"This should mean cheaper DRAM pricies for consumers." - good joke :P
This will mean higher profits for samsung not lower prices, many ram manufacturers have already decided to reduce production in order to keep prices high. We need more companies producing ram to bring down prices. China has just built some but they are using 22nm or something and only low volume.
The price of DRAM has already been in free fall for months are we are already back to late 2016 levels, and on course to reach 2016 or even lowest $/GB Memory in 2019 assuming no further rebounce.
These 10nm class DRAM will allow them to sustain their operation and profits even in 2016 DRAM price level, which is good for everyone.
We are NOWHERE NEAR late 2016 prices, unless you mean high latency 2400 Mhz DDR4. I paid $219 for 2x16GB Trident Z 3000Mhz CL14 DDR4 and bought two sets for a total of ~$440 in March 2017 for a new Ryzen build. Now it's still hard to find CL14 DDR4 3000; mainly because anything with that latency is also sold at higher latency/speeds and prices.
Considering the current DRAM glut is said to be mainly linked to Intel CPU shortages it might still be reasonable to keep investing in more fabs or new tech as those shortages will not last too long considering the grand scheme of things where product production has a lead up time of years.
Is mobile phone, or other non PC, DRAM also getting soo cheap?
Product *development* has a lead time. Product *production*, meaning ramping up production, has close to no lead time. How is knowing the shortage won't last encouraging them to invest in new factories? New factories means higher production (so supply), which means lower prices. So you expect them to spend money to build factories that make their product worth even less?
Come on, don't they teach any economic concept in school these days?
The only thing that could help now (other than any legal investigation that might prove a price fixing cartel) is for the new Chinese players to force themselves into the market by flooding it.
I argued it has more to do with Mobile Phone shipment slowed down more than Intel CPU shortages. Intel isn't producing less chips, they were more demand than supply, and you would imagine the demand also means DRAM purchase. And yet this didn't happen. i.e The 5% missing shipment, or actually 20% Q4 YoY decline had more DRAM capacity left than what ever Server + PC market had combined.
Dram is a commodity, generally speaking. More ram will eventually lower prices for end users, but in the meantime it's just more money for the manufacturer. *cough*notacartel*cough*
the problem with ever increasing automation of production is that it increases the fixed cost of production; you can't fire a machine. even if you turn it off, you still have to pay the vig. the only way around the problem is to move more product. the problem with that is the manufacturer doesn't control aggregate demand for its product, so the only out is to reduce price in hopes that demand will follow. not necessarily.
And the slower machine that you replace but paid for just goes in the dumpster? Money well spent. No, you'll stick with you old machine until it's paid off and made a profit. Otherwise it's not a business, it's charity to the company who sells you those machines.
Micron in todays earning call already announced they plan to idle 5% of their capacity to "help" with supply. This is in addition to previous scale backs in production increases that were supposed to come online.
I'm confident you don't know how much semiconductor tools and automation equipment/software costs, nor do you know how much engineers are paid in this sector. I guarantee you that automation isn't driving an increase of production costs. Go study up on how much it costs to DESIGN a device at 28/14/10/7nm for your answer.
Most especially because the tooling for DRAM is generally a whole lot simpler than tooling for logic (tools used for high aspect ratio etch and immersion lithography excluded). DRAM production is truly printing money because it is insanely fault tolerant by virtue of redundant features.
"I'm confident you don't know how much semiconductor tools and automation equipment/software costs, nor do you know how much engineers are paid in this sector. I guarantee you that automation isn't driving an increase of production costs."
are you talkin' to me?? are you talkin' to ME????? :):)
anyway, R&D is mixed into the fixed cost number, just as the cost of the physical capital. again, the variable/marginal cost approach to production and price is the econ's way, not the bean counter's. all the up-front costs are fixed once production starts. the bean counters want to push down average cost; in the case of semi-conductor, they ain't much production labor. there's materials and electricity.
without an increasing demand for whatever eventual user products are aimed at, units of output goes down, and average cost inevitably goes up. the bean counters and C-suits are not happy.
This will hopefully mean lower RAM prices are here to stay (at least for a while): If Samsung can produce 20% more RAM per wafer with similar yields, they'll have lower cost than other producers, meaning they'll be able to undercut competitors while making money. In the absence of truly cartel-like price fixing (which I wouldn't entirely discount), that means there's going to be real price competition.
