Apple looks towards flash-enabled notebooks

[AppleInsider]

Intel Corp's plan to incorporate NAND flash memory into its next-generation notebook platform will enable PC manufacturers such as Apple Computer to develop systems that are twice as fast in some operations yet sustain longer battery life.

Already a fixture in Apple's most popular iPod digital music players, NAND flash is a type of memory that contains no moving parts and retains data even without a power supply.

During a developer forum this past March, Intel revealed plans to add the technology as a feature of its Santa Rosa notebook platform due out in the first half of next year, saying it would offer the main benefit of almost instantaneous PC boot times.

At the forum, Intel mobility chief Sean Maloney conducted a demonstration in which he booted two PCs, one with 256MB of flash memory, and the other without. The PC with flash booted in about half the time.

Maloney said the technology can scale way beyond a 256MB flash buffer, potentially running a PC's entire operating system from flash instead of from the hard drive. "It just comes down to what's the cost curve on NAND," he said.

While Apple is not listed amongst the PC manufacturers who have committed to the Santa Rosa platform, people familiar with the Mac maker say it has been working closely with Intel engineers to implement NAND flash into a future-generation of its MacBook notebook lines.

Those same people say that Apple is amongst the best positioned in the PC industry to adopt the technology broadly due, in part, to its existing supplier arrangements with the world's five-largest NAND suppliers.

Apple's top selling notebook line: the 13-inch MacBook

At Intel's fall developer forum this week, Intel chief executive Paul Otellini spoke at length about Santa Rosa and offered some early benchmarks for NAND-enabled notebook PCs. He said users could expect faster boot times, 2X faster application load times, and a 2X reduction in the time need wake a system from hibernation.

"The platform that launches next year will have NAND on the motherboard for the first time that we know of out there," said Otellini. "[It] significantly improves notebook performance and battery life."

Essentially, Otellini said, the NAND flash will act as a buffer cache so that frequently accessed data can be stored on the NAND flash rather than the hard drive.

"This means it is more available to the main memory and the microprocessor," he said. "It doesnt have to go out to the drive. Since it doesnt have to go out to the drive as often, it consumes less battery life under notebook mode."

Before embarking on its next-generation notebooks in 2007, Apple later this year will refresh both its MacBook and MacBook Pro lines with Intel's new Core 2 Duo mobile processors.[ View this article at AppleInsider.com ] (http://www.appleinsider.com/article.php?id=2092)

[BMWintoxication]

so it acts like a buffer..

if someone finds a way to do this for the current notebooks, it'd sell.

[cwoloszynski]

I think that this is a great item for Microsoft, since they have to reboot so often. For me, I reboot when a new security fix requires it. Otherwise, my iBook (my MBP arrives tomorrow) just shuttles from home to work with the lid closed.

Perhaps I am missing a key value to this. Anyone else think that improving the boot time of a MB or MBP is not much of an issue for OSX users? Perhaps Bootcamp users want this!

[bdj21ya]

I think that this is a great item for Microsoft, since they have to reboot so often. For me, I reboot when a new security fix requires it. Otherwise, my iBook (my MBP arrives tomorrow) just shuttles from home to work with the lid closed.

Perhaps I am missing a key value to this. Anyone else think that improving the boot time of a MB or MBP is not much of an issue for OSX users? Perhaps Bootcamp users want this!

It will be extremely awesome for use with Boot Camp, making OS switching much less onerous. However, if you read the article carefully, it points out that programs will be able to boot faster as well if recently used programs are left stored in the NAND memory. Think about the boot times for some Adobe products for example, and you'll see that this innovation could be a big deal to just about every computer user.

[IQ78]

I think that this is a great item for Microsoft, since they have to reboot so often. For me, I reboot when a new security fix requires it. Otherwise, my iBook (my MBP arrives tomorrow) just shuttles from home to work with the lid closed.

Perhaps I am missing a key value to this. Anyone else think that improving the boot time of a MB or MBP is not much of an issue for OSX users? Perhaps Bootcamp users want this!
...
Battery Life. Anything that keeps a hard drive spun down for as long as possible can really save battery life. Speed + Battery Life. Also, there are plenty of consumers that just can't get behind the sleep thing. They feel they must shutdown their computers when they are not using them. Even if they do put them to sleep, this should help with the wake-up speed as well.

IQ78

[ecking]

It's gonna be a while until this trickles down in the consumer space.

Probably only a high end macbook pro will have this within the next 2-3 years. And that's assuming intel has the formula right, right now. It could take intel another 1-2 years easy to perfect it.

[aiolos]

It's gonna be a while until this trickles down in the consumer space.

