "The technology was initially seen as extremely difficult to achieve, with doubts expressed about its feasibility in 2013.[1] The regions being written must be heated in a tiny area - small enough that diffraction prevents the use of normal laser focused heating - and requires a heating, writing and cooling cycle of less than 1 nanosecond, while also controlling the effects of repeated spot-heating on the drive platters, the drive-to-head contact, and the adjacent magnetic data which must not be affected. These challenges required the development of nano-scale surface plasmons (surface guided laser) instead of direct laser-based heating, new types of glass platters and heat-control coatings that tolerate rapid spot-heating without affecting the contact with the recording head or nearby data, new methods to mount the heating laser onto the drive head, and a wide range of other technical, development and control issues that needed to be overcome.[2][3]"
I wonder if there's any possibility of using some sort of nano-array of magnets, in the write head. This could help address the issue of not disturbing adjacent tracks, by potentially rewriting their values at the same time as the current track is being written.
Of course, the annoying thing about that is how it basically turns every write into a read-modify-write operation, essentially like what we have with SMR. I hate SMR as much as anyone, but it seems there comes a point where areal density can't be increased without having to address the issue of "over-spray".
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
40 Comments
Back to Article
Sivar - Thursday, December 2, 2021 - link
"Meanwhile, it never mentioned microwave-assisted magnetic recording (MAMR) that was expected to precede HAMR"Was "preceded" the intended word here?
Ryan Smith - Thursday, December 2, 2021 - link
Precede is accurate. Until now, WD's roadmap was that MAMR would come before HAMR.Sivar - Friday, December 3, 2021 - link
I see, thank you for the clarification, and great article, and timely for some of my plans.nandnandnand - Thursday, December 2, 2021 - link
"HAMR is extremely important, great technology, it is still several years away before it is commercialized, and you can bet your datacenter on it"I remember the good old days when we thought HAMR could be coming to consumers around 2015-2016.
LtGoonRush - Friday, December 3, 2021 - link
So do the HDD manufacturers literally have no plan to keep offering relevant products? It just isn't plausible that there's a market for HAMR drives, the slightly lower cost-per-GB isn't worth the awful performance, density, heat, and power.Wereweeb - Friday, December 3, 2021 - link
Yeah, just put everything in magnetic tape /sThe electronics industries don't revolve around your computer. Datacenters need absurd amounts of immediately-accessible data. Facebook is not going to put every stupid post you made 10 years ago in an SSD.
Kjella - Saturday, December 4, 2021 - link
The text of the post probably is... all of Wikipedia is around 150GB compressed. It's hard to comprehend, but computers could easily record and store every keystroke from the day you're born to the day you die.mode_13h - Sunday, December 5, 2021 - link
@Kjella right, the big data hogs have moved from text to richer media, like photos and videos. Since a lot of people are including these in their posts, @Wereweeb's point still stands.And, while text went through some evolution from plain ASCII to unicode + markup, it never ballooned in size the way that higher-resolution, higher-framerate, and higher-dynamic-range cameras are causing photos and videos to grow.
bsd228 - Wednesday, December 8, 2021 - link
my wife took 150gb worth of gopro video footage on our last vacation. And it's not even all that much length, it's just that 4k/60 is huge.Cat pics and videos are exabytes alone.
Crazyeyeskillah - Wednesday, December 8, 2021 - link
tv's are in for a bad day when trying to convince people to move to 8k when nothing will be able to stream / support it for another 10 years.Zoolook - Sunday, December 12, 2021 - link
Wikipedia is currently around 20 GB compressed, text isn't very high density and is easily compressed.https://en.wikipedia.org/wiki/Wikipedia:Size_of_Wi...
mode_13h - Monday, December 13, 2021 - link
Yeah, but the images & other media files are already compressed and much, much bigger. According to that link you cite:"The size of the media files in Wikimedia Commons, which includes the images, videos and other media used across all the language-specific Wikipedias was described as well over 23 TB near the end of 2014."
In the past 7 years, it's surely grown by orders of magnitude.
alfalfacat - Friday, December 3, 2021 - link
Lol, someone got HAMR confused with SMRYou have no idea what you're talking about, higher capacity to drive lower TCO-per-GB is exactly what the market (hyperscalers + large enterprise) wants.