Samsung themselves have already announced they have nop interest in fighting for market share. All producers have made announcements of production scale back including idling factories to help protect margins with the demand slump.
Memory prices even with the slump are still well above marginal product costs. All lowering production costs do now is help add to manufacturer margins.
As for whether or not there is a cartel the real issue is that the memory market is really only 3 manufacturers. Unless there is a massive breakthrough technology (this isn't it either) none of them have the ability to significantly take market share from each other without resulting into a race to the bottom condition (basically flooding the market). Conversely it's in all their best interests to maintain the status quo and a slight under supply to maintain the highest possible margins and therefore income.
The barrier of entry is also insanely high so you cannot use typical capitalism thinking. Hence the only potential new entrants to the market is those bankrolled by the Chinese government who want such capacity due to other interests (national) and are willing to pump money regardless of profits.
"Memory prices even with the slump are still well above marginal product costs."
for the econ types, a fair argument but not for the bean counters. for them, the metric is <b>average cost</b>, and for such heavily capitalized production, as output falls average cost rises (the cost of the machines is spread over decreasing number of units; labor and materials being minor). the only way to win that battle is for price to increase more than proportionately. and the only way to do that is to collude with other producers.
Yes, damn these charities not giving away free memory.
You realise they need to make profits to justify the investment, right? You realise that prices go up and down, and they need to make hay while the sun shines? You realise that if they don't make reasonable profits, you don't get the tech you desire? You realise... no, you probably don't.
If you look at earnings before taxes in many industries (does not matter if it's tech, finance, energy), then "reasonable" is not the word that comes to mind.
Not saying that companies do not need to make a profit on the goods and services they provide - that's the reason they are doing what they are doing - oh, and of course to make the world a better place for everyone - I almost forgot that.
But there is a point when "reasonable" no longer applies and once it does, this is also the point when it is at the consumers' expense as profits are higher than they should be in an ideal free market economy.
Really funny how Samsung plays around with the "10nm class" technology.
"Samsung says that its 3rd Generation 10 nm-class manufacturing technology (also known as 1z-nm) for DRAM enables it to make 20% more 8 Gb DDR4 memory chips per wafer than the 2nd Gen 10-nm class (aka 1y-nm) does"
1z, 1y, 1*. Basically saying that number replaced by a letter is anything but a 0. They were probably 19nm for 1y and 15-16nm for 1z.
I'll just throw this next to my 10Yottabyte-class SSD (0.00000000000001 Yottabytes for anyone who's asking). Still Yottabyte-class I guess since I managed to use it in the same sentence while being mathematically correct.
First generation is X actually (so 1X), followed by 1Y then 1Z. You are probably right theat 1X was close to 20 (so probably 19 or 18). 1Y in mid 10s and now 1Z in low 10s.
Giving generations like that avoid the marketing names foundries have been using for logic chips. Ex: 16nm, 12 nm (which is really only an optimized 16nm), 10nm... and so on. There is very little relation nowadays between the node name and its critical dimensions.
Being an OLED and presumably pentile screen, does this not mean the effective resolution is 1080x(2/3)= 720p? Seems like such a waste to have effective 720p screen on such a big phone. No matter much the price it is not a good tradeoff IMHO
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
28 Comments
Back to Article
spaceship9876 - Thursday, March 21, 2019 - link
"This should mean cheaper DRAM pricies for consumers." - good joke :PThis will mean higher profits for samsung not lower prices, many ram manufacturers have already decided to reduce production in order to keep prices high. We need more companies producing ram to bring down prices. China has just built some but they are using 22nm or something and only low volume.
ksec - Thursday, March 21, 2019 - link
The price of DRAM has already been in free fall for months are we are already back to late 2016 levels, and on course to reach 2016 or even lowest $/GB Memory in 2019 assuming no further rebounce.These 10nm class DRAM will allow them to sustain their operation and profits even in 2016 DRAM price level, which is good for everyone.
