I’ve just watched WD’s expose on SMR tech and underlying hows and why of the tech.
They are saying that HDDS will simply have to go SMR, there is no other option.
Underlying explanation is simple - even with all the new tech, there is a limit how physically small the smallest domain can be. That limit is imposed by the write head ( and NOT read head). Smaller discrete domains demand ever stronger magnetic fields, which can’t be generated.
Tech has came up with a “solution”/workaround of having “energy assist”( sort of turbo for write head) that allows the tech to jump a notch or two higher, but that’s it. Basically they use other means to pump (heat) energy into the rough area where then magnetic write head uses that as a base to make a final jump over the top of needed energy for bit magnetization.
But since that has limits, it has been combined with SMR trick - so write head writes the wider track. That track is wider than what the (narrower) read head can resolve, but then next track has offset less than the write head width, so it partially plows over the part of the previous track.
This is not a problem for reading it, since read head is much narrower. But rewriting it destroys the data on subsequently written following tracks.
So far, so good. But judging from their materials, one would expect for resulting track to be much narrower than the write head. Like 2-3x narrower. So, at 1/3x of the track width and at 1/3 of the bit length, one would expect to see ~10x the capacity.
And yet, all that we can see today is measly ~15%.
Companies usually guide you to more profitable products. SMR is more profitable than CMR. Volumn could be an excuse. The disgusting WD 6TB SMR was an example.
There’s a difference between seeing a 10x improvement in the lab and being able to do that repeatedly over a 5 year warranty and multiple DWPD without data loss, errors, performance degradation, firmware crashing due to OOM, etc. Those safety factors combined likely reduce the 10x theoretical factor of improvement to 1.15X.
To be slightly pedantic: we can generate stronger magnetic fields no problem, the issue is increasing the flux. There are ways to do this with some of the higher mu_0 materials before even broaching metamaterials.
EAMR adjusts media coercivity before a magnetic write, but it’s implementation isn’t limited to heat. We’ve got optically modulated coercivity materials as well that are much more promising than brute force method of approaching T_C.
A better way to think of this is that the magnetic media is so resistant to demagnetization that an electromagnet on the head alone can’t demagnetize it, making it more robust.
The tracks aren’t shingled nearly as close as the marketing will have you believe. Also guard bands are still used generously.
Not really. THere aren’t that many customers that would fuss around the difference between say 36TB and 40TB.
ESPECIALLY if using it meant forgetting everything they knew about HDDs, relearning the new stuff and then completely rewriting, drivers, adapting filesystems and good part of user code.
Well, if one gets in front of shareholders and starts spouting big numbers about SMR, you’d expect to see him being pretty confident that he’ll be able to reach them.