DRAM Capacity Crunch Expected to Persist in 2018

Megalith

24-bit/48kHz
Staff member
Joined
Aug 20, 2006
Messages
13,000
According to DRAMeXchange, tight supply conditions that have characterized the DRAM chip market throughout 2017 are expected to persist for much of 2018, with suppliers planning only limited capacity additions. Suppliers will likely adjust their product mixes to allocate more production capacity to server DRAM in 2018 due to increased demand for data center building by technology giants such as Google, Facebook, and Amazon.

DRAMeXchange also said that prices for mobile DRAM chips are forecast to rise 10 to 15 percent in the fourth quarter, outgrowing the overall DRAM market, as OEMs ramp up production of smartphones. The expected increase will move mobile DRAM prices past PC DRAM prices for the first time this year, the firm said. DRAM prices have risen consistently throughout the year amid a shortage and higher demand created by their use in a greater number of products.
 
I should have enough ram until ddr5 come out. 3200 is good enough until then.
 
On June 14th, 2016, I ordered a Crucial 64GB Kit (4 x 16GB) DDR4-2133 ECC UDIMM for my FreeNAS server. It was a bit overkill at the time, but it only cost me $368 for the kit. Now, the price is $800. The 32GB kit is $400.
 
This is one of the reasons I decided to upgrade one of the older servers at the office instead of buying a new one.

Upgraded from a single 2.6Ghz 8 core Xeon to dual 2.8Ghz 10 core Xeons, and from 64GB ram to 256GB for around $1,200 (refurbished parts)

A new server with similar specs/performance would have cost over $10,000 (just the memory would have been almost $4,000)
 
This is one of the reasons I decided to upgrade one of the older servers at the office instead of buying a new one.

Upgraded from a single 2.6Ghz 8 core Xeon to dual 2.8Ghz 10 core Xeons, and from 64GB ram to 256GB for around $1,200 (refurbished parts)

A new server with similar specs/performance would have cost over $10,000 (just the memory would have been almost $4,000)

Looks like I upgraded at the right time too. We have a UCS system with 8 10 core Xeons and 1TB now. :D (across four blades) Picked up the second 512GB a couple months back. Glad I didn't wait until now. I love seeing low-green bar-graphs in VMWare.
 
Glad I got 16GB of C14 3200 back in February when prices were up, but not crazy like they are now. I should be good for awhile.
 
It seems like they are artificially inflating prices on dram by continuing to limit supply. I don't understand how more companies don't get into it, unless there are so many patents involved that it's monopolistic in nature.
 
This is one of the reasons I decided to upgrade one of the older servers at the office instead of buying a new one.

Upgraded from a single 2.6Ghz 8 core Xeon to dual 2.8Ghz 10 core Xeons, and from 64GB ram to 256GB for around $1,200 (refurbished parts)

A new server with similar specs/performance would have cost over $10,000 (just the memory would have been almost $4,000)

I've been wanting to upgrade the Westmere-EP era dual Xeon I have in the basement for some time, to something single socket, with lower power use, but whenever I run th enumbers I just can't justify it. The 192GB of Registered DDR3-1333 RAM I have would be too costly to replace with DDR4 RAM.

This is total bull.
 
This keeps up my new gaming rig may only get 8gb of ram...like I've had for 5 years. 32gb is already out the window, prices for 16gb are getting hard to swallow.
 
Between this & the current state of graphics card pricing, these companies are cutting their own throats.
 
Damn price gouge. I bought my 32gb 3200 cas 14 kit back in January for $210. That same kit is hovering around $360+ sense early summer.

On the plus side however I was able to get 64gb ddr3 ECC super cheap a few months ago.
 
They're doing this on purpose,keep production low and demand high,profits, profits, profits.

It is possible, but I kind of doubt it. Demand has skyrocketed for DRAM chips and really there are not that many companies that produce them. Its the same thing that happened with NAND Flash. There is talk of it potentially happening with Lith-Ion batteries next. The warning signs for these shortages have been around for a few years, but demand just grows and grows every single year.
 
No wonder new PC sales keep sliding. Costing more to replace with new than to keep and upgrade old hardware.
 
It is possible, but I kind of doubt it. Demand has skyrocketed for DRAM chips and really there are not that many companies that produce them. Its the same thing that happened with NAND Flash. There is talk of it potentially happening with Lith-Ion batteries next. The warning signs for these shortages have been around for a few years, but demand just grows and grows every single year.