Freeb!rd - Thursday, March 21, 2019 - link
Memory prices are down from their highs, BUT...We are NOWHERE NEAR late 2016 prices, unless you mean high latency 2400 Mhz DDR4. I paid $219 for 2x16GB Trident Z 3000Mhz CL14 DDR4 and bought two sets for a total of ~$440 in March 2017 for a new Ryzen build. Now it's still hard to find CL14 DDR4 3000; mainly because anything with that latency is also sold at higher latency/speeds and prices.
valinor89 - Thursday, March 21, 2019 - link
Considering the current DRAM glut is said to be mainly linked to Intel CPU shortages it might still be reasonable to keep investing in more fabs or new tech as those shortages will not last too long considering the grand scheme of things where product production has a lead up time of years.Is mobile phone, or other non PC, DRAM also getting soo cheap?
close - Friday, March 22, 2019 - link
Product *development* has a lead time. Product *production*, meaning ramping up production, has close to no lead time. How is knowing the shortage won't last encouraging them to invest in new factories? New factories means higher production (so supply), which means lower prices. So you expect them to spend money to build factories that make their product worth even less?Come on, don't they teach any economic concept in school these days?
The only thing that could help now (other than any legal investigation that might prove a price fixing cartel) is for the new Chinese players to force themselves into the market by flooding it.
ksec - Friday, March 22, 2019 - link
I argued it has more to do with Mobile Phone shipment slowed down more than Intel CPU shortages.Intel isn't producing less chips, they were more demand than supply, and you would imagine the demand also means DRAM purchase. And yet this didn't happen. i.e The 5% missing shipment, or actually 20% Q4 YoY decline had more DRAM capacity left than what ever Server + PC market had combined.
e1jones - Thursday, March 21, 2019 - link
Lol, right!?Dram is a commodity, generally speaking. More ram will eventually lower prices for end users, but in the meantime it's just more money for the manufacturer. *cough*notacartel*cough*
FunBunny2 - Thursday, March 21, 2019 - link
the problem with ever increasing automation of production is that it increases the fixed cost of production; you can't fire a machine. even if you turn it off, you still have to pay the vig. the only way around the problem is to move more product. the problem with that is the manufacturer doesn't control aggregate demand for its product, so the only out is to reduce price in hopes that demand will follow. not necessarily.quiksilvr - Thursday, March 21, 2019 - link
You can totally fire the machine by replacing it with a FASTER machine.rpg1966 - Friday, March 22, 2019 - link
I don't think you understand what's involved here.close - Friday, March 22, 2019 - link
And the slower machine that you replace but paid for just goes in the dumpster? Money well spent. No, you'll stick with you old machine until it's paid off and made a profit. Otherwise it's not a business, it's charity to the company who sells you those machines.limitedaccess - Thursday, March 21, 2019 - link
Micron in todays earning call already announced they plan to idle 5% of their capacity to "help" with supply. This is in addition to previous scale backs in production increases that were supposed to come online.FullmetalTitan - Friday, March 22, 2019 - link
I'm confident you don't know how much semiconductor tools and automation equipment/software costs, nor do you know how much engineers are paid in this sector. I guarantee you that automation isn't driving an increase of production costs. Go study up on how much it costs to DESIGN a device at 28/14/10/7nm for your answer.FullmetalTitan - Friday, March 22, 2019 - link
Most especially because the tooling for DRAM is generally a whole lot simpler than tooling for logic (tools used for high aspect ratio etch and immersion lithography excluded). DRAM production is truly printing money because it is insanely fault tolerant by virtue of redundant features.FunBunny2 - Friday, March 22, 2019 - link
"I'm confident you don't know how much semiconductor tools and automation equipment/software costs, nor do you know how much engineers are paid in this sector. I guarantee you that automation isn't driving an increase of production costs."are you talkin' to me?? are you talkin' to ME????? :):)
anyway, R&D is mixed into the fixed cost number, just as the cost of the physical capital. again, the variable/marginal cost approach to production and price is the econ's way, not the bean counter's. all the up-front costs are fixed once production starts. the bean counters want to push down average cost; in the case of semi-conductor, they ain't much production labor. there's materials and electricity.
without an increasing demand for whatever eventual user products are aimed at, units of output goes down, and average cost inevitably goes up. the bean counters and C-suits are not happy.
sing_electric - Thursday, March 21, 2019 - link
This will hopefully mean lower RAM prices are here to stay (at least for a while): If Samsung can produce 20% more RAM per wafer with similar yields, they'll have lower cost than other producers, meaning they'll be able to undercut competitors while making money. In the absence of truly cartel-like price fixing (which I wouldn't entirely discount), that means there's going to be real price competition.limitedaccess - Thursday, March 21, 2019 - link
Samsung themselves have already announced they have nop interest in fighting for market share. All producers have made announcements of production scale back including idling factories to help protect margins with the demand slump.Memory prices even with the slump are still well above marginal product costs. All lowering production costs do now is help add to manufacturer margins.
