Follow Slashdot stories on Twitter

 



Forgot your password?
typodupeerror
×
IT Hardware Technology

DRAM Prices To Slide More Than 40% in 2019 Because Chip Makers Can't Forecast (theregister.co.uk) 81

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. From a report: 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. "A weaker pricing environment for memory and some other chip types combined with the US-China trade dispute and lower growth in major applications, including smartphones, servers and PCs is driving the global semiconductor market to its lowest growth level since 2009," said Gartner analyst Ben Lee. [...] The upshot of this is that global semiconductor revenues are expected to drop 9.6 per cent year-on-year to $475bn. This is down 3.4 per cent on Gartner's earlier forecast and likely could be revised again before the end of 2019 is upon us. Given the volumes of DRAM swilling around the supply chain that have forced down price, oversupply is on track to spill into the first and second quarters of the next calendar year.
This discussion has been archived. No new comments can be posted.

DRAM Prices To Slide More Than 40% in 2019 Because Chip Makers Can't Forecast

Comments Filter:
  • So, is this the best time to buy? I could upgrade a couple of machines. Or are the prices still dropping?

    And, I should probably turn in my geek card here, but do SSDs run on DRAM, too? Can I get some cheap disk as well?

    • by bobbied ( 2522392 ) on Monday July 22, 2019 @01:55PM (#58967480)

      And, I should probably turn in my geek card here, but do SSDs run on DRAM, too?

      Yea, hand that card over..

      SSD's are made from FLASH not DRAM. Where there is a small amount of DRAM on the controller, DRAM is by it's very nature "volatile" which means if you remove the power, it loses the data it was storing. Non-volatile FLASH is what actually stores your data in an SSD and the small amount of DRAM is only a buffer to store data in transit.

      In short, there might be a slight cost reduction (maybe a few pennies) but I doubt the DRAM glut is going to affect your SSD pricing.

      However, I think there is a glut in FLASH memory too, so SSD's have been getting progressively cheaper and cheaper, while they keep getting bigger and bigger.

      • by Quirkz ( 1206400 )

        Thank you for the answer. As soon as you mentioned volatility, I realized I should have known that. Apparently I'm getting rusty on my hardware.

      • Still a good time to buy. I've been putting off updating one of my desktops because I've got 24GB of DDR3 in that machine and I don't want to have to re-purchase all that RAM in DDR4 when I upgrade. A big price drop makes it a lot more feasible.

    • There actually are DRAM disks, and before SSDs they were used quite a lot. They're still around, and they're still the fastest "storage" devices, but they're quite expensive, and they have to have their own batteries or they lose their contents any time the system loses power. Most of them take DIMMs of some sort so that they can be serviced, upgraded, etc.

      • I remember somewhere around 1995 my friend pulled a SCSI hardware RAM disk out of a dumpster and he was the legend of the neighborhood for months.

        He put 4 megabytes in it and jaws were bouncing off the floor.

    • As other have mentioned - this will not help drop the cost of SSD drives. But it does beg the question - how difficult is it to rework a fab configured for DRAM to manufacture FLASH storage? Assuming it is not that difficult, the cost of FLASH should mirror that of DRAM. So a glut of DRAM indicates not only a good time to purchase RAM but also a good (but not as good) time to purchase SSD drives.

      But at the end of the day the best rule of thumb is this; Get what you need when you need it. Any money yo

  • oh that's awesome (Score:3, Insightful)

    by fat man's underwear ( 5713342 ) <tardeaulardeau@protonmail.com> on Monday July 22, 2019 @01:48PM (#58967426)

    I can finally upgrade my Commodore 64 to the full 576K by putting in all the chips in my 1764 REU!

    https://en.wikipedia.org/wiki/... [wikipedia.org]

    • Re: (Score:2, Interesting)

      by Anonymous Coward

      As funny as that is you might want to wait a bit since this is coming out in December of 2019.

      https://retrogames.biz/the-c64

      Also that 8 Bit Guy is building a new Commodore like computer supposedly. He recently appeared on the CoCoTalk this past weekend and talked a lot about what he was going for and where things are. Yeah I know, he was talking about Commodore stuff on a TRS-80 pod cast but it was still pretty good stuff. Also the other part of that show talked about a warehouse full of old computer equipm

    • How quaint [youtube.com]
  • I'm goint to assume the writer meant "swirling" - but, c'mon ...

