Samsung's Solid-State Disk Drive Unveiled

Iddo Genuth writes "After unveiling their upcoming hybrid hard drive, Samsung — along with a number of other manufacturers — is planning to begin shipping solid-state drives during 2007. Unlike the upcoming hybrids, solid-state drives should work with windows XP as well as Vista." The drives will be introduced in 1.8- and 2.5-inch form factors for notebooks. While streaming performance can't equal that of hard disks, Samsung claims that random-access performance is more important and that (e.g.) Vista users would see a 4x speedup in many key operations. Pricing was not announced.

Not on XP?

Could someone tell me why one type of drive wouldn't work with a specific version of Windows? Shouldn't they be able to write drivers for that?

Re:Not on XP?

Could someone tell me why one type of drive wouldn't work with a specific version of Windows? Shouldn't they be able to write drivers for that?

Ah, you must be new here. It's not that it wouldn't work, it just doesn't, you dig? No? Well, here's a Vista t-shirt.

Re:Not on XP?

Could someone tell me why one type of drive wouldn't work with a specific version of Windows? Shouldn't they be able to write drivers for that?

Obviously because Microsoft paid them a certain amount of money to make it an extra reason to force people to upgrade.

Re:Not on XP?

Solid-state drives are flash drives with a PATA/SATA connector, and will work like a regular hard disk, as far as the motherboard and the OS are concerned. Therefore working whatever OS you're using.

Hybrid drives, OTOH, are relying on two different technologies, and it seems the choice of using disk or flash is up to the OS. It means that if your OS isn't Hybrid-drive aware, you probably will end up with using the disk and losing its flash ability. Vista OTOH will be able to put some files on the flash part.

Re:Not on XP?

You seem to have three hands.

Re:Not on XP?

You seem to have three hands.

Sooo, no need to type one handed like the rest of us.

SuperFetch

According to Microsoft, "SuperFetch understands which applications you use most, and preloads these applications into memory, so your system is more responsive".

Seems nice in theory, but the first thing I do to any XP machine that someone tells me is running very slow is to kill those quick start apps in the bottom right corner. Their use of processor and/or memory definitely slows the machine down overall. I'd much rather wait an extra second for an app to load so the system runs faster overall.

So they better have improved their techniques with this SuperFetch. If it causes many more context switches or reduces memory available to apps people are actually running then it'll be a hinderance. At the very least it should be automatically turned off for systems with less than an ideal amount of memory.

Re:

If it's done right, then it'll be handy. IIRC, linux uses free pages of memory for disk cache, and if an application needs more pages, it just invalidates the disk cache pages, and allocates them to the app.

If Windows caches applications into free memo

Re:SuperFetch

I saw this "SuperFetch" idea and it is a total rip off of NeXT's "prebinding" system. Often, when you install something on Mac OS X (since version 10.0), there is a little status message in the installer that says "Optimizing System Performance...". This command calls a program that sits in "/usr/bin" that loads memory addresses of each program in a cache for faster launch times. After prebinding, applications load faster at startup.

There is also a daemon on Mac OS X that dynamically prebinds applications that have not been prebound. One condition of prebinding is that all the Libraries must be dynamically linked and prebound themselves. If one dependant library is not prebound, then the whole thing gets marked as something "not to prebind."

To see the actual programs on Mac OS X, do a
ls /usr/bin | grep prebinding

Maximum lifetime of flash...

Doesn't flash memory have a maximum lifetime (R/W cycles)? If so, are these new drives designed to "degrade" gracefully so that as the flash "rots", more and more data is stored to the drive instead of the memory? If so, this would mean that the drives would "slow down" over time right?

Re:Maximum lifetime of flash...

Hard disks also have maximum lifetimes. Both HDDs and flash drives reallocate damaged blocks to compensate for the problem. The question is how the two compare in practical use.

Re:

TFA states that current flash technology has a lifespan of about 10 years. Unlike hard drives, when flash fails you can still read from it, just not write to it. This means that, when your drive wears out, you just dump the contents to the new one, which

Reminds me of when...

Reminds me of when a company in the 70's built a solid-state swapping "drum" memory system for IBM S/370 mainframes. Of course, that one wouldn't fit in a 2.5" form factor.

Solid State = Sexy

The greatest immediate benefit from the transition to solid state storage will, of course, be reduced power consumption.

Coupled will fuel cell technology, mobile computing is finally going to live up to its potential.