Yeah, but the part that was telling to me was that none of the manufacturers have plans to increase capacity.

It does kind of sound like there might be a bit of a "gentlemans agreement"/collusion going on here.

You don't usually find an industry that is booming, with demand predicted to skyrocket, sitting on its hands not expanding capacity...
 
I cannot wait til China opens up some fabs, sadly we have a bit of a wait
You're going to be very disappointed. Their memory will be several generations behind current memory mfg's products (if it works at all).

Yeah, but the part that was telling to me was that none of the manufacturers have plans to increase capacity.

It does kind of sound like there might be a bit of a "gentlemans agreement"/collusion going on here.

You don't usually find an industry that is booming, with demand predicted to skyrocket, sitting on its hands not expanding capacity...
You need to look at the earnings history for the memory companies - their industry is considered cyclical because the price of memory jumps up, they all increase capacity and flood the market with product, which then crashes the price. They all lose money when they expand capacity; it seems they have finally learned their lesson now that there are only a handful of memory companies. It is not collusion, but common business sense that they are not striving to increase capacity by an means other than making denser chips. With 3D NAND, increasing layer counts increases the mfg time, so although they make denser chips, they take longer to complete (but the mfg's are building new NAND fabs). Also note it takes several billions of dollars to build a new fab - not exactly a no-brainer, especially when spending several billion dollars will most likely reduce profits instead of increasing them.

AI/Big Data/Smartphones/IoT as well as a recovered PC market have really jacked up demand and it will only continue to increase moving forward. Autonomous vehicles is another area that is increasing demand. Everything uses memory these days, including refrigerators, TV's, and thermostats. This was known over a year ago and was an amazing investment opportunity for anyone who paid attention. Micron (MU) had a share price of $10 1.5 years ago and is now over $40 - and should continue to increase for the next 2-4Q.
 
They're doing this on purpose,keep production low and demand high,profits, profits, profits.

Basic principles of Microeconomics dispute this. A shift in supply due to a shortage will indeed raise prices, but quantity demanded will decrease accordingly. Sorry for the impromptu microecon course...

In the example below, you can see equilibrium is at $25/ton for 5 tons of grapes, which is valued at $125. Due to a shortage in grapes, there are only 2.5 tons in supply at the price of $30/ton, which is valued at $75. The grape farmers here would love to make more money by supplying more grapes, and likewise buyers want more grapes, but can't or wont buy them at $30/ton. No one wins here.
decrease_in_supply.png


In terms of system memory, consumers will feel this the most. Are we going to roll over and pay the premium when we can simply squeeze a few more months/years out of our last gen machines? Remember we're talking about system memory here, its demand curve isn't going to be nearly as inelastic as energy/fuel.
 
Basic principles of Microeconomics dispute this. A shift in supply due to a shortage will indeed raise prices, but quantity demanded will decrease accordingly.
Makes sense to me; the memory mfg's are shifting their product mix to be more server and mobile heavy (thus, less consumer products).
 
holy shit. Sucks to be you guys who are building a PC.

$75 for a 8gb stick of ddr4.

ram and gpu prices through the roof. Glad my laptop came with 16gb and gtx1060 while it was cheap
 
Prices rise when they stop producing.
Basic principles of Microeconomics dispute this. A shift in supply due to a shortage will indeed raise prices, but quantity demanded will decrease accordingly. Sorry for the impromptu microecon course...

In the example below, you can see equilibrium is at $25/ton for 5 tons of grapes, which is valued at $125. Due to a shortage in grapes, there are only 2.5 tons in supply at the price of $30/ton, which is valued at $75. The grape farmers here would love to make more money by supplying more grapes, and likewise buyers want more grapes, but can't or wont buy them at $30/ton. No one wins here.

In terms of system memory, consumers will feel this the most. Are we going to roll over and pay the premium when we can simply squeeze a few more months/years out of our last gen machines? Remember we're talking about system memory here, its demand curve isn't going to be nearly as inelastic as energy/fuel.


Yeah...we aren't talking about grapes though.
 
Or to word the headline as if it were the no bullshit edition:

DRAM shortage to persist indefinitely due to us making as much money as we can get away with.

