News Toshiba exec claims hard drives are 7X cheaper than SSDs and will continually evolve for large datacenters

[Admin]

A Toshiba exec believes that hard drive technology will continue to improve, enabling HDDs to be used well into the future admist growing energy costs.

Toshiba exec claims hard drives are 7X cheaper than SSDs and will continually evolve for large datacenters : Read more

 

[USAFRet]

Cheaper per GB, sure.
In data centers? Sure.
Even in the 100TB in my NAS? Sure.

For home and workstation use? Not a chance.

 

[hotaru251]

HDD's in data center has a limit.

A drive can only get so large with current tech limitations before a failure results in data loss risk due to having so much data its physically impossible to copy the data to a new drive within the time frame.

and ssd have a lower power consumption & heat (benefit of no moving parts) output (which data centers spend a lot to dissipate) and thats ignoring the reliability of ssd over mechanical drives.

now will hdd actually dieout? probably not however they will become more and more rare for sure.

 

[bit_user]

The cost of SSDs most SSDs today is nowhere near the "7X" multiplier Kaese suggested.

Then again, NAND-makers and SSD companies were losing money, in 2023. Such pricing is unsustainable, invalidating any long-term predictions based on it. I saw one prediction that SSD prices will rise up to 55%, in 2024.

We'll have to wait and see which prediction comes true in 2028.

Both predictions are from industry players with a vested interest in their respective outcome. If you want a good prediction, look to an independent market analyst.

 

[George³]

Before almost 6.5 years:
Dude, where is my 100TB hard drive?
The only thing that is certain is that HDDs have fallen far behind in capacity advancements compared to past promises and development plans.

 

[thestryker]

HDDs also have an advantage in power consumption/GB when talking about always on for higher capacities.

I mostly just hope that the shifting market doesn't make it a lot more expensive to get drives for local storage as a regular customer.

 

[FoxtrotMichael-1]

This is an analysis of datacenter storage, not consumer storage. From that perspective data costs for SSD are absolutely 7x the cost of HDD because Enterprise SSDs are absolutely bonkers in cost due to overprovisioning. If you look at your average consumer NVMe drive they aren’t rated in terms of TBW while Enterprise drives are. Even a relatively small enterprise SSD is well over $1000. HDDs will continue to have a place in the datacenter for years to come, if anything due to the need for data tiers of different speed and cost.

 

[Giroro]

"HDDs maintain a gap in cost per capacity with flash storage of around a factor of seven. "

Its reassuring he never actually said anything as grammatically confused as "7x cheaper". This implies he knows that multiplying by 7 makes numbers bigger, and how to talk about numbers in general.

 

[SethNW]

I would say it is half/half, like HDD will likely die for home use, with exception of long term local storage maybe. But for business use, they will keep using them. They won't just switch all storage to more expensive SSDs. Since per GB, HDD will be cheaper and not everything needs speed and snappiness of SSDs. And of course the better be cheaper, as they are lower in performance. Nevertheless regular users wise, as internal storage, HDD will die. Even games started to take advantage of SSD.

 

[das_stig]

HDD's have years of life left in them for Enterprise, in the hybrid config of hot data on SSDs and cold data on HDDs, will provide the best cost/performance benefits.

 

[hotaru251]

HDDs also have an advantage in power consumption/GB when talking about always on for higher capacities.

i'd find that questionable.

HDD by deisgn have a higher power draw as they are mechanical in nature. (it takes energy to physically move soemthing)

i'd be interested in seeing em compared.

largest ssd i think is like 100TB vs 20TB for hdd. (but they are crazy expensive)

 

[bit_user]

HDD by deisgn have a higher power draw as they are mechanical in nature. (it takes energy to physically move soemthing)

It takes energy to manage flash, refresh DRAM, and keep the controller in a power state where it can respond with the minimum latency.

The nice thing about HDDs is that the power needed to keep the platters spinning is roughly invariant to the density of the platters. You can only fit so many of them in there, so the frictional losses don't change as capacity grows.

i'd be interested in seeing em compared.

you can look up specs as easily as the rest of us.

largest ssd i think is like 100TB

I think only about 60 TB, if we're talking about NVMe drives. Only slower, read-oriented QLC models are such high-capacity.

20TB for hdd.

28 TB, I think.

 

[thestryker]

i'd find that questionable.

https://www.solidigm.com/content/solidigm/us/en/products/data-center/d5/p4320.html

rated at 5W idle and 15W active for 7.68TB

https://www.westerndigital.com/products/internal-drives/wd-gold-sata-hdd?sku=WD202KRYZ

rated at 5.8W idle 6.9W active for 20TB (you'd need to look at the PDF as they don't list it on the page)

Now HDD makers do have a tendency to not show peak, but you can check reviews here on Tom's and you'll see averages are usually 7-8W on their tests with peaks spiking due to the mechanical nature of HDDs.

 

[hotaru251]

I think only about 60 TB, if we're talking about NVMe drives. Only slower, read-oriented QLC models are such high-capacity.

consumer yes.
100TB is not consumer drive it uses enterprise MLC and costs 40 grand
https://www.techradar.com/news/at-100tb-the-worlds-biggest-ssd-gets-an-eye-watering-price-tag

 

[George³]

consumer yes.
100TB is not consumer drive it uses enterprise MLC and costs 40 grand
https://www.techradar.com/news/at-100tb-the-worlds-biggest-ssd-gets-an-eye-watering-price-tag

Yes, back far away in the past in March 2018. But this is exceptional device which isn't have successor by capacity to this moment despite all meantime developed and released much modern and much higher bit density nand flash chipy. I was really excited in 2018 and waiting for 200TB in 2020, 400 TB 2022...I was fool.

 

[hotaru251]

despite all meantime developed and released much modern and much higher bit density nand flash chipy.

advancement consumer ssd have doesnt benefit enterprise.

Consumer ssd are getting larger and stuff but its at the cost of drastically less endurance.

Data center prefer not going past MLC as its sweet spot for durability & speed.

and why would they(company made 100TBssd) advance something that has no competition & a limited audience?

 

[Co BIY]

If the "need" for data storage continues to grow at the current rates there may be room for all the technologies regardless of whether they are the most "efficient".

It seems like software that classifies and controls the storage of data according to priorities , reduces redundancy and eliminates unneeded data is an area with potential for growth.

There is probably someone here on Toms with deep knowledge on the state of the art with regards to that.

 

[bit_user]

Data center prefer not going past MLC as its sweet spot for durability & speed.

Not true. I already mentioned ~60 GB models (from Solidigm and Micron, perhaps others) that are QLC. These are sold as light-duty, read-oriented models.

TLC is now the main workhorse for mixed-workload datacenter drives.

and why would they(company made 100TBssd) advance something that has no competition & a limited audience?

Dunno, but I don't see much news or any reviews of their stuff on StorageReview.com, so I assume they're a niche player.

 

[bit_user]

It seems like software that classifies and controls the storage of data according to priorities , reduces redundancy and eliminates unneeded data is an area with potential for growth.

That stuff is old hat. You can read about some of that sort of middleware on nextplatform.com, but I find that site isn't very approachable to outsiders.

In some ways, I think storage virtualization is really what enabled cloud computing. Oh, and data de-duplication has been around for ages.

There is probably someone here on Toms with deep knowledge on the state of the art with regards to that.

Not me, but I know people who are/were plugged into that world.