News Western Digital envisions 80TB HDDs in 2030, 100 TB HDDs to follow — new HDMR tech enables record-breaking storage density

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
And 20 yrs ago, a typical drive was 160GB.
30 years ago, 140MB.

Technology advances.
And who can forget this beast: 250MB in 1979


250mb-hard-drive-1979.jpg
 
  • Like
Reactions: bit_user
Well, they have. For example, you can get 20 TB Seagate Expansion for $230. Maybe the better way is to get refurbished enterprise drives since they are higher quality to start and have been field tested. Improvements have slowed down since 2011 but they are cheaper, especially when adjusted for inflation (GPU prices rose faster than inflation, and are reliant on ever-more expensive nodes).
Off the back of a truck or used? The absolute lowest $/GB right now is 1.6 cents per GB. That is for a 5400rpm drive in the 8-12GB range. The cheapest 20GB drive available right now is $350.
 
I am skeptical of this. 10 years ago, HD makers can only do 2TB per platter, and now it is 3. Hard drive size got bigger because they put more platters and stuffed Helium in the drive.
Leaving aside the timeline of platter density increases, you do have a point that PMR (Perpendicular Magnetic Recording, now often referred to as CMR) has basically reached a density limit and new technologies are needed to continue increasing platter density. SMR (Shingled Magnetic Recording) helped a little, but it's not applicable to every use case and they really needed EAMR (Energy-Assisted Magnetic Recording) to further increase densities. For a while, I think it was Western Digital talking about about microwaves, but that seems to be abandoned and we're back to heat.

I do not believe there has been a meaningful increase in the platter count, any time recently. I think the upper limit has been around a dozen platters per drive, for quite a while. Recently, dual-actuator drives were introduced - but, those were about performance, not density.

Also, I might be wrong on this, but I think helium wasn't about density, but rather to improve energy efficiency. The main thing it does is reduce air turbulence over the platter surface and probably air resistance to the actuator arm.
 
Hard drives prices for a given capacity have always dropped over time and continue to do so whenever a higher capacity is released.
I think that's because once the industry figures out how to build denser platters, they can then make lower-capacity drives by using fewer platters than before. The platters themselves are more expensive, but simply needing fewer of them is what makes the drives cheaper.

If so, we could pack a lot more data on a 5.25 or even 8 inch platter! An 8 inch circle has 522% of the area of a 3.5 inch circle.
Yeah, I do wonder why the industry hasn't gone to even larger diameters than 3.5", in the modern era. Could mechanical stability be an issue, as diameter increases?
 
Last edited:
Unlike "normal raid", the array would be usable during the entire time.
Every RAID I've used remained available during rebuilds. Performance was impacted, of course.

Unraid defaults to a check a month. If the check takes over a week, the drives will be running all out 25% of the time while generating extra heat and using peak power for all that time and degraded performance.
Enterprise systems usually have something like a "patrol scrub" function that does consistency checks in the background (i.e. at a reduced rate and generally subordinate to normal reads or writes).

Modern hard disks will also tend to do periodic online self-checks, where they scan for failing blocks and rewrite or relocate them before they become unreadable. Check the SMART logs of your drives and you'll probably see records of when they ran.

Maybe these drives will be 2-3x faster than "normal" drives today? I believe SATA can go up to 600MB/s, and my 16TBs are less than half that.
If a hard disk scales density in both dimensions of a platter, the media transfer rate only increases proportionately to the square root of the platter density. So, it doesn't keep pace with capacity and that's why rebuild times have grown ever longer.
 
Last edited:
80TB mechanical in 2030 - 5 years from now. That is not news. What is news is NVME.
Huh? What's news about NVMe?

AFAICT, CXL storage devices are where most of the action is, these days.

Beware the Carrington effect.
You mean Carrington Event?
What I'm reading says that the main risk is to our power and copper-based telecommunications infrastructure. Aside from drives getting fried due to power surges, storage media isn't generally what you should be concerned about, in such an event.
 
And 20 yrs ago, a typical drive was 160GB.
30 years ago, 140MB.

Technology advances.
I don't think he's questioning whether or not it will eventually happen, he's questioning the time scale.

Seagate hasn't exactly delivered on their roadmap.

seagate-roadmap-2021-capacities_575px.png


30TB drives were supposed to be available in 2023, but you can't find anything larger than 24TB currently. I don't recall even seeing an announcement for a 40TB drive last year which puts 50TB next year in question.
 
