God DRAM you! Prices to slide more than 40% in 2019 because chip makers can't forecast The laws of botched supply and demand forecasting are coming home to roost for the semiconductor industry in 2019 with DRAM average sales price set to fall 42.1 per cent. The latest ladle of doom and gloom was poured onto the sector this morning by Gartner, days after IC Insights delivered its dark prognosis for chip makers …
No doubt the number was truncated for publication: if the fall had been 20% in the first 173 days of the year (to 22 June), then extrapolating to the full 365 days would have given 42.1965% for the full year. But presenting a forecast to that much precision would have been silly.
Well, I already switched to SSD for my bulk storage (4x1TB in RAID-Z1). I have bought about 8TB of SSD storage for the last year or so and the prices keep falling.
However, you can combine the best of both worlds (on the HDD side being nothing more than the price, of course) by using ZFS with additional SSD(s) for L2ARC device(s). Once the cache has been warmed up it performs admirably. The only problem is that with ZFS on Linux the cache is empty after reboot but if you boot no more than once a month that's pretty much negligible.
Or for those on Windows, you can use tiering with Storage Spaces, provided you have at least two spare SSDs (they have to be mirrored to prevent data loss apparently). You'll have to do everything through Powershell though if you're using Win10, only the server OSs have a GUI for tiered storage.
Performance is about what you'd expect, fast when it's using the SSDs, slow when using the spinning rust. My use case is so far from the intended use that I can't really comment on how well the prioritisation works (it's supposed to keep frequently used files on the fast tier).
You can also use the SSDs as a straight-forward cache (and in conjunction with tiering).
I'd rather my data be safer than on an SSD no matter what the speed.
Once SSD start using memristors then I may be interested. Wearing out semiconductor junctions my forcing electrons through (SSD NAND) seems a step backwards for me.
I have an SSD for booting and root partition. Temp files, large files, files that may need wiping when deleted are all on HDD's. I dont even care if the HDD is a 5200rpm or 7200rpm device. Makes little difference although the slower one would run cooler.
I don't think the impact of the looming trade war between S.Korea and Japan has been taken into account, as if it eventuates it will lead to a significant increase in DRAM prices due to the near obliteration of the S.Korean DRAM manufacturing capacity (as they source 92% of their photoresist chemicals from Japan).
This is already having an effect on SSD and client RAM prices. Server RAM is so far unaffected but it's probably just a matter of time.
On top of that, the Pound is currently looking poorly which could add a few percent to the price if you're buying in the UK.
As requested:
The photoresists are the really crucial item and the story is completely different here. South Korea will be hard-pressed to easily alt-source enough material when 92 percent comes from the Japanese. Further, these photoresist chemicals are comprised of complex molecules and quality is of utmost importance as TSMC’s half a billion dollar blunder earlier this year illustrates. TSMC used either expired or tainted photoresist and had to throw out thousands of costly wafers.
Combined, Japanese suppliers control about 90% of the global resist and etching gas markets. The new screening process is expected to slow down exports, potentially hurting the South Korean electronics makers that buy the materials.
If everyone believes that RAM-Chips are going to be cheaper tomorrow, people will stop buying RAM-Chips now and the prices will fall.
Same goes for news about prices going up which cause a surge in demand causing higher prices.
Industry publications are full of that.
always have been, you have the big boys who vary production to try to controll prices,
but there are the oportunist shops that produce what ever is expensive at the time
so once demand increases due to supply shortegaes, increasing prices more fabs come online, then there is a glut of supply so prices fall and the fabs move on to something else so supply falls again and prices rise
My general impression is that the very high fixed costs for semiconductor fabs (up front capital costs plus skilled labor), tend to make it cheaper to produce products and sell for whatever one can get than to not produce. That's sort of like oil wells. Once you've spent millions to drill the thing and committed to land lease payments and such, you might as well pump whatever hydrocarbons you can even if the flow is not what you hoped for. There's a 30 year old report at https://www.nap.edu/read/1890/chapter/5 that tends to support that argument.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