Crooked Korean fuckers.
 
Luckily most of the people here at [H] are running an overclocked Sandy which is, what... 90% as good as the best stuff out there today?

Man, I'd love to upgrade just because upgrades are fun. But in seven years we've had equal performance with higher prices...
 
Yeah...we aren't talking about grapes though.

The graph demonstrates a shift in supply due to a shortage and the loss to the economy as a result of being below equilibrium. If all you took away from that was "HERRRRR GRAPES" then our public school system has failed you.
 
Last edited:
Basic principles of Microeconomics dispute this. A shift in supply due to a shortage will indeed raise prices, but quantity demanded will decrease accordingly. Sorry for the impromptu microecon course...

In the example below, you can see equilibrium is at $25/ton for 5 tons of grapes, which is valued at $125. Due to a shortage in grapes, there are only 2.5 tons in supply at the price of $30/ton, which is valued at $75. The grape farmers here would love to make more money by supplying more grapes, and likewise buyers want more grapes, but can't or wont buy them at $30/ton. No one wins here.
decrease_in_supply.png


In terms of system memory, consumers will feel this the most. Are we going to roll over and pay the premium when we can simply squeeze a few more months/years out of our last gen machines? Remember we're talking about system memory here, its demand curve isn't going to be nearly as inelastic as energy/fuel.

You are assuming there is no collusion and easy access for new suppliers to come online. Since both have been shown to be true in this case...why would you think that "basic econ" would apply?

All high entry markets collude. Fuck, you know what happens during a "bad grape season", they have no issue with "extra water" making it into the system. But if it is a high yield season...WTF, keep that damn water away. They all cheat in these type of markets. But what would I know living in a wine a valley for a while. ;)

In case where there is a high entry fees, suppliers will often manipulate supply to optimize overall profit. Why wouldn't you...you would be doing your investors wrong by not doing so. Remember, the CEO's and all EMPLOYEES are beholden to the investors..not the consumer.
 
Collusion is a way to manipulate supply (i.e. shift) and barriers to entry will generally impact the elasticity of supply (e.g. the slope\curve). It doesn't change the fact that whether a tsunami wiped out their factories or fat cats decided to restrict supply, it's still a shift in the supply curve. I'm sure the decision to adjust their mix to favor server/mobile memory proved to have better margins and therefore profit than serving consumer-grade desktop memory.

My original point is that in a shortage situation your'e, by definition, not at equilibrium, which is not beneficial for a product with a typical demand elasticity. This was made to counter the argument that they're doing this on purpose to screw consumers into paying more for desktop memory; it doesn't gel because consumers will demand less at that price and both sides lose. They can have their cake and eat it too if there's a way to grow their consumer desktop memory supply without impacting the rest of their mix.
 
Last edited:
The graph demonstrates a shift in supply due to a shortage and the loss to the economy as a result of being below equilibrium. If all you took away from that was "HERRRRR GRAPES" then our public school system has failed you.

Yeah, and there are no differences between DRAM markets and grape markets, right? I think your ECON 101 class is the problem here.
 
Yeah, and there are no differences between DRAM markets and grape markets, right? I think your ECON 101 class is the problem here.

No, reading comprehension is your problem. :D

The graph demonstrates what a shortage means for buyers and suppliers: there's a net loss for both sides. People want to buy more RAM, and suppliers want to make more money and meet that demand, but cannot. The argument that they're constraining supply to gouge PC RAM buyer's doesn't make sense considering the demand for PC RAM is not inelastic.

Try again. Please say grapes one more time to prove you've nothing to add. Yknow what, nevermind... you're gone.
 
Last edited:
Glad i built my rig from Natex with 128G of DDR3 ECC with a dual xeon and mobo when they were on sale for like $600!
 
No, reading comprehension is your problem. :D

The graph demonstrates what a shortage means for buyers and suppliers: there's a net loss for both sides. People want to buy more RAM, and suppliers want to make more money and meet that demand, but cannot. The argument that they're constraining supply to gouge PC RAM buyer's doesn't make sense considering the demand for PC RAM is not inelastic.

Try again. Please say grapes one more time to prove you've nothing to add.

Look, "not inelastic" means elastic, so I think this might be a sign. DRAM (PC RAM? No.) is (relatively) an inelastic product. Grapes...now back to class for you.
 
Back
Top