Samus - Friday, December 3, 2021 - link
A 22TB platter capacity with 8TB NAND doesn't make a 30TB hard drive. It makes a 30TB storage device. Interesting how they're spinning this.boozed - Friday, December 3, 2021 - link
Nice punBilly Tallis - Friday, December 3, 2021 - link
OptiNAND is not a new name for the existing hybrid drive concept, so it's incorrect for you to assume they'll hit 30TB by simply aggregating 22TB of magnetic storage with 8TB of NAND. They can't afford to put 8TB of NAND on a single hard drive, and will actually be increasing the real areal density of the magnetic storage.mode_13h - Saturday, December 4, 2021 - link
If you're talking about OptiNAND, that's not what it is."... they are stressing the fact that it is not a hybrid drive (SSHD). Unlike SSHDs, the OptiNAND drives do not store any user data at all during normal operation. Instead, the NAND is being used to store metadata from HDD operation in order to improve capacity, performance, and reliability."
Source: https://www.anandtech.com/show/16920/western-digit...
Samus - Tuesday, December 14, 2021 - link
I mean, they can say whatever they want, and it sounds like they are lying, because if the goal of OptiNAND is to increase capacity - whether through "metadata" or whatever - if its increasing capacity, user data is inherently stored there. My guess is they are using parity to cover data loss in the event of NAND block failure, damaged sectors, or power loss corruption. Who knows. They won't disclose exactly how it works.At the end of the day what they HAVE disclosed if they are going to make 30TB drives using 22TB platters. That implies there is at least 8TB of NAND, and while that is extremely expensive, I doubt a 30TB storage product is going to cost less than $1500 when the 22TB drive is going to cost $800.
flyingpants265 - Wednesday, December 15, 2021 - link
Did you read the part where it says they need a 35% density increase?flyingpants265 - Wednesday, December 15, 2021 - link
36%*. God damnitmode_13h - Friday, December 17, 2021 - link
> it sounds like they are lyingIf they were lying, there would be a big board *full* of expensive NAND chips attached to the HDD, and it would be much more expensive. This isn't something you can fudge.
> if its increasing capacity, user data is inherently stored there.
Are you more interested in being stubborn or actually learning something. The article I linked is fairly clear about what sort of information it holds. I'm not going to spoon-feed you more quotes, when you can just go read it, yourself.
> My guess is they are using parity to cover data loss
Disks have already done that for *decades*!! There's no real advantage to be had from moving that to NAND storage.
> They won't disclose exactly how it works.
Did you try reading what they *did* disclose?
> what they HAVE disclosed if they are going to make 30TB drives using 22TB platters.
As flyingpants265 already mentioned, this is to be accomplished by "increasing areal density of its ePMR disks by about 36%". So, that will make them 3 TB platters.
Silver5urfer - Friday, December 3, 2021 - link
Western Digital is scum. I just discovered they removed Helium from all HDDs lower than 12TB. Including WD Gold, I think this happened in 2020. Since all the drives I have are from 2019. I had a DOA from WD which was made in 2020 but had no Helium for a 10TB Red Pro.I have an 8TB Red CMR from Easy Store which has Helium, damn HGST had even 4TB drives with Helium filling. Now if consumer wants Helium they have to pay a ton of cash. HAMR is only there for 16TB and 18TB drives in WD Gold Portfolio or Ultrastar DC since Red maxes out at 14TB. Note, WD Red has higher 65C max operating temp while Gold is at 60C max and higher sound levels.
Insane segregation, they are real trash and sadly due to HGST technology they are the reliable option until Seagate proves themselves. Also speaking of Seagate I think their Ironwolf recently got Helium treatment.
ksec - Friday, December 3, 2021 - link
>I just discovered they removed Helium from all HDDs lower than 12TBBecause those doesn't need to pack as many disc? I am not even sure WD ever mentioned those model had helium in the first place?
