Micron Reveals 232-layer NAND Flash During Investors Day

[Valantar]

In all honesty I wouldn't mind seem more 2-bit MLC and even SLC more common and available in the consumer space again like when SSDs first came out.

It would be nice to have the option to get a 64GB (or larger) SLC drive for a scratch drive or something that's going to take a ton of abusive writes and still stand up.

You could get one of those Optane m.2 drives for that - and they also vastly outperform SLC NAND Flash at the same capacity, especially in random workloads and low queue depths. Much better suited to that task - but also, crucially, they have never sold in any significant quantity. The number of people willing to pay many times more for a small cache/scratch drive just isn't enough to sustain those kinds of products in the market.

What would be cool would be if drive vendors allowed for more flexibility in configuring their drives for the end user. For example, they could allow for alternative firmwares applied through their own drive management software that force the entire drive to run only in pseudo-SLC mode, at the cost of 2/3rds of its capacity - turning your 1000GB drive into a 333GB one, etc. This is likely a non-trivial undertaking though, given that the controller firmware would need significant changes and the increased performance would put more stress on the controller, increasing heat output under sustained loads. But it would be an option, much cheaper than bespoke products, and it would be pretty neat even if extremely few people would end up using it. (And of course there's increased risk involved as such a firmware change would require reformatting the drive, increasing the likelihood of PEBKAC errors.)

 

[Jokii]

I've been watching SSD prices and they're roughly the same as in 2019, when comparing similar performance/features/warranty. The best price per TB are still 1TB size drives (aka it's cheaper to buy two 1TB SSDs than one 2TB SSD). A 4TB SSD still costs an arm and a leg (and 4TB is nothing if you want to store your movie collection).

All this talk about speed, PCIe 4.0 5.0..., but a lot of us still have to use HDDs at the end of the day, because large SSDs are so expensive.
I wouldn't mind even a 3.5" SATA 500MB/s SSD if the price was reasonable. It would still be much faster, quieter, cooler than a HDD. Why is that not a thing?

 

[ARF]

I've been watching SSD prices and they're roughly the same as in 2019, when comparing similar performance/features/warranty. The best price per TB are still 1TB size drives (aka it's cheaper to buy two 1TB SSDs than one 2TB SSD). A 4TB SSD still costs an arm and a leg (and 4TB is nothing if you want to store your movie collection).

All this talk about speed, PCIe 4.0 5.0..., but a lot of us still have to use HDDs at the end of the day, because large SSDs are so expensive.
I wouldn't mind even a 3.5" SATA 500MB/s SSD if the price was reasonable. It would still be much faster, quieter, cooler than a HDD. Why is that not a thing?

Greed aka pursuit of ever growing profit margins, and unsustainable progress driven solely by the greed.
You can buy a 14 TB HDD for 250.
Or you can buy a 4 TB SSD for 350.

 

[trparky]

$456 for only 64 GBs of storage? Holy price tag Batman!

 

[Valantar]

I've been watching SSD prices and they're roughly the same as in 2019, when comparing similar performance/features/warranty. The best price per TB are still 1TB size drives (aka it's cheaper to buy two 1TB SSDs than one 2TB SSD). A 4TB SSD still costs an arm and a leg (and 4TB is nothing if you want to store your movie collection).

All this talk about speed, PCIe 4.0 5.0..., but a lot of us still have to use HDDs at the end of the day, because large SSDs are so expensive.
I wouldn't mind even a 3.5" SATA 500MB/s SSD if the price was reasonable. It would still be much faster, quieter, cooler than a HDD. Why is that not a thing?

Greed aka pursuit of ever growing profit margins, and unsustainable progress driven solely by the greed.
You can buy a 14 TB HDD for 250.
Or you can buy a 4 TB SSD for 350.

