Slashdot Mirror

Maybe Apple will even let you upgrade this system...!

We've heard that one of our customers bought some bad Ram or SSD or something and had a bad experience and also they like thin machines. So we've decided to solder it all down.

Still for a hefty premium you can order the machine with extra Ram.

https://www.apple.com/shop/buy...

How does $400 for an extra 8GB Ram and 128GB of SSD sound like?

Now you'll say "That sounds expensive. I could buy the RAM for $91.76

https://www.newegg.com/Product...

And a 128GB SSD for about $133

https://www.newegg.com/Product...

Presumably Apple buy stuff cheaper than I do. E.g. look at the spot prices for NAND and SSDs here

http://www.dramexchange.com/

So why is it it $400? Other than that you've got me over a barrel.

And, more to the point if the system were upgradeable I could buy the unupgraded version now and buy more Ram and a larger SSD later when/if I need them. At which point they'll will be cheaper.

Like I did with my Macbook Pro 2012. That was about $1200. Now for a future proof machine I need to pony up $1699.

Suddenly Windows 10 doesn't seem so bad. I bet I could even get XCode running in a VM on an Asus laptop if I had to.

1Gbit see line 4 http://www.dramexchange.com/
128M....thx :-)

Hmmm.... how about buying large quantities? A "DDR2 1Gb 128Mx8 800MHz" chip costs less than 2 US$ on the spot market. So you can get 2GB for below 4 bucks as manudacturer (Source: http://www.dramexchange.com/).

CU, Martin

A new RAM format means the price of the old format declines a little, stagnates, then climbs up incessantly.
http://www.dramexchange.com/

I misread http://www.dramexchange.com/

Actually MLC flash for flash drives is awesomely cheap 32Gb 4Gx8 MLC averages at $6.30. So $1.57 per GB. So a drive to hold a movie would be worth $16.

But it's still much more expensive than a pressed DVD.

Flash will kill hard disks in notebooks at these prices though.

Regular flash works just fine for swap. If you write nonstop at top speed to a standard chip, you'll wear I'd out in about fifty years. Thus I don't understand why we should care about an even longer lifetime.

That used to be true with SLC chips. It's not true with MLC.

http://www.storagesearch.com/ssd-slc-mlc-notes.html

It's a simple matter to plug new data for MLCs into the calculation I did for the worst case wear-out process for flash SSDs - which I called the Rogue Data Recorder.

Instead of the 64GB example I used then, I'll assume the MLC SSD has 128GB capacity. MLC SSDs have more capacity than SLC. And more capacity means longer operating life - before cells wear out.

I'll still use the 80M bytes / sec sustained write speed - because the fastest MLC products (in Feb 2008) can already do that. (Meanwhile the fastest SLC products have moved up in the world and are about 50% faster.)

The next factor is where we hit the big problem... Instead of a write endurance rating of 2 million cycles (for the best SLC) - I can only use a figure of 10,000 for MLC. MLC has a much lower rating due to the complex interaction of discriminating multiple logic levels reliably coupled with the intrinsic failure mechanism of wear-out.

Plugging these numbers in the same calculation gives an estimated MLC flash SSD operating life (at max write throughput) which is 6 months! (instead of 51 years for a 64GB SLC SSD).

All the affordable SSDs I've seen from Intel and Samsung are based on MLC flash because it costs much less per bit. Down to $2 per GB in fact. SLC currently costs 2-4x as much. E.g.

Here are the average prices for flash

32Gb 4Gx8 MLC 9.27
16Gb 2Gx8 SLC 15.61
16Gb 2Gx8 MLC 3.97
8Gb 1Gx8 SLC 6.31
8Gb 1Gx8 MLC 2.34

SLC is 2.7x more expensive for 1Gx8 and 3.9x more expensive for 2Gx8. So it's not surprising that most SSDs are MLC based. But if you write at full speed to them they will die very quickly.

Incidentally look at the price of 4Gx8 MLC. $2.31 per gigabyte. Pretty damn cheap.

Anyone can check the spot price for flash anytime by looking at this site:

http://www.dramexchange.com/

Scroll down to the flash section.
SLC is the good stuff used in the big fast SSD's you get from people like Apple.
MLC is the slower, less long lasting, stuff commonly used in thumb drives.
$2.08 for a Gigabyte in MLC
$6.70 for a Gigabyte of SLC

If you want to know the long term price improvement rate for flash, you can join that site for $1000 a year or if you want the cheap version, I've been tracking retail flash (MLC) prices for 9 years at my site here:
http://www.mattscomputertrends.com/flashmemory.html

29$ a GB and you are modded informative? Here is a report reflecting the current Nand Flash prices. Shows that its not far off in future that these things will be available for desktops on a reasonable and competitive price.

