Posted by
emmett
on from the do-you-suffer-from-long-term-memory-loss dept.
CitizenC writes: "ACK! According to this C|Net story, RAM prices are expected to go up again this week, due to the low supply and high demand. Buy your RAM now!"
I'm not convinced that these RAM price fluctuations are for real. Remember when that plastic factory in Korea burned down a few years back? It was immediately reported that RAM prices would skyrocket, because this was just 1 of 2 factories in the world that made this particular type of plastic for RAM.
Then, about a month later, some interesting articles appeared. Just a few. (This was pre-web-journalism, for the most part, so I can't find anything to link to.) The reports were that the amount of plastic that was in warehouses was unbelievable. There was enough for years of RAM, more than enough time to rebuild the factory and replenish the stocks. Further, this other factory in the world was fully capable of producing enough plastic to satisfy the world's demand. (If anybody has a more accurate recollection of this than I, please correct my errors.)
Yet, still, the industry continued to insist that the price would go up, which it did. Smacked of price-fixing to me, like oil from the mid-east.
It's kinda sneaky everytime something like this happens. When Coca-Cola demand goes up, the price doesn't increase. When carpet demand goes up, the price doesn't increase. Yet somehow we're to believe that the RAM industry is so grossly incompetent as to be unable to adequate predict the demand for their sole product more than a month in advance?
Call me a conspiracy theorist, but it seems weird to me.
When Coca-Cola demand goes up, the price doesn't increase. When carpet demand goes up, the price doesn't increase. Yet somehow we're to believe that the RAM industry is so grossly incompetent as to be unable to adequate predict the demand for their sole product more than a month in advance?
They don't call it "supply and demand" for nothing.
In the case of Coke and Carpets there is an ample supply of both, so both are cheap regardless of demand. If demand goes up it's easy to increase production by hiring more workers and buying more raw materials.
With RAM chips the supply is (supposedly) not ample. Demand fluctuates. Increasing production to meet demand requires building or converting a fab. Ramping up production to meet demand takes a while, during which time prices are high.
Disclaimer: IANAE (I am not an economist).
High SDRAM prices are bad for Rambus
by
Guppy
·
· Score: 5
On the surface, it sounds like a SDRAM price hike would be good for RDRAM, as it would close (slightly) the price gap between the two sorts of memory. In reality, though, this is bad for Rambus. There are plenty of other issues with Rambus (die size, yield, packaging, royalties, etc...), but the biggest one in simple economics.
A year or two ago, the big carrot that Rambus had to offer to manufacturers was a potentially profitable product. SDRAM prices were in the dumps, so nobody was making any money no matter how much they could churn out. Fast forward to present day, and just about every fab is guaranteed a nice fat profit, just for producing a commodity chip. The only way to justify the risk and cost of switching production from SDRAM to RDRAM is if RDRAM fetches a huge margin. Hence low supplies and sky-high prices for Rambus.
Intel has been pushing manufacturers to "voluntarily" cut their margins and drop prices on RDRAM. For once, though, it seems the memory makers have the upper hand, and there have been no takers so far.
High demand fueled by these rumors
by
gklyber
·
· Score: 5
Of course, with articles like this, everyone will go out and buy RAM. There may be high demand right now, but these rumors will make demand skyrocket and make the prices even higher. Somebody at C|Net must want their memory stock to improve in value.
Re:High demand fueled by these rumors
by
Betcour
·
· Score: 4
Out of the last 5 RAM price hike, the IT press has successfully predicted 47 of them.
Contract prices for the standard 64-MB chip, now at $6, will rise about a dime
Now, correct me if I'm wrong, but isn't a dime only a few cents? So this isn't like the doubling in price which happened last September.
--
perl -e 'fork||print for split//,"hahahaha"'
$0.10 more a stick? Try $0.10 more a chip.
by
Marcus+Aanerud
·
· Score: 4
I think the article is confusing "MB" and "Mb". Megabytes is what we typically measure our computer storage in. Megabits is what chip vendors measure the individual chips by. 8 Megabits = 1 Megabyte. What's that mean? If the price of a 64Mb chip is $6, you need 16 of those chips to make a 256MB DIMM, and 8 of them to make a 64MB DIMM. That's $96 for the actual RAM chips on a 256MB DIMM, and $48 for the chips that go on a 64MB DIMM. That sounds a lot more realistic.
So, if there's a price hike from $6 to $6.10, the cost of the 256MB module is now $97.60. Up $1.60. Not too big of a deal until you consider markups, profit margins, and hype-stories like the one on c|net. If people see that story, they're bound to believe it, and rest assured, the good resellers are going to be reading articles that concern them (like this one).