As for whether or not there is a cartel the real issue is that the memory market is really only 3 manufacturers. Unless there is a massive breakthrough technology (this isn't it either) none of them have the ability to significantly take market share from each other without resulting into a race to the bottom condition (basically flooding the market). Conversely it's in all their best interests to maintain the status quo and a slight under supply to maintain the highest possible margins and therefore income.
The barrier of entry is also insanely high so you cannot use typical capitalism thinking. Hence the only potential new entrants to the market is those bankrolled by the Chinese government who want such capacity due to other interests (national) and are willing to pump money regardless of profits.
FunBunny2 - Friday, March 22, 2019 - link
"Memory prices even with the slump are still well above marginal product costs."for the econ types, a fair argument but not for the bean counters. for them, the metric is <b>average cost</b>, and for such heavily capitalized production, as output falls average cost rises (the cost of the machines is spread over decreasing number of units; labor and materials being minor). the only way to win that battle is for price to increase more than proportionately. and the only way to do that is to collude with other producers.
Soulkeeper - Thursday, March 21, 2019 - link
No promise of lower power consumption ?benzosaurus - Thursday, March 21, 2019 - link
Sounds like it's about time for another tsunami.Notmyusualid - Friday, March 22, 2019 - link
So are we going to get modules with better CL levels, or is this just for THEIR benefit?rpg1966 - Friday, March 22, 2019 - link
Yes, damn these charities not giving away free memory.You realise they need to make profits to justify the investment, right? You realise that prices go up and down, and they need to make hay while the sun shines? You realise that if they don't make reasonable profits, you don't get the tech you desire? You realise... no, you probably don't.
Irata - Friday, March 22, 2019 - link
If you look at earnings before taxes in many industries (does not matter if it's tech, finance, energy), then "reasonable" is not the word that comes to mind.Not saying that companies do not need to make a profit on the goods and services they provide - that's the reason they are doing what they are doing - oh, and of course to make the world a better place for everyone - I almost forgot that.
But there is a point when "reasonable" no longer applies and once it does, this is also the point when it is at the consumers' expense as profits are higher than they should be in an ideal free market economy.
close - Friday, March 22, 2019 - link
Really funny how Samsung plays around with the "10nm class" technology."Samsung says that its 3rd Generation 10 nm-class manufacturing technology (also known as 1z-nm) for DRAM enables it to make 20% more 8 Gb DDR4 memory chips per wafer than the 2nd Gen 10-nm class (aka 1y-nm) does"
1z, 1y, 1*. Basically saying that number replaced by a letter is anything but a 0. They were probably 19nm for 1y and 15-16nm for 1z.
I'll just throw this next to my 10Yottabyte-class SSD (0.00000000000001 Yottabytes for anyone who's asking). Still Yottabyte-class I guess since I managed to use it in the same sentence while being mathematically correct.
frenchy_2001 - Monday, March 25, 2019 - link
First generation is X actually (so 1X), followed by 1Y then 1Z.You are probably right theat 1X was close to 20 (so probably 19 or 18). 1Y in mid 10s and now 1Z in low 10s.
Giving generations like that avoid the marketing names foundries have been using for logic chips.
Ex: 16nm, 12 nm (which is really only an optimized 16nm), 10nm... and so on.
There is very little relation nowadays between the node name and its critical dimensions.
BenJeremy - Friday, March 22, 2019 - link
Am I the only one that remembers the glorious Samsung 4GB Magic RAM sticks from years ago?Remember when $20 for 4GB was a low price in 2012? Glad RAM is so much cheaper these days... wait, what?
Dolda2000 - Friday, March 22, 2019 - link
Any info on what these higher densities mean for things like row hammer attacks?ianmills - Thursday, March 28, 2019 - link
Being an OLED and presumably pentile screen, does this not mean the effective resolution is 1080x(2/3)= 720p? Seems like such a waste to have effective 720p screen on such a big phone. No matter much the price it is not a good tradeoff IMHO