Nand flash is about $7 per GB

http://www.dramexchange.com/#fspot

So 640GB should be about $4480. That's the spot price on the chips too, a device with them in will be more expensive. And it would need to be a hell of a device. 160*32Gigabit (4Gbyte) devices for example. That would be a hefty PCB. And the controller that performs the way they say is not going to be cheap either.

Contract Price for two 2GB flash chips is $31.20 http://www.dramexchange.com/ so I don't think the parent poster is that off. Soldering them in with whatever interface they need is probably less than $5.

You're also neglecting a few things--like physical size, weight, reliability, and power. Hard drives are by far the largest cause of laptop returns, so flash keeps the overall price lower. A 2.5" drive would not fit in such a small laptop, so you would need a 1.8" or smaller drive, which would cost more than (your rather optimistic) $30, and be quite a bit slower to boot (no pun intended...) Flash's much lower power consumption means a cheaper, lighter, battery; and makes cooling of the device easier (cheaper!)

Where the hell have you people been, DRAM prices are at the bottom now . If anything, DDR2 is going to go up as manufacturers switch to DDR3, just as happened with plain old DDR once upon a time.

And no, DRAM isn't "always at their lowest". Here's a clue; go to dramexchange and check their handy graph at the upper left-hand corner.

If you're out to get some DDR2, today is the day to shop.

As I understand it, over 2GB SD cards (i.e. 4GB upwards) use the new SDHC standard which is partially backwards compatible, i.e. uses the same form factor and interface, but older standard SD 1.0 or 1.1 devices - i.e. most of them from 2006 or earlier - can't read SDHC cards. Obviously, new SDHC devices can generally read MMC or older SD cards without issue.

Unfortunately, the quote you made is quite right, except that you want to write at a rate 10k greater in order to beat out hard disks, so that 0.744 years instead of 744, which doesn't look so hot (or 7 years if you're willing to accept a drive that is almost as fast as a conventional drive, which is almost acceptable).

I measured the write rate on my desktop machine, which was building software at least half the time. It's the average write rate, not the peak one. Even though a modern hard disk can write at 30MB/sec or whatever, the average write rate is much, much lower. And it's the average write rate that you should use in that equation. As I point out, on my laptop where I do web browsing and email, the average write rate is lower still.

You also assume that wear levelling means getting the statistical average number of writes out of every cell, which it doesn't. Realistically, that factor chops off roughly half of your write capacity.

The 100K writes is what the manufacturer guarantees. Since they don't want to get sued, they hopefully understate things a bit. From what I've heard, most flash erase units can do much more than this. In fact, most of the EU will be ok for more than 100K writes, you just need to swap out sectors that go totally bad from a replacement pool. Some of them will just get a few stuck bits but you can do ECC so 1 bit / sector errors are correctable, only ones with more than one bit stuck need to be swapped. Finally, you have to bear in mind that modern hard disks are not that reliable. I've had several desktop ones fail after around 18 months. Old laptop drives seem to have head crashes too, pretty much at random no matter how carefully you avoid walking around with them running. And the life is proportional to disk size, you can double the life by doubling the size of the disk. That last part should solve the problem on its own, given Moore's law.

If you had a wear levelled flash disk, the failure would be 100% predictable - the SMART data could give you a predicted fail date based on when the replacement block pool will be exhausted. I think this is better than the unpredictable failure you get from hard disks, which seem to go from 100% ok to dead without any warning from SMART.

On the other hand, 10^8 writes with pram will be plenty.

Yeah, I agree. Actually, I read some other article where an Intel guy said that the phase transitions don't inherently wear the cell like writing or erasing flash does - in fact he more or less implied that PRAM didn't wear out. But other places quote 10^8 or "trillions of writes" which sounds like more than 10^12. Maybe more recent PRAM has an longer life than 10^8 writes, or future PRAM will.

The only problem I can see is that it's possible to buy NAND flash now for $7 per GB. I haven't seen any quotes for PRAM. If it starts off being expensive and lags NAND chip capacity, it might not catch on.

At the moment, 32GB of NAND flash is about $580.

http://www.dramexchange.com/

A 40GB hard disk is about $58. Which is a pretty bad price differential.

But the cost of NAND flash should drop 43% per year over the next 5 years

http://blogs.zdnet.com/ITFacts/index.php?blogthis= 1&p=9615

So that 32GB NAND flash device should be $580 * ( 0.57 ^ 5 ) = $34. That's less than a small hard disk, and you can probably charge a premium based on the low power consumption / small form factor. My guess is that people will pay $100-$200 premium for a NAND flash laptop, so I'd expect to see 32GB NAND flash ultraportables in the next couple of years.

I checked the write rate to the Physical Disk on Windows in perfmon (this counts writes to the block device driver) on my work machine, and it's not too bad, about 100K/sec averaged over an hour of building a massive chunk of software. Mainstream desktop use should have a much lower rate. E.g the laptop I'm writing this on is writing at 3K-20K/sec average at the moment. Most of that is short 40K spikes, with a small duty cycle too, so an daily average should be lower.