And, of course, you need to add to the price of the RAM chips assembly of the actual DIMM, the circuit board that the chips are put on, testing of the board, distribution, further markups in retail, and so on. All together, that maks a 256MB module about $200-$270, depending where you buy.
May I suggest the RAM price index:
by
Wakko+Warner
·
· Score: 4
Go here to see if prices really do go up. This page tracks and graphs the prices of CPU, SDRAM, and RDRAM and has done so for a couple of years now. Very interesting site.
- A.P. --
"One World, one Web, one Program" - Microsoft promotional ad
-- "Remember when the U.S. had a drug problem, and then we declared a War On Drugs, and now you can't buy drugs anymore?"
Your standard 64MB Chip costs $6.00 from Micron. That price is going up about 10 cents. The consumer purchases a 64MB chip for about $126.00. (Based on Simple 64MB EDO 168-pin DIMM)
Now, $6.00 goes into $126.00 21 times. Take $.10 and multiply it by 21, and you get $2.10.
So in a worst-case scenario (being the company charges you DOUBLE the increase in price) the price of a 64MB DIMM from Simple Technology goes from $126.00 to $130.20. I'd hardly call that skyrocketing. Even if you pay QUADRUPLE the price increase, it's still under $10 more.
-- Give him Head? Be a Beacon?
--
-- Give him Head? Be a Beacon?
(If you can't figure out how to E-Mail me, Don't.:P)
Obligatory prediction poll
by
PollMastah
·
· Score: 3
What is the most likely Slashdot article to appear next?
Yet Another Fiasco involving Napster
Yet Another {DMCA, RIAA, UCITA, insert-your-favorite-enemy} bashing session.
Famous Person X (or not-so-famous person X) writes something inflammatory about {Napster, Open Source, Linux, GPL}. News site gets slashdotted, Slashdot gets filled with zealot posts.
You can run your Linux box off potatoes! Serious!
RAM prices are dropping! Run for cover! CPU prices are rising! Ride the tide!
Scientists have discovered Yet Another Innovation! Rabid slashdotters rush to post uninformed, nonsensical posts that get modded up to +5!
Hemos re-posts something redundant! Addicted slashdotters complain that slashdot isn't reading itself, but they're too hooked to find a better site!
Slashdot posts a movie review. Sane slashdotters complain that the movie sucks. Insane slashdotters elaborately explain why Natalie Portman is hotter than the lead actress.
--
Poll Mastah
Re:Obligatory prediction poll
by
swinge
·
· Score: 4
Addicted slashdotters complain...
9. carefully constructed parody of slashdot appears and gets moderated up, ostensibly making fun of slashdot but clearly mired in it, simultaneously complaining about yet adding to the noise.
slashdot would be better if we ditched all the emotion. Emotion is stuff that does not matter.
Nothing makes prices skyrocket like panic buying
by
redelm
·
· Score: 5
C'mon/. , you're playing right into the hands of the DRAM mfrs. They don't like the current low prices and have lost a bundle on Rambus. So they're taking this low-profit time to do some maintenance/retooling on some fabs. Perfectly normal.
They've also got lots of chips in inventories to meet sales. A plant fire when demand is tight is one thing, but a planned shutdown when demand is slack is quite another.
But starting a buying panic is very much in their interest. CNet bought the story, hook-line&sinker. Now you. Fortunately, the hobbyist market is fairly small, and I doubt can move prices. The big OEMs (Dell, Compaq, HP, IBM) are smarter than to fall for this.
It's fairly obvious that prices are not 'skyrocketing'. But if they did, as happened last year, and stayed high, what would happen?
When software was just beginning to get bloated (I remember being shocked that CorelDraw required a massive 20Mbyte), it became expensive to get a large enough hard disk (80Mbyte or so). Disk prices, at least at first, didn't drop enough for consumers to keep up with the increased demand for disk space. This led to a brief period of disk compression programs such as Stacker, which slowed down your machine a little but let you fit about 60% more stuff on your disk (typically). Eventually {Double,Drive}Space was included as a standard feature of MS-DOS (and then Win95). But now that disks are so cheap, I doubt if anyone bothers with it on a new system.
Similarly, a couple of years later, starting just before the launch of Win95, it started getting expensive to fit a system with the amount of RAM needed to run modern bloated applications. Products such as RAM Doubler appeared for the PC and Mac. (The PC version of one such program turned out to be a total fraud - it didn't compress things at all - but the others did actually do something.) In common with the disk compressors, the 'Double' in the name is misleading; you don't get anywhere near the same performance as a system with twice the RAM.
The way these RAM compressors (or at least Quarterdeck's offering) work is to set aside an area of memory, say 25% of physical RAM, for compressed pages. When the remaining memory gets full, pages are compressed and stored in the buffer instead of being paged to disk. When the buffer gets full you do have to go to disk, but you can write compressed pages and several of them at once, so disk activity is less. Compressing the pages takes CPU time, but most PCs have fast CPU relative to disk speed so the speed burden isn't too great.
