Slashdot Mirror
Samsung's 64-GB Solid-State Drive
Anonymous Howard writes "Just a couple of weeks ago Sandisk introduced a 32-GB solid-state drive. Now Samsung has one-upped them, unveiling a 64-GB solid-state drive. They are expecting to begin shipping in the second quarter of this year. Samsung says the device can read 64 MB/s, write 45 MB/s, and uses just 0.5 W when operating (0.1 W when idle). In comparison, an 80-GB 1.8-inch hard drive reads at 15 MB/s, writes at 7 MB/s, and consumes 1.5 W when either operating or idle. No pricing yet."
This would be perfect for my iRiver H320 MP3 player, since (according to TFA) it's in the 1.8" form factor which almost every HD MP3 player uses.
How can it be one-upping them A-DATA already annouced 128GB SSDs two months ago?
They're marketed as drop in replacements, currently built to notebook drive standards. The downside, as someone else mentions, is that they are flash based. While flash has gotten better recently, I'd be squeemish about having an OS that constantly writes to the drive even when nothing is (apparently) happening on it.
"It is a miracle that curiosity survives formal education." -Albert Einstein
Not totally certain, but I would guess so. Depending on how big the pagefile was, perhaps the drive does some intelligent load-balancing that would keep you from frying it too quickly, but it might be better to keep another drive around for that, or loading the machine up with enough RAM to keep it from swapping often.
I used to know people who swore that, after adding RAM, the best thing you could do for speed was to add a small-but-fast SCSI hard drive and use it for nothing but your swapfile. I've never gone that route personally, because I've never thought it worth the expense, but I bet it would make for a pretty nice system. And I also suspect there have to be a lot of SCSI drives on the used market if you look.
"Ladies and gentlemen, my killbot features Lotus Notes and a machine gun. It is the finest available."
Hm, based on the cheapest (without rebates) memory available at $8.50/GB, figure 20% markup between the manf and retailer, thats $6.8/GB.
e _digital_2gb.htm
$435 for memory
+10% for R&D
+10% for manf (including controller, parts, etc)
-10% for manf efficency when producing 64GB/run
COST $479
RETAIL:
+20% for geewhiz-newtoy-factor/supply shortages
+10% for retail
YOUR COST: $630
sources:
http://www.pricewatch.com/flash_card_memory/secur
Another prediction: SSDs will offer such huge power and performance advantanges, they will sell like crazy and drop in price by a factor of 70% within 1 year from now.
But doesn't perform as well when used by an operating system.
Since flash doesn't have sectors that are faster than others; Thus, this is incorrect.
Flash chips each have a read and write speed limit the more of them you have in parrelel the faster it can read/write. It's trivial to make the chips within a flash drive have JBOD(F) properties.
This is one of the major advantages for me, disks that will be able to max out gigabit+ ethernet with increadible seek times, data redundancy, and massive througput.
As disks get bigger it may become nescessary to have some space for a read/write buffer (normal HD's have ram for this) which will increase the life or need for higher MTBF sections, both of these properties are showing up in variations on flash.
So if you have a flash disk with 1 Increadible MTBF chip, 1 super speed no storage sector (like ram), 1 massive storage space, and a bunch of standard flash you can have all the advantages of every kind of disk with the internal controller handling performance and wear leveling (not a trivial programming problem but one which we have a bunch of excellent solutions in place for).
My personal problem with flash disks is that industry seems to be holding back development, trying to develop an upgrade cycle instead of realeasing a perfect solution.
I can get a 1GB microsd flash card for $15 about 400-600(conservative) of them would fit into a 2.5 disk enclosure. With JBOD and wear leveling across the chips and I'm assuming it would be cheaper because you wouldn't need hundreds of cases/interfaces a 200GB drive with read/write speeds of 100-300 Gb a sec and seek time of
Hmm, well maybe the price does need to come down but the other concerns about flash seem unjustified, write wear isn't a problem, it's not scary. losing all your data to a HD failure, now that's scary.
The numbers you used, approx 900/100, are also a special magic point on 32-bit CPUs under Linux. Above about 960MB, Linux uses "highmem" mode on x86, and that slows things down. A 32-bit x86 PC runs faster when you restrict it to 960MB instead of letting it use the full 1024MB.
For those of you who wonder how a computer could run faster w/ a little swap in RAM instead of just using all the RAM, the answer is complicated. Mainly, all the VM algorithms assume the existance of swap, and so when they get backed into a corner, they expect to be able to dump a bunch of stuff overboard into swap. They only start making the really hard choices once swap fills up. If you take away swap, then you hit the "out of swap" condition much more readily.
You might be thinking "ah, but it's all just a shell game! You'll still run out of swap at the same time, since your total memory is fixed!" Not true. The OS prioritizes disk buffers and other caches relative to the work it's doing and the RAM available to it. RAM dedicated to a RAM disk is not available for other purposes. Thus, a RAM-based swap partition dedicates some portion of RAM to only hold dirty program pages. No disk buffers, no network buffers, no inode information. Just dirty program pages. By forcing austerity on these other discretionary structures, you can compensate for the VM's inbuilt assumption it can just "dump things to swap", and that running out of swap occurs "almost never."
--JoeProgram Intellivision!