This paper has a formula for disk lifetime
http://www.st.com/stonline/products/literature/an/ 10122.pdf

i.e. lifetime in days is

Size of flash * Number of erase cycles * FS overhead
/
Bytes written per day.

Let's assume that the 32GB array is 8*32Gbit chips. Each chip is then 4GByte

Plugging in the figures, we get

4GB * 100,000 * 0.7
/
100KB * 24 * 60 * 60

I get 33981 days, or 93 years. This is mostly because with wear levelling, the lifetime is proportional to NAND flash array size, since wear levelling should spread erases evenly. If you could wear level over the whole 32GB, it would be 744 years!

It's actually pessimistic, since the fs overhead is lower on Inode based filesystem like NTFS or ext2/3. E.g in their example, the file needs 5000 clusters in the FAT, or 20K. On NTFS, file extents are stored as runs - i.e. a contiguous file will be one extent in the 1K Inode. Most of the time, file extents are not growing too.

At the moment, the fastest NAND has a per chip write speed of 10MB/sec, and reads about 100MB/sec. I think you could do some kind of RAID like approach, especially for reads. And since we're talking about a 8*32Gbit chips for this device the read speed should be 800MB/sec, neglecting speed ups from new interfaces and process shrinks. And bad blocks could be tracked on a finer granularity too, so you could continue to use a erase unit until all the blocks went bad. And 1 million erase cycles is probably not impossible to achieve in 5 years, since some NAND devices do that as you say.

People often quote pagefile writes as a reason for NAND being unusable, but I'm not really convinced. Even with a completely unmodified system, a worst case filesystem choice (FAT32 with small clusters), and an application which constantly grows files, and a NAND chip where every block has a 100% fail rate after 100K erases, the lifetime of a large NAND disk is still orders of magnitude greater than a hard disk, which is likely to die in 5-10 years or so at this level of usage, since it wears from both read and write.

Take a look here.. this is a good place to check the "memory" spot market.. note, these are DRAM prices, not DIMM or memory stick or flash card prices..

If you notice, the SDRAM "chips" (I'll use "chips" versus DRAM to avoid sounding like I have a stutter) cost about as much as DDR chips, and in some cases, even more.. how this relates to your question is, Flash capacity is coming up real fast, which leads to competition, which leads to price drops.. DDR, on the other hand doesnt have the sudden increase in rise of DDR output like Flash does (in fact, if you are familiar with the DDR spot market, the price has been steadily rising recently due to shrinking supply so the price should be going up as I type).. on the other hand, you have shrinking capacity for SDRAM (since most manufacturers would rather produce DDR wafers over SDRAM wafers) which has led to an increase in price from what many of us may remember paying for, say, a 256MB stick of SDRAM..

To figure out a rough estimate of how much a stick of RAM will cost based on the spot price, take the DRAM price, say $4.78, multiply by 8 ($38.24), add like $1 for logic parts (caps, resistors, etc) and labor (significantly less if the memory module is manufactured in China).. and you arrive at a "cost" of about $39.24.. tack on 5-10% for profit.. $41-43 for a 256MB DDR module.. today

In short, its all supply and demand..

This is a bit late since memory prices have been trending down in the past few days. Take a look at http://www.dramexchange.com where they present pricing in a stock-chart-like format. Seems like a great way to track the pricing on something as commoditized as PC components.

64 bits should be enough for anyone.

No really, I mean it.

Clever, Ed. For those who don't get it, he's quite right: 64 bits *will* be enough for anyone.

For those still stuck in mid-90's video game wars, "bit-edness" in the real world refers (technically) to the size of your general purpose integer registers, which, for most intents and purposes, refers to how many memory addresses you can easily and quickly address. 32 bit addressing tops out at 4GB, a value which is often too small for e.g. large databases, which thus tend to live on 64-bit big iron machines. (MS has a hack to give x86 processes access to 36 bits of space, but it requires OS intervention.)

64 bits, on the other hand, works out to 16 billion GB. (That's 16 exobytes IIRC.) For reference, that's roughly 40 times as much memory capacity as there currently is DRAM produced (of all types, for all markets) worldwide in a year, at this January's rate.

I don't have the figures on hand for hard drive production, but I would guess as a first approximation that 16 billion GB is not quite equal to the total number of bits of digital storage of all kinds manufactured throughout computing history up until today. (I'd guess it's too small by a factor of 3 or so.)

In other words, it's quite a lot. Presumably computing will have run into some very different paradigm (wherein the bit-edness of the "CPU" is no longer an applicable term) before any computer has a use for >64 bit addressing.

(FWIW, today's 64-bit processors don't offer all 64 bits of data addressing yet, because no one has a need for more than 40-something, so that's what they offer.)