I don't think he's questioning whether or not it will eventually happen, he's questioning the time scale.

Seagate hasn't exactly delivered on their roadmap.
Makes me wonder whether the plateauing of HDD densities could be having an inflationary effect. The rate at which new data is generated only seems to be accelerating, which means the rate at which cloud & datacenter must buy additional HDDs should be increasing, since density hasn't been keeping pace. This should be pushing data storage costs off of their historical depreciation curve that people are used to, and the mismatch might end up manifesting in the form of higher prices, in some cases.

P.S. technology is hard, especially as we get closer to physical limits. So, I don't really blame the industry for not holding to its roadmaps. Everybody seems to have underestimated how quick the transition to HAMR would be. Let's not forget that these are mechanical devices that operate with extremely exacting precision and have to do so very reliably and with usable performance. To me, it's pretty mind-boggling stuff, like building a formula one car that can stop on a dime - any dime - and does it over and over, hitting its mark every single time.
 
Seagate's website seems to indicate they are currently producing a 30TB CMR drive and 32/36TB SMR drives. Since they are no where to be found in retail, that would indicate they are all going to hyperscalers. Realistically, there probably isn't a whole lot of demand for drives that size among the general public.
 
So not off the back of a truck, repurposed from Chia mining farms.

Did you miss the warnings on this site?
Different product category. Best Buy has external drives constantly on sale which are not being using in Chia farms and resold. External drives are typically cheaper because they only carry a one year warranty. Hard drives are produced with very low margins, so any returns are a painful hit to profitability. Reducing the warranty from 3 or 5 years to 1 year allows for lower prices.
 
  • Like
Reactions: usertests
Different product category. Best Buy has external drives constantly on sale which are not being using in Chia farms and resold. External drives are typically cheaper because they only carry a one year warranty. Hard drives are produced with very low margins, so any returns are a painful hit to profitability. Reducing the warranty from 3 or 5 years to 1 year allows for lower prices.
Yeah, there is plenty of margin. Seagate has a profit margin just under 40% and we all know that their NAND products will have a slimmer margin so I suspect they are getting 60-70% margins on their spinning rust. Competition has dwindled to almost nothing so the HDD industry is just riding the cash cow until NAND matches them on price and makes them irrelevant.
 
Yeah, there is plenty of margin. Seagate has a profit margin just under 40%
Their FY 2025 Q2 report (dated Jan 21) states gross margin of 35.5%. The last 3 quarters were good, for them. Year-on-year, their gross margin increased about 12%.

and we all know that their NAND products will have a slimmer margin
I don't know that. I'd suspect it has a lot to do with the market conditions. However, their revenue is 93.3% from hard drives, so it doesn't really matter what their margins on SSDs actually are.

so I suspect they are getting 60-70% margins on their spinning rust.
That math doesn't work, given how much of their revenue is from HDDs. It's got to be close to their gross margins.
 
Yeah, there is plenty of margin. Seagate has a profit margin just under 40% and we all know that their NAND products will have a slimmer margin so I suspect they are getting 60-70% margins on their spinning rust. Competition has dwindled to almost nothing so the HDD industry is just riding the cash cow until NAND matches them on price and makes them irrelevant.
The fact the HDD industry has dwindled to so few competitors pretty much disproves your point outright. If it was so profitable, why did everyone leave the market? Starting in 2023 and continuing into 2024, Seagate lost money five quarters in a row.

Seagate-rev-history-to-Q3-fy2024.jpg


AI hyperscalers' surge in demand for enterpise storage has improved Seagate's outlook. That has no relevance to the margin on a retail hard drive sold to consumers.
 
The fact the HDD industry has dwindled to so few competitors pretty much disproves your point outright. If it was so profitable, why did everyone leave the market?
Nobody 'left the market'. It was all consolidation. After buying up their rivals, in whole or just divisions, there are only three HDD manufacturers on Earth right now and they simply match each other's pricing and compete through marketing. There are no challengers in the market trying to gain market share by being the value option.

Also, your chart is BS. Seagate's most recent three quarters they had income of $515M (Q2), $305M (Q3), and $336M (Q4) for the last three quarters. The business had also been paying down debt from acquiring Samsung's HDD division, so it appears their income was being deliberately depressed to lower debt.
 
Last edited: