Only wondering if they would ever going down in price and compete cost wise? I saw an 8TB SSD the other day for nearly 500 of my local currency, while the HDD equivalent was 150.
I think that SSDs probably have a longer life span (assuming), so perhaps the cost can balance out?
Probably not. HDDs need to be better in some categories than SSDs in order for them to be viable. If SSDs were ever able to reach price parity with HDDs, all of a sudden HDDs wouldn’t be as attractive and wouldn’t sell well so they have an incentive to make them cheaper than SSDs.
I wouldn’t necessarily expect an SSD to last as long as a HDD. Most SSDs after a year or two of having data “at rest” on them will start to have the NAND cell voltage degrade in a way that slows their read speed down by several orders of magnitude, slower than a HDD.
Most SSDs don’t have GC or wear leveling algorithms that will account 100% for the aforementioned problem.
Ratio in price/TB is more like 5:1 in most markets. Recent decrease in Flash prices made this more 4:1ish. And you want to compare similar product lines. Not some junk SSD like those 8TB Samsung or the stuff you buy at gas stations. An SSD isn’t better if the read/write speed converges to HDD levels or even below that. SSDs vary vastly in quality, while HDDs are pretty much homogenous at this point.
SSDs slowly degrade over time, where HDDs and the data are fine until some mechanical stuff breaks.
And if you buy 1TB SSDs, Flash is cheaper than a 1TB HDDs. I just proved that SSDs are already cheaper today.
This is not how things work unless you want the numbers to fit into your narrative/bias.
Overall we see a linear increase in HDD capacities and a more exponential increase for Flash for a given form factor. Price difference per TB will continue to decrease over time. HDDs will still keep their niche in the next 10 years.
Also, as flash chips scale, and get more dense, they are getting more affordable, but less durable.
HDD’s have pretty much plateaued, and the vendors are eeking out small (ish) margins on what they have, while having to invent new lasers / magnetic heating techniques to get even more dense, and stay relative.
Flash will Probably come down to HDD prices eventually and probably overtake (It’s already like, 5 pence per gig. Crazy prices compared to £1 / gig in the OCZ days) but the flash then, won;t be the super durable SLC / MLC of yesteryear… Will always be better for RND/Seek work tho
Flash drives are also (generally) smaller , more droppable and lighter, making pallets of then cheaper to ship, where HDD’s will again, be challenged in that aspect.
We’re not there yet. And don’t assume HDD prices being static while everything else is moving. There is too much R&D, factories and investment in HDD production to just quit. Once SSD come into dangerous proximity, HDD prices will drop too. Until we see manufacturers selling their HDD factories and dropping out of the market. But with Exabyte clouds consuming 100.000s of HDDs, it won’t happen anytime soon. Consumer market obviously already switched to Flash for the most part. So this impact has already happened. Didn’t change a bit.
I reckon it is not the endurance of NAND flash that is the concern, but cheap ICs and lack of QC. Most users who don’t abuse the paging file won’t go over 10% max TBW, and for everyone else there will be a market for TLC drives.
I generally agree, in 2023 endurance for most stuff and users is irrelevant. But this doesn’t change QLC drives usually also ship with lack of cache or bad controllers/firmware. Cheapskate-tier is still bad quality. People like buying stuff on a budget and accept the drawbacks. HDDs had this with SMR and things like 5.25" drives in the 90s. You still get what you pay for.
Hang on, I thought Puget or BackBlaze (or was it ServeTheHome?) said that SSDs degrade only on write but not on read and HDDs degrade on both write and read so we are now approaching a net advantage of SSDs vs HDDs in thr long run?
The physical structure of NAND cells degrades with every write and reduced data retention half life of a cell is the result (because the cells for the most part follow exponential decay of charge similar to how a radioactive material decays).
The reads themselves don’t actually degrade the drive but time will degrade the charge in NAND cells which leads to significant read speed problems after a couple of years of data sitting on a SSD without being re-written (or even less time if the SDD has been worn with writes).
In theory it would be possible for the SSD manufactures to implement a controller algorithm that either periodically samples NAND cells and rewrites them if they are weak (burning PE cycles) or just indiscriminately rewrites them based off of age and the problem would be solved… its just I don’t know of any that do this, likely because it would cause the SSDs to wear themselves out faster.
There’s a post with some benchmarks of this affect in the “Good high-end NAS cases don’t exist, Should the community make one?” thread.
Reads and writes aren’t typically what degrades HDDs, “violent” seeks and vibration are. Even though a HDD will typically give something like 550TB read/written per year as the amount of throughput it should be subjected to, that number is largely derived from the amount of expected violent seeks the drive will see.
Well… Yes, actually. Some would argue we are already there.
It is a complex issue for sure, but the main limiting factor is that HDDs are simply too big for their interface now. SATA III still isn’t saturated by most HDDs.
Let’s do a quick cost/capacity analysis. I’m going to compare the cheapest SSDs (in 2.5" and NVME) with the cheapest HDD at the same capacity. Source PC Part Picker:
Capacity
SSD NVMe Cost
SSD 2.5" Cost
HDD Cost
1 TB
$38.99
$34.99
$26.99
2 TB
$69.98
$68.98
$35.99
4 TB
$199.99
$179.99
$49.64
8 TB
$899.99
$499.00
$99.99
16 TB
-
$2499.99
$224.99
This may look like it is bad for SSDs, but then factor in the fact that 3 years ago 512 GB drives cost as much as 2TB drives do now. That is twice the doubling. At the same rate, we will be looking at $200 NVMe 16TB drives at around the start of 2026. Add to that, the fact that very few people need more than 4TB of storage today, and HDDs are pretty much useless today in anything but a NAS. You can find some small niche usecases for HDDs of course, but mainstream? Nah, forget it.
Now, capacity wise… Where is the HDDs? Well, Seagate has promised 30TB drives by 2023, and could increase this to 60TB by 2027. To get this high capacity, Seagate had to spend a lot of money. This means the new HDDs must come with a high price tag.
However… Since the mass market is in SSD chips now, it is probable that economies of scale will outcompete HDDs before 2030. I think 16TB at $300 will be some sort of sweet spot, as it looks like right now. Let’s see how long that will take!
We had 16TB NVMe SSDs for years now. 32TB is available in limited supply already. You should add this to the list. First hit on 16TB NVMe = 1500€ on 7450 Micron Pro retail price incl. VAT so make this 1300 for the american way of doing things, although you pay 1500
Stopping at 8TB is like saying HDDs cap out at 4TB because there aren’t any larger 2.5" HDDs out there.
The 32TB disk mentioned by @TimHolus changes hands for 4000€ in retail. So yeah, we probably stick to 16TB for the nice <100€/TB values. But 200 PBW value is certainly a nice figure. We can write more to Flash than to Optane now, exciting times.
There was this article about China limiting rare earth metal exports to the US last week and I thought that yeah Fancy headphones will get more expensive now and then wendell commented that HDDs will get more expensive.
As a side note, Micron was banned in China recently as a retaliation to banning of Huawei, Hikvision, etc.
Expect HDDs to get more expensive and SSDs to get cheaper.
True, though not at m.2 form factor yet (the 16 and 32 TB are u.2). But yes you can reach 100 TB drives with SSD if you really want to, but they will cost you an arm, a leg and a kidney.
Ah, I was focused on the Consumer market. SATA on m.2 is pretty much dead - Really no point anymore other than for doing say, 2xSATA drives in a single 2.5" bay if you REALLY need more m.2 slots. It is legacy now by all accounts. In the server market that might be different though.