And I love this William Gibson quote from 1991:

It wasn't until I could finally afford a computer of my own that I found out there's a drive mechanism inside- this little thing that spins around. I'd been expecting an exotic crystalline thing, a cyberspace deck or something, and what I got was a little piece of a Victorian engine that made noises like a scratchy old record player. That noise took away some of the mystique for me; it made computers less sexy. My ignorance had allowed me to romanticize them.

"Pricing was not announced"

...because they don't want you to get a bad case of sticker shock. If texas memory systems (http://www.texmemsys.com/) is any guide, these things won't be comparable to platter drives in cost per GB per performance. Maybe they've figured out a way to manufacture the things not too expensively per GB but the performance will be wretched. And even though most apps will not care unless you have a stopwatch people will look at the raw numbers and shy away. Just see all the trouble AMD had with the Pentium 4 vs Athlon XP CPU GHz wars.

Oh good!

Vista users would see a 4x speedup in many key operations.

So now this might get Vista running half as fast as every other operating system, right?

Cheap Spinning Media has come a long way too

I did an eval of SSD back, oh five years ago for my employer. These were SSD's attached via SCSI to Sun boxes running Solaris and Sybase. Based on the results I saw then, I have two problems with this:

>Vista users would see a 4x speedup in many key operations.
Back in the day, we were seeing 10-20X improvements over spinning media in Random Access. 4x is almost not worth it, depending on price - give spinning media another year or two and they'll match that gain.

>Pricing was not announced.
Of course not, because it's going to be outrageously expensive!

Re:inflection point is coming

What the...WE HAVE SHARP METAL DISCS SPINNING @ 7200prm ON OUR LAPS?!

HEAVENS TO BETSEY!

Re:

I wouldnt bet on that. The gap between what can be stored on a given area of magnetic/optical disks compared to some type of solid state memory is actually getting larger , not smaller.

Re:Bzzzt!!!! It uses flash ram.

And if anyone had actually read the article, they would see that according to Samsung, the Flash technology in use in the drives has a lifetime of TEN years (your IDE / SATA HD likely wont last that long btw). They also note how much the R/W cycle issue has improved in the last few years.

Oh wait, this is /., we don't read the articles we just write silly comments first!

Re:Bzzzt!!!! It uses flash ram.

I've always wondered about this. Most modern flash seems to get 100k writes (many more reads). Fast flash is on the order of 13MB/s write.

With load balancing, you wouldn't notice a failure until all the locations were rewritten just shy of 100,000 times. So the drive will "fail" in once you've written 40GB of data 99,999 times, or almost 4PB of write ops. At 13MB/s, that's just under 10 years of 100% duty cycle writes. If you presume you'll read that data once at 20MB/s, and you allow only an 82% duty cycle overall (to make the math easy), then your drive should last 20 years.

I don't know about you, but I don't have any 20 year old computers or drives. The computer I had 20 years ago (PS/2 model 30, iirc) used 720k floppies, and a 20MB hard drive was a $400 option. Wait, check that. I do have a copy of Windows 1.04 on floppy disk here. It fits on three 720k floppies.

Re:Bzzzt!!!! It uses flash ram.

The usual way they construct it is like this:
1. Fill your drive 95%
2. Trash the remaining 5%. Your disk will now die in 1/20th of the time, that is a matter of months

IMO even that theoretical problem could be solved by active swapping, that is using some of your write cycles to move information internally. If you spent 100 of your 100k cycles doing that noone would notice. So when you're trying to trash those 5%, those 5% would swap places with the other 95%, even though there's no free space. For all I know maybe they do already, but if it was a problem that is the solution (this was sooo obvious. I bet it's patented).

Re:

Let's look at the application, notebooks. There are quite a few pros for solid-state drives here: 1) HDDs are loud, 2) HDDs are hot (especially as you increase RPM), 3) HDDs are sensitive to motion, 4) HDDs require more power, 5) HDDs are marginally heavi

Re:

Bzzt, wrong. I don't know how many applications you're loading up, but 32 Gigs is plenty for my entire windows C: drive. I'll keep all my applications and operating system on fast, quiet SSD, and I'll happily store my 400 gigs of music and video on magne

Re:Dedicated OS Harddrive?

Don't forget spinup time. Spinup time is pretty important, especially for notebooks. Notebooks are bound to suspend the disk a lot more than desktop drives, so to safe power.