Samsung Announces Solid State Laptop 114
An anonymous reader writes "Samsung has announced they'll be manufacturing solid-state laptops, with an eye for a June release in Korea. Everything you wanted from a laptop: faster boot times, quicker storage access, less noise, longer battery life. Laptop Logic has the story." From the article: "Now to the features of this laptop: Celeron M 1.2GHz, 12.1-inch screen, 512MB DDR2, Wireless LAN 802.11b/g, Digital Multimedia Broadcasting TV, and measuring 2.5 pounds. Price? $3,700 and only available in Korea in June."
Disk Space (Score:3, Informative)
So 32gb of total storage. Not too bad, really.
Re:Seek Time & Reduced Heat (Score:3, Informative)
Re:DIY clone? (Score:4, Informative)
No extra power required and it fits in a 2.5 inch drive bay.
Re:DIY clone? (Score:4, Informative)
Of course, if you have to ask, you can't afford it.
You're Right, I'm an Idiot (Score:5, Informative)
After doing more thorough researching, I found expensive 32GB SSDs but also 16GB SSDs at around $600 here [etech4sale.com] & here [dvnation.com]. I know that size is not always directly related to price but I guess the release of this laptop with 32GB means they've found cheaper ways to produce the 32GB versions.
The $3k price tag is probably pretty reasonable considering that two drives equating to the same size would run around $1,200. Heck, after thinking about the number of writes to the disk they're good for, it might benefit you to have your OS and apps on a drive apart from your userspace drive (a la Unix security layout).
Again, I apologize for not researching my link in the original post and for wasting your time. I only hope the discussion isn't waylayed by people pointing out my ignorance.
Re:Seek Time & Reduced Heat (Score:2, Informative)
Re:Seek Time & Reduced Heat (Score:5, Informative)
There are two major flaws in such a mechanism. The first flaw is that it assumes that you can spread the load evenly among these spares. This varies from manufacturer to manufacturer. We will assume that the flash is made by someone whose algorithm is good. The second flaw is that, AFAIK, generally speaking, existing blocks are not remapped unless they are actually being written. The reason for this is that flash writes are relatively slow. To illustrate why this is a problem, it is necessary to give a fairly concrete example.
Assume that you have a flash device that is full (so that every logical block is mapped to a physical block). Assume that the vendor left an extra 25% overhead for wear leveling. If you rewrite random portions of the disk, there will be significant improvement in longevity, as each block is only rewritten 4/5ths as often. However, if you rewrite the same single block, you only have to wear out the 25% overhead plus one block. Assuming a 32GB disk, that's one million writes times 8GB + 1 block. (I won't speculate on the block size.)
But it's worse than that. You probably don't have a single 32GB flash part. It's probably a bunch of flash parts that are, at most, a gig or two, and in all likelihood, smaller. If each part is only 512MB, you can create a catastrophic failure of that part (effectively preventing any further writes to the part) with only one million writes times 128MB (a fourth of 512MB) + one block. Thus, even if you randomly use all the blocks on a single flash part and ignore all the others (which is typical usage for the first part of the disk, assuming the consecutive flash parts are mapped linearly), that takes the time down from 101 years to less than 2 years. If you do the "repeatedly write one block" technique, that reduces it from a couple of years to under 5 months.
Now all of these problems are, to some extent, solvable, but they would require a cross-chip wear leveling mechanism, coupled with OS intervention (or a relatively smart on-disk controller) to periodically remap random blocks without modification. I'd be surprised if either of these was being done.
In theory, wear leveling makes flash a reasonable alternative to a hard drive. In practice, flash needs to be at least an order of magnitude (and preferably two orders) more reliable before I would ever trust it with anything more than photos in a digital camera... and even for photos in a digital camera, I carefully limit the number of times I will reuse a flash card to ensure that I never run into these sorts of problems.