Having said that, I didn't notice any wonderful speedup from Quarterdeck's RAM booster (can't remember the name) on a 16Mbyte machine, and it made Windows less stable. But done right, it might work.
Could you do something similar for Linux? One tactic might be to create a RAM disk with some of your memory, and use something like the crusty old DoUbLe (or however it's spelt) code to make it a 'compressed partition'. Then get Linux to swap to that before going to disk.
However you wouldn't get the benefit of pages being compressed when they go to disk, nor of compressed pages being simply moved from the RAM-swap to the swap partition. AFAIK you can tell Linux to use one swap device in preference to another, but you can't ask for a tiered swap scheme where pages from one device spill onto the next. (Please correct me if I am wrong.)
Another problem is that with DoUbLe, the size of the device varies according to how well things have compressed so far, and swapping requires a fixed-size device. So none of this would probably work at all.
But there are plenty of ultra-fast compression algorithms out there (LZO is fast and GPLed) which you might stick in the kernel. Then get the kernel to compress pages and dump them in a big pile somewhere, and to swap them out only when there is no alternative (or when the system is idle).
I have no idea whether such a scheme would make a significant difference; I'd guess that many of the most bloated programs do have highly compressible data areas though. If it did work, then Linux would be able to survive a memory shortage better than other OSes. It might also help ultra-cheapskate manufacturers making thin clients or mobile devices.
Slashdot Mirror
I'm not convinced that these RAM price fluctuations are for real. Remember when that plastic factory in Korea burned down a few years back? It was immediately reported that RAM prices would skyrocket, because this was just 1 of 2 factories in the world that made this particular type of plastic for RAM.
Then, about a month later, some interesting articles appeared. Just a few. (This was pre-web-journalism, for the most part, so I can't find anything to link to.) The reports were that the amount of plastic that was in warehouses was unbelievable. There was enough for years of RAM, more than enough time to rebuild the factory and replenish the stocks. Further, this other factory in the world was fully capable of producing enough plastic to satisfy the world's demand. (If anybody has a more accurate recollection of this than I, please correct my errors.)
Yet, still, the industry continued to insist that the price would go up, which it did. Smacked of price-fixing to me, like oil from the mid-east.
It's kinda sneaky everytime something like this happens. When Coca-Cola demand goes up, the price doesn't increase. When carpet demand goes up, the price doesn't increase. Yet somehow we're to believe that the RAM industry is so grossly incompetent as to be unable to adequate predict the demand for their sole product more than a month in advance?
Call me a conspiracy theorist, but it seems weird to me.
-Waldo
On the surface, it sounds like a SDRAM price hike would be good for RDRAM, as it would close (slightly) the price gap between the two sorts of memory. In reality, though, this is bad for Rambus. There are plenty of other issues with Rambus (die size, yield, packaging, royalties, etc...), but the biggest one in simple economics.
A year or two ago, the big carrot that Rambus had to offer to manufacturers was a potentially profitable product. SDRAM prices were in the dumps, so nobody was making any money no matter how much they could churn out. Fast forward to present day, and just about every fab is guaranteed a nice fat profit, just for producing a commodity chip. The only way to justify the risk and cost of switching production from SDRAM to RDRAM is if RDRAM fetches a huge margin. Hence low supplies and sky-high prices for Rambus.
Intel has been pushing manufacturers to "voluntarily" cut their margins and drop prices on RDRAM. For once, though, it seems the memory makers have the upper hand, and there have been no takers so far.
Of course, with articles like this, everyone will go out and buy RAM. There may be high demand right now, but these rumors will make demand skyrocket and make the prices even higher.
Somebody at C|Net must want their memory stock to improve in value.
The semiconductor industry is about to be slashdotted!
Now, correct me if I'm wrong, but isn't a dime only a few cents? So this isn't like the doubling in price which happened last September.
perl -e 'fork||print for split//,"hahahaha"'
I think the article is confusing "MB" and "Mb". Megabytes is what we typically measure our computer storage in. Megabits is what chip vendors measure the individual chips by. 8 Megabits = 1 Megabyte. What's that mean? If the price of a 64Mb chip is $6, you need 16 of those chips to make a 256MB DIMM, and 8 of them to make a 64MB DIMM. That's $96 for the actual RAM chips on a 256MB DIMM, and $48 for the chips that go on a 64MB DIMM. That sounds a lot more realistic.
So, if there's a price hike from $6 to $6.10, the cost of the 256MB module is now $97.60. Up $1.60. Not too big of a deal until you consider markups, profit margins, and hype-stories like the one on c|net. If people see that story, they're bound to believe it, and rest assured, the good resellers are going to be reading articles that concern them (like this one).