  • by Anonymous Coward

    This is not a bad thing. It means consumers will pay less for DRAM. In an era when government funnels as much money as it can from the poor to the rich, anything that can save the working class a few bucks is a good thing.

  • Nothing a flood or fire can't fix...

  • I built a new box back in 2016 and 2*16GB cost me €129. The price is now €133 - much lower than it was last year. So it looks like I'll be bringing the machine up to it's max ram sometime soon.
  • the inevitable DRAM price fixing and production collusion that the manufacturers have engaged in multiple times.
    • This is the real reason prices are falling.

      Several years ago the new investigations into their collusion started, so they upped capacity while backing off on the collusion (lower prices mean more sales, after all.)

      At the same time, the desktop market began its decline and the cell phone market hit saturation.
  • by Anonymous Coward

    This is what really annoyed me back in my financial/MBA training - posting as AC so that I don't out myself as scum - we were taught that we could predict the future.
    Yes. The future.

    When it comes to business - which is based upon a SOCIAL "SCIENCE" of economics - using trends and regression lines is just retarded. I got into some heated discussions with my professors about this. Business and economics is NOT based on natural physical laws. It's not like planets orbiting a sun or anything.

    It's abou

    • by Archangel Michael ( 180766 ) on Monday July 22, 2019 @02:44PM (#58967858) Journal

      I actually think you're mostly correct. That being said, forecasting is still a necessary evil that will be wrong most if not all of the time.

      The reason is because the predictive nature of forecasting doesn't have to be 100% accurate to be useful. I'll use Caveman analogy for my reasoning.

      Caveman Ugg is preparing for winter, because he knows it is coming. He can prepare for long winter, medium winter, short winter. If he gets it wrong, depending on how wrong he gets it, it could be deadly.

      He prepares for a long winter by working extra hard during the spring, summer and fall seasons, gathering all that would be needed for a long winter. The risk he takes now (extra hunting, gathering) risks CURRENT health / wellness for future success. He might get killed going for the big Mammoth, which would really help him during a really long winter. Risk = Death now, Reward = Survives long winter. In the case of short winter, looks really foolish. In the case of Medium winter looks like he over did it just a bit, but otherwise is okay.

      He prepares for average winter, by working normally during the spring, summer, fall seasons. He experiences average risks for those three seasons, and can survive all but the longest winter. Looks reasonable in most cases.

      He prepares for short winter. He spends more time with L'Ugga (his female), less time hunting gathering. He's living the good life for Spring, Summer, and fall seasons. Anything other than a short winter, he dies. It was a good life.

      Most of the time, forecasting Ugg figures out that most of the time most winters are average, occasionally long and occasionally short. Most of the time he's right (average winter). Except last winter, which was unusually cold and long (damn IceAge) but some how he miraculously survived by the skin of the mammoth blanket he ate. What Ugg doesn't know, as he prepares for next winter is that it will be an easy one, but he won't see it because he won't fall into the trap of a long winter again, and is killed trying for that one last beast. Turns out, Ugg sucks at hunting bear.

      Forecasting is a bitch, because even if you're right most of the time, if you're wrong even once in a bad way, it will kill you. Predicting is fraught with all sorts of dangers, often not apparent until too late. That doesn't mean it is worthless.

      • by Aighearach ( 97333 ) on Monday July 22, 2019 @03:42PM (#58968256)

        Grog plan for long winter. Grog smash Ugg at start of winter. Grog live the good life.

        This is what happened to the DRAM market. Some companies intentionally over-produced to harm their competitors and force a round of mergers that benefited them. Then the whole industry was stuck with low prices and everybody cut back on capacity because they had no choice. Now capacity is ramping up again. Prices will come down, but then they will continue to go down until somebody is hurting. This cycle will likely continue for a long time, until eventually production costs get low enough that nobody gets killed by the low prices and it becomes a very narrow margin business, at which point the big companies that dominate analog ICs will start making and selling them at some fixed percent markup, like transistors. Then there won't be any companies that specialize in DRAM.

        Same for Grog and Ugg. They will eventually be replaced by large tribes with communal planning for winter.

    • by Luckyo ( 1726890 )

      Technically you're wrong. Practically, you're 100% correct.

      You're wrong in that demand does in fact obey physical laws. You're right in that physical laws are at so many degrees of separation, that accounting for them is effectively impossible using only limited human cognitive abilities and computing power. You need something that is able to comprehend a far greater whole of inputs and actually able to cross reference them. This is not something that either we as humans or our computers are very good at.