Are OEMs keeping prices high through framing these as premium products? To some extent, sure. But the HDD-SSD price difference is mainly down to component costs, and higher capacity = more flash. And flash is the most expensive part of any SSD of significant capacity. HDD platters are on the other hand relatively cheap per capacity, allowing for good price scaling as capacities increase - the motor, housing, controller PCB and the like represent a higher relative cost compared to the storage medium itself than in an SSD, meaning that the more storage medium you have, the less you notice the cost of the more expensive components. That's why HDD's have never, ever gone below a base cost of ~$40 unless on clearance/being sold at a loss - you just can't make them that cheap. SSDs on the other hand have extremely low base costs, as they have tiny PCBs, controllers are cheap and relatively simple (especially on the low end, higher end ones can be a tad expensive), and cheap ones can even omit DRAM for cost savings. That's why you can find decent low capacity SSDs at around $20.

This means that HDDs and SSDs scale across capacities in different ways:
- HDDs start out slightly expensive, but get progressively cheaper as they increase in capacity. New innovations and ultra-high capacity models break this trend somewhat (helium drives, exotic write methods, extreme platter density), but it is generally true.
- SSDs start out very cheap, but price/GB stagnates quickly as you pass the point where the only difference is the amount of flash. And as high capacity SSDs are expensive and thus relatively niche products, they don't benefit from economics of scale to the degree that smaller ones do, rendering them more expensive/GB than their mid-capacity siblings.

Explaining SSDs being more expensive than HDDs only by "corporate greed" or expecting SSDs to match HDDs for price/GB at any point in the foreseeable future just entirely fails to take into account the realities of manufacturing, distributing and selling these products. Unless we see some massive ground-breaking improvement in the fabrication of NAND flash, it'll be years still until we see 2TB SSDs come close to even 4TB HDD pricing today, and that's ... well, that's just reality. The flash for a 2TB SSD alone costs several times more than the retail price of a 2TB HDD.

 

[ARF]

Very good explanation about the pricing.
The thing is that it doesn't work for the users who would not be convinced to pay the premium because the technology doesn't allow to make high capacities cheaper.

Then try to find another technology for faster memory storage which can be manufactured cheap at capacities like 10 - 20 TB.

We need that capacity and speed, and cheap.

 

[mechtech]

Are you willing to pay the price though?

https://www.mouser.com/ProductDetail/Virtium/VSFPM8PC064G?qs=mAH9sUMRCtuAyyXuuQeUng%3D%3D

SFSA128GM1AA4TO-I-NC-616-STD Swissbit | Memory Cards, Modules | DigiKey

Order today, ships today. SFSA128GM1AA4TO-I-NC-616-STD – Solid State Drive (SSD) FLASH - NAND (SLC) 128GB SATA III M.2 Module 3.3V from Swissbit. Pricing and Availability on millions of electronic components from Digi-Key Electronics.

www.digikey.com

Which is ridiculous, because I remember buying an SSD when they first came out, it was SLC, and about $350 for 120GB ish With all the dies shrinks since then SLC should be cheaper now then it was then.

 

[R-T-B]

Which is ridiculous, because I remember buying an SSD when they first came out, it was SLC, and about $350 for 120GB ish With all the dies shrinks since then SLC should be cheaper now then it was then.

Not only that but that drives performance metrics aren't exactly stellar either...

 

[TheLostSwede]

I've been watching SSD prices and they're roughly the same as in 2019, when comparing similar performance/features/warranty. The best price per TB are still 1TB size drives (aka it's cheaper to buy two 1TB SSDs than one 2TB SSD). A 4TB SSD still costs an arm and a leg (and 4TB is nothing if you want to store your movie collection).

All this talk about speed, PCIe 4.0 5.0..., but a lot of us still have to use HDDs at the end of the day, because large SSDs are so expensive.
I wouldn't mind even a 3.5" SATA 500MB/s SSD if the price was reasonable. It would still be much faster, quieter, cooler than a HDD. Why is that not a thing?

Depends where you live I guess, as it's not cheaper to get two 1TB drives here. You're right about 4TB and larger drives though.

 

[Valantar]

Very good explanation about the pricing.
The thing is that it doesn't work for the users who would not be convinced to pay the premium because the technology doesn't allow to make high capacities cheaper.

Then try to find another technology for faster memory storage which can be manufactured cheap at capacities like 10 - 20 TB.

We need that capacity and speed, and cheap.

I mean, we all want infinitely fast CPUs and GPUs for cheap as well, but that doesn't mean anyone in the world has the foggiest idea of how to make that into an actual material reality. AFAIK there are no such storage technologies even on the horizon - the focus for future storage tech seems to be more along the lines of non-volatile RAM (expensive!) or massive capacity archival storage (cheap, but generally not suited for home use). I haven't seen even a single report of some "better than flash for the same use case and cheaper" tech even in conceptual design stages. There have been people arguing that stuff like MRAM would do that, but those promises have been a decade or more in the making and have yet to come evel close to panning out.

And there's good reason for this: for most people, storage needs have dropped significantly in the past decade. Most people are perfectly happy with a 1TB SSD in their laptop (or 128GB in their phone). Of course this is a bit misleading, as they're instead just using someone else's HDDs (aka cloud storage), but the installed capacity for the average PC sold has dropped precipitously since the early 2010s. Streaming and cloud storage killed per-device high capacity storage, and power users have moved to NASes and storage servers, which happily run HDDs as they're generally fast enough for mass storage use and cheaper than anything else that's a reasonable alternative is likely to be in the foreseeable future. Of course there are arguments for current cloud storage practices being rather unsustainable (in terms of cost, drive replacements, and being premised on an idea of growth offsetting low incomes, which also applies to streaming services that are currently struggling in a saturated market), but the solution to that for home users is getting a HDD-equipped NAS, not inventing a new storage technology.

HDDs exist, are fast enough for nearly all mass storage needs, and are quite cheap per capacity. You don't need more than ~150MB/s for media storage, which is also typically mostly sequential reads, which HDDs handle just fine. Very few people have an explicit need for a multi-TB game library, for example. I mean, we can always dream of more for less, but that doesn't mean it will happen in the near future.

There's an inherent problem with this dream: (nearly?) all solid state storage is dependent on silicon lithography, and silicon lithography is expensive compared to making a HDD. Stacked 3D flash is a good way of offsetting this, as it significantly increases per die area storage capacity without sacrificing endurance or performance. So unless one finds a way to drastically increase density, cut lithography costs, or both, this is just going to continue on its current trajectory of slow-and-steady flash price drops. And it's not going to catch up to HDD pricing at mid-to-high capacities. That just isn't technologically feasible. Flash allows for better performance and higher density, but not lower pricing outside of the extreme low end of capacities.

 

[Aquinus]

Very few people have an explicit need for a multi-TB game library, for example. I mean, we can always dream of more for less, but that doesn't mean it will happen in the near future.

People should just use Linux and then you could use dmcache with an NVMe drive to do read and optionally write caching against physical media, like a RAID-1 of two 8TB drives or something. Tada! It'll feel like the fastest hybrid drive that you've ever used in your life with the capacity and resilience that "today's consumers" crave.

 

[Valantar]

People should just use Linux and then you could use dmcache with an NVMe drive to do read and optionally write caching against physical media, like a RAID-1 of two 8TB drives or something. Tada! It'll feel like the fastest hybrid drive that you've ever used in your life with the capacity and resilience that "today's consumers" crave.

Lol, yeah, NVMe/Optane caching in front of slower mass storage is one of these persistent great-on-paper ideas that never really works out in real life, no matter the OS and software involved.

 

[Denver]

Perhaps a HDD in raid 0 or a fast HD(Mach.2 Exos 2X14) is the solution with the best balance between high capacity x speed x price for now.

 

[trsttte]

leave one as a raid backup and it would still be cheaper than buying a single 4TB SSD

Raid is not a backup

Perhaps a HDD in raid 0 or a fast HD(Mach.2 Exos 2X14) is the solution with the best balance between high capacity x speed x price for now.

For bulk storage sure, but regular use it simply can't compare with any modern SSD (even a SATA one). It's not just speed, but also latency and IOPS, SSDs are simply in a different league.

For a large boot drive you can get a 1tb drive for ~70€ today, granted you can get HDDs like the WD blue for about 30 or 35€ but that's one of the slowest hdds on the market.