Silver5urfer - Friday, December 3, 2021 - link
The Physical drives do not have breather holes. That's the best way to check if the drive is Helium filled or not. After HGST acquisition WD made all the WD Red with Helium only because that's how HGST technology existed. By 2019 and 2020 that got changed. Next with He you get lower power consumption, naturally lower power consumption = less heat and more longevity.mode_13h - Saturday, December 4, 2021 - link
I agree with ksec, that maybe what changed is the number of platters. With fewer platters you have less surface area and thus less drag. That lessens the need/benefit of helium.mode_13h - Saturday, December 4, 2021 - link
> I just discovered they removed Helium from all HDDs lower than 12TBHelium is a non-renewable resource. If the drive can still hit its operational parameters without it, then it's better not to use it. Its price has also gotten increasingly volatile, in recent years.
> due to HGST technology they are the reliable option
Didn't HGST also bring 2nd gen SMR technology, ushering in the wave of SMR consumer drives?
willis936 - Monday, December 6, 2021 - link
Non-renewable in what sense? The Earth is making more of it than we currently consume and will be for at least millions of years.mode_13h - Tuesday, December 7, 2021 - link
But it's not being produced where we can easily get it. We've had a glut of helium supply because it's a byproduct of natural gas extraction, but that won't continue forever.https://en.wikipedia.org/wiki/Helium_storage_and_c...
Zoolook - Monday, December 13, 2021 - link
No it does not, there is an abundance in space and in the sun, but on earth not so, it's only created by radioactive decay and about a tenth of what we use yearly is created yearly, see an issue here?Currently we are using old helium trapped together with natural gas in rock far beneath the surface but we will run out of that, so until we learn to harvest it from outside our atmosphere it is a limited resource.
mode_13h - Friday, December 17, 2021 - link
> until we learn to harvest it from outside our atmospherePfft. There's no way that's ever going to be cost-effective. It'd always be cheaper to scrounge for a little more, underground.
pugster - Monday, December 6, 2021 - link
If they can make a hard drive without helium, it is probably better in the long run. If helium leaked from one of these sealed drives, it would render it useless.willis936 - Monday, December 6, 2021 - link
If the hermetic seal fails on any HDD it is rendered useless.mode_13h - Tuesday, December 7, 2021 - link
That's a false-equivalence. The filters/seals on HDDs containing normal atmospheric gas only need to keep out dust. In the case of helium, the molecules are very small and the difference in density creates constant osmotic pressure for it to leak out.jasonriedy - Friday, December 3, 2021 - link
But can you program / progrom HAMR with Sycl?mode_13h - Saturday, December 4, 2021 - link
LOL. No, I doubt it's a technology from the USSR.mode_13h - Saturday, December 4, 2021 - link
It would be cool if they called next-gen HAMR technology JackHAMR. Maybe they do that with the drives utilizing suspension control, in their PZT actuation.FunBunny2 - Saturday, December 4, 2021 - link
I suppose there's some nice long paper somewhere, but a 25 words or less context filler here would be helpful: why is it easier to nuke a drive with microwaves than a bit of heat? seems counterintuitive from an energy generation and control point of view. corollary: why does heat make a 'better' drive than microwave? that 5G network isn't doing much with microwaves, either.Wereweeb - Saturday, December 4, 2021 - link
From wikipedia: "The technology was initially seen as extremely difficult to achieve, with doubts expressed about its feasibility in 2013.[1] The regions being written must be heated in a tiny area - small enough that diffraction prevents the use of normal laser focused heating - and requires a heating, writing and cooling cycle of less than 1 nanosecond, while also controlling the effects of repeated spot-heating on the drive platters, the drive-to-head contact, and the adjacent magnetic data which must not be affected. These challenges required the development of nano-scale surface plasmons (surface guided laser) instead of direct laser-based heating, new types of glass platters and heat-control coatings that tolerate rapid spot-heating without affecting the contact with the recording head or nearby data, new methods to mount the heating laser onto the drive head, and a wide range of other technical, development and control issues that needed to be overcome.[2][3]"mode_13h - Sunday, December 5, 2021 - link
I wonder if there's any possibility of using some sort of nano-array of magnets, in the write head. This could help address the issue of not disturbing adjacent tracks, by potentially rewriting their values at the same time as the current track is being written.Of course, the annoying thing about that is how it basically turns every write into a read-modify-write operation, essentially like what we have with SMR. I hate SMR as much as anyone, but it seems there comes a point where areal density can't be increased without having to address the issue of "over-spray".