      An

  • In the boom times, DRAM make money hand over fast. In bad times, it barely breaks even. It's been that way at least back into the 1980s.

    Forecasts are close to useless when it takes 18-36 months to design, build and equip a cutting edge wafer fabrication plant. At $10 billion plus, it's also a very, very expensive game only a few companies can play.

    Adding to that is the fact that some fabs in China [battleswarmblog.com] seem immune to business conditions, backed by big communist party interests, while others are entirely mirages, announcing grand plans to much fanfare only to never materialize, either unable to attract investors or having their funds magically siphoned off to some other enterprise..

    Forecasting on a year-by-year basis in such a cyclical space is pretty much useless. You invest as much as you dare, reap huge profits during the booms, and just try to hang on during the busts.

    It's a crazy business. The metaphor used is imagine a train whose cars attached to each other with giant Slinkies. Computer equipment takes off, then DRAM manufacturers ramp up, then semiconductor equipment manufacturers in the caboose ramp up, and by the time they're finally at full speed, the engine halts during another downturn, and the other cars come barreling along behind to crash as well.

  • Revenues of only $475 billion??? How will the industry survive?

    • by tsqr ( 808554 )

      Revenues of only $475 billion??? How will the industry survive?

      Revenue != Profit. But you knew that, right?

  • 10 years ago, we got 4GB of RAM for $20. It's taken this long just to get back to that, thanks to price fixing by the manufacturers.

    There is no reason why it shouldn't be $20 for 16GB or more at this point in time.

    • 10 years ago, we got 4GB of RAM for $20.

      Because we all know that nothing has changed in RAM technology in 10 years except the price...

      Also if you bought 4GB of RAM for $20 10 years ago, why? Were you building a computer for someone you hated? Do you just crave bluescreens and data corruption? Just because you can get something cheaply doesn't mean you *should*.

      thanks to price fixing by the manufacturers

      Except the price fixing was only about 20% of the problem which was a massive change in the demand side.

      • Since it was DDR3 and dual channel, it was two sticks of 4GB RAM for 8GB systems, which, at the time, was fine for many users. Also, Samsung's RAM sticks were awesome... they overclocked and just worked in everything (same could not be said for many other brands - even those with fancy heatsinks).

        As for my price fixing claim, I'll stand by it. It was the same problem as the HDD business, which used the Thailand floods as blame for artificially keeping the prices inflated. Let's go back even further, to 1994

        • Samsung's RAM sticks were awesome... they overclocked and just worked in everything (same could not be said for many other brands - even those with fancy heatsinks).

          I don't have to overclock my RAM, because it's already fast, and has faster timings than Samsung DIMMs. No matter how much you overclock RAM with slow timings, you'll never get it up to the same speed as RAM with fast timings.

  • This happened with DDR3 just before the DDR4 switch. This is happening....just in time for DDR5 and new, fresh, artificial supply shortages!
    • .just in time for DDR5 and new, fresh, artificial supply shortages!

      Changing over manufacturing isn't an artificial shortage. It's pretty much a requirement of a new production line. The problem is PC makers switching while the price is still at the bottom of the curve. Once production ramps up for DDR4, it will remain low until DDR5 production begins. DDR3 is expensive again now because nobody is really producing much of it.

      What they didn't anticipate was the Intel CPU shortage. PC makers can't switch to AMD in an instant, so prices go sky high and consumers buy fewe

  • Perhaps they should give their forecasting-AI more RAM.

  • They use verbiage intended to make this sound like a bad thing. This is definitely a good thing.

  • ...affects the price of measures of drinks that weigh 1.7718451953125 grams. Anyway, the dram measurement had pretty much disappeared by the time they introduced the metric system.
  • Every time RAM prices got too low? " the glue factory, the only one in existance" would burn down, etc.

  • No as far as customers are concerned.
  • After all supply and demand still appears to reign supreme on the market.
    With the release of the Ryzen 3000 series, at least here in Germany, I've observed a high demand for the high quality Samsung B dies in RAM. This made their availability go down since you'll hardly can find them in stock anywhere. Subsequently this increased prices from even the cheapest offers by over 15% for something like a G.Skill F4-3200C14D-32GTZ kit.
    Maybe when G.Skills Trident Z Neo hit the market with some serious over supply

Technology is dominated by those who manage what they do not understand.

Working...