And, of course, you need to add to the price of the RAM chips assembly of the actual DIMM, the circuit board that the chips are put on, testing of the board, distribution, further markups in retail, and so on. All together, that maks a 256MB module about $200-$270, depending where you buy.
- A.P.
--
"One World, one Web, one Program" - Microsoft promotional ad
"Remember when the U.S. had a drug problem, and then we declared a War On Drugs, and now you can't buy drugs anymore?"
Your standard 64MB Chip costs $6.00 from Micron. That price is going up about 10 cents. The consumer purchases a 64MB chip for about $126.00. (Based on Simple 64MB EDO 168-pin DIMM)
Now, $6.00 goes into $126.00 21 times. Take $.10 and multiply it by 21, and you get $2.10.
So in a worst-case scenario (being the company charges you DOUBLE the increase in price) the price of a 64MB DIMM from Simple Technology goes from $126.00 to $130.20. I'd hardly call that skyrocketing. Even if you pay QUADRUPLE the price increase, it's still under $10 more.
-- Give him Head? Be a Beacon?
-- Give him Head? Be a Beacon? :P)
(If you can't figure out how to E-Mail me, Don't.
What is the most likely Slashdot article to appear next?
Poll Mastah
C'mon
They've also got lots of chips in inventories to meet sales. A plant fire when demand is tight is one thing, but a planned shutdown when demand is slack is quite another.
But starting a buying panic is very much in their interest. CNet bought the story, hook-line&sinker. Now you. Fortunately, the hobbyist market is fairly small, and I doubt can move prices. The big OEMs (Dell, Compaq, HP, IBM) are smarter than to fall for this.
It's fairly obvious that prices are not 'skyrocketing'. But if they did, as happened last year, and stayed high, what would happen?
When software was just beginning to get bloated (I remember being shocked that CorelDraw required a massive 20Mbyte), it became expensive to get a large enough hard disk (80Mbyte or so). Disk prices, at least at first, didn't drop enough for consumers to keep up with the increased demand for disk space. This led to a brief period of disk compression programs such as Stacker, which slowed down your machine a little but let you fit about 60% more stuff on your disk (typically). Eventually {Double,Drive}Space was included as a standard feature of MS-DOS (and then Win95). But now that disks are so cheap, I doubt if anyone bothers with it on a new system.
Similarly, a couple of years later, starting just before the launch of Win95, it started getting expensive to fit a system with the amount of RAM needed to run modern bloated applications. Products such as RAM Doubler appeared for the PC and Mac. (The PC version of one such program turned out to be a total fraud - it didn't compress things at all - but the others did actually do something.) In common with the disk compressors, the 'Double' in the name is misleading; you don't get anywhere near the same performance as a system with twice the RAM.
The way these RAM compressors (or at least Quarterdeck's offering) work is to set aside an area of memory, say 25% of physical RAM, for compressed pages. When the remaining memory gets full, pages are compressed and stored in the buffer instead of being paged to disk. When the buffer gets full you do have to go to disk, but you can write compressed pages and several of them at once, so disk activity is less. Compressing the pages takes CPU time, but most PCs have fast CPU relative to disk speed so the speed burden isn't too great.
Having said that, I didn't notice any wonderful speedup from Quarterdeck's RAM booster (can't remember the name) on a 16Mbyte machine, and it made Windows less stable. But done right, it might work.
Could you do something similar for Linux? One tactic might be to create a RAM disk with some of your memory, and use something like the crusty old DoUbLe (or however it's spelt) code to make it a 'compressed partition'. Then get Linux to swap to that before going to disk.
However you wouldn't get the benefit of pages being compressed when they go to disk, nor of compressed pages being simply moved from the RAM-swap to the swap partition. AFAIK you can tell Linux to use one swap device in preference to another, but you can't ask for a tiered swap scheme where pages from one device spill onto the next. (Please correct me if I am wrong.)
Another problem is that with DoUbLe, the size of the device varies according to how well things have compressed so far, and swapping requires a fixed-size device. So none of this would probably work at all.
But there are plenty of ultra-fast compression algorithms out there (LZO is fast and GPLed) which you might stick in the kernel. Then get the kernel to compress pages and dump them in a big pile somewhere, and to swap them out only when there is no alternative (or when the system is idle).
I have no idea whether such a scheme would make a significant difference; I'd guess that many of the most bloated programs do have highly compressible data areas though. If it did work, then Linux would be able to survive a memory shortage better than other OSes. It might also help ultra-cheapskate manufacturers making thin clients or mobile devices.
Does any of that make sense?
-- Ed Avis ed@membled.com