Is this the ultimate limit for hard drive capacity or will they keep getting larger?
>The theoretical maximum size for a SATA drive is 16 exabytes (EB)but the 560MB/s SATA3 limit will make it impractical at some point and will have to be replaced with PCIe based U.2 (external version of M.2)
transferring a full 32TB disk at 560MB/s would take around 16 hours. there has to be a better way.
>>109081260Hell no. The theoretical limit for a 3.5 inch hard drive is somewhere at the atomic level, and that, with bit pattern media and HAMR, is somewhere at 200TB per PLATTER.So, if we could stack 10 platters, we could get 2 Petabyte drives and that would be the absolute limit for HDD.
>>109081260>shartgate
>>109081260Re-silvering one of those drives would take forever
Why does it matter, regular consumers won't be able to afford it anyway.
>>109081457>16 hoursWould it take the same amount of time for 4Gb chunks and maybe small 10kb files?
Use case?
>>109081457Yeah it's called SSDs and nvme but they're a thousand dollars per terabyte
>>109081457I'm really surprised they don't use SAS, or invent something better than SATAUSB cables get faster every year, why can't SATA
>>109081934because the transfer rate of the interface isn't the limiting factor
>>109081934 very conservative companies resistant to change, retooling existing manufacturing lines, upstream and downstream impact, re-marketing campaigns, etc, etc...>>1090817234GB chunks would take about the same time. 10kb files could take a lot longer because of random seeks. best practice is to zip small files together and transfer the compressed archive.
>>109081934>USB cables get faster every year, why can't SATASATA hasn’t received major speed updates because it maxed out its physical capabilities on copper wires years ago, and the industry intentionally pivoted to NVMe over PCIe, or any of it's innumerable fucking derivatives like U2 for high-speed storage. SATA is effectively "finished" evolving because faster speeds require a direct pathway to the CPU, which legacy SATA cables can't provide.
>>109081934sata isn't the limiting factor for transfer speed of spinning rust
>>109081457a 32TB SATA or SAS disk would be a member of a parity array and thus have a faster transfer rate. filling a single drive at that capacity would be like watching the grass grow.
why aren't 2 platter disk twice as fast. 4 platter disks 4x as fast etc.those heads should raid0 those platters.
>>109081260After about a terabyte, don't drive failures become stupid common? I'd be scared to have more than like 2TB on a single drive because those fuckers die.
>>109082273I remember 6 GB/s over a pair of differential channels is the maximum they can push without signal integrity going to various places, or the physical interface on SSD/HDD and mobo arranging to create problematic impedances.Only way for SATA to be faster is add more data lanes, a thing that is already done by PCIe slots.
>>109082606I guess it's not feasible anymore because all the heads are physically connected to each other to save costs and complexity on the actuators, but because the tracks are so narrow these days then slightly different thermal expansion of the different platters means only one head can be precisely aligned to a track at once, and that reading a different platter or side means re-aligning to that other track.
>>109082629It's a Seagate drive, they fail anyway
I'm tired of HDDs.They are fragile, slow and fragment data. Also expensive now thanks to AI datacenters.What happened to the ceramic-on-glass that WD promised us?
>>109081893You can 8 TB for a thousand dollars. That's still too much. The cost has got to come down at some point or the capacity has got to go up (so you get more for your money).
>>10908126064TB announcement in two weeks
>>109082629>After about a terabyte, don't drive failures become stupid common?no
>>109081934HDDs are slower than the SATA interface.
>>109082506but I thought rust was safe and performant?
When are we getting quartz-etching femtosecond laser drives?
>>109081260I'd love to have such a big drive, but it's always a risk. 16TB is already a lot. I'm terrified... imagine I have to resilver.
>>109082606no issue when everything is written to the platters with the all heads moving like one. the only drawback would be increased blocksize. blocksize * platters, basically. weird it hasn't been done...or at least tried.
>>109081526Is WD any better these days?
>>109081934Parallel SATA and more actuators inside the driveHire me Seagate
>>109081260>>109081524If 10TB was $200, then 100TB (which they have on their road map by 2030) should be $200 too when they arrive, they contain the same amount of hardware (motors, heads, plates), only better technology, the R&D costs are going to be gained back anyway.I won't believe their lie that 1TB = $20, 10TB = $200, 100TB = $2000.Once the AI pops, those big 100TB HDDs are going to become very common, and the demand for them should be low since most need only one of them, you can store way too much data on it to need a second one.
>>109083598They are both alright, even some toshiba drives are okay
>>109081524Bull shit. You cannot access that data with such precision with such form factor.
>>109081893> SSDs and nvme but they're a thousand dollars per terabyteFound the cheap ones yesterday
>>109081260Western Digital already showcased 40TB drives right now at computex.
>>109083989>they contain the same amount of hardware (motors, heads, plates), only better technology,Not true, they use different materials on the platters which have higher magnetic resistance so they can make the magnetic grains smaller without worrying about random bit flips, and the write heads need higher power or microwave/energy/heat assistance to write these platters, which translates to more durable, more expensive write heads so they don't fucking melt within a month.Take it this way, HDDs used to be built to the heat/speed tolerances of a Sedan before, now they are built to the heat/speed tolerances of a F1 race car. They physically use more durable parts, not the same ones they used before. Until production ramps up that these parts cost the same to manufacture, they will be keep being more expensive, but if the new parts use materials which are simply too rare to produce in higher numbers, then prices will not ever come down unless some new rare earth deposits are found and get mined, increasing global supply.
>>109084088you and a few decades back you'd have thought that it's not possible to drive displays with such precision that you can fit a 4k display into a mobile phone, yet we have reached the point that it's possible. Same thing.
>>109084161>unless some new rare earth deposits are found and get minedElon finna open space mines n sheeeit.