Probably only a high end macbook pro will have this within the next 2-3 years. And that's assuming intel has the formula right, right now. It could take intel another 1-2 years easy to perfect it.
but won't it be in the santa rosa?

[rain]

Woo hoo... slowly going back to hard programming.
I will always stand on my position that the Commodore 64 was the best personal computer ever.
I used one for 11 years and NEVER lost work to a crash, or 'Application suddenly quit'.

Eventually, we will have computers with no moving parts at all.

[fuyutsuki]

Eventually, we will have computers with no moving parts at all.

Well that part I can agree with.*:lol:

Flash is so where it's at, only not quite today.* I'm totally getting a 16gb compact flash card to put in an IDE adapter for my PowerBook, whose hard drive has seen better days anyway.* Currently there's just not enough capacity in flash, but 16 gigs should handle Tiger / Leopard and my apps with some room for swap.* FireWire for the rest.* I'm a finnicky bugger for silence and fancy the experiment.* I dub thee "FlashBook".

Um but yes, the extended hard drive buffer idea of Intel's is sound.* The platforms for future generations of MacBook±Pro should be pretty interesting.

[Elixir]

if NAND can boot programs faster than that is a huge plus, even now the new MBP takes some time to boot certain programs.

i like the idea of NAND being used in laptops a lot, it just seems better overall.

is there any downside to using JUST nand in the future?

[Ireland]

Eventually, we will have computers with no moving parts at all.
I'd say the fans will be moving for some time to come, seen as laptops are only getting hotter! :p

[Ireland]

If nand on the motherboard makes things faster, then why wouldn't they be interested in doing it with the desktops too?

[Kickaha]

Slapping in another DRAM slot on a desktop is easier than on a notebook. DRAM is cheaper than NAND flash. If you're just going to be using it for a cache, then on a desktop, RAM is the way to go.

[Bageljoey]

i like the idea of NAND being used in laptops a lot, it just seems better overall.

is there any downside to using JUST nand in the future?

Ummm, I can think of one... CO$T:no: :no:

[Lust]

I'd say the fans will be moving for some time to come, seen as laptops are only getting hotter! :p
Hmm, how about those ion cooler things? I can't remember exactly what they're called. I saw them on www.slashdot.org not too long ago. Seemed pretty cool.

[Ireland]

Ummm, I can think of one... CO$T:no: :no:
Yeah that, and with Nand there's a limited amount of times you can read/write to it.
/but they're working on that.

[OfficerDigby]

AFAIK it's already part of Vista so...

So we get Nand on the board and Nand on the hard drive. 2007 is year of the Nand!

[SteveGTA]

with Nand there's a limited amount of times you can read/write to it.
/but they're working on that.

interesting, is it a big enough number to the point where it almost doesn't matter?

[Lust]

interesting, is it a big enough number to the point where it almost doesn't matter?
The math has been done for this *many* times. According to *their* math, it should last for quite some time. Much longer than you will probably retain the notebook. So I wouldn't worry too much.

[AaronHarmon]

It's gonna be a while until this trickles down in the consumer space.

Probably only a high end macbook pro will have this within the next 2-3 years. And that's assuming intel has the formula right, right now. It could take intel another 1-2 years easy to perfect it.

Don't count on it taking that long. I"ve been reading about this tech for around a year now. Not only is Intel working on it. So is every major HD manufacturer. Sansung, WD, ect, are all working on flash inbeded HDs. While intel is looking at putting the flash on the MB. I've read about prototype drives with a gig of flash on them already. The only drawback is it will require at the very minimum bios updates to existing MB's (even in the case of flash inbeded HDs to my knowledge), or possibly in some cases a new MB altogether. It is likely that we will see both, Flash inbeded MB's running Flash inbeded drives. The only question is which is faster, and how does the system decide where to store critical, heavy access files. I have read that with the larger flash drives, they will be able to almost double the battery life on some if not all portables (which will also decrease wear and improve their longevity), but also decrease power consumption on desktops (something that the corperate customers would love). This tech will be available alot sooner than 2-3 years. You'll see it in 2007.

[AaronHarmon]

If nand on the motherboard makes things faster, then why wouldn't they be interested in doing it with the desktops too?


They will have it for desktops, I don't know if they will inbed it on the MB in a desktop though. But the Flash inbedded HDs will be available for desktops. We are not just talking about faster boot times, we are talking about less spin up (less noise, les drive failure), faster seek times and better/faster throughput. This stuff will speed up computing across the board. One of the big selling points for laptops though is the much improved battery life, not to mention the fact that they can charge portable customers a premium and they know we'll pay it. Can you imagine laptop with 8-10 hours on a single charge, they are not all that far off.

[AaronHarmon]

Slapping in another DRAM slot on a desktop is easier than on a notebook. DRAM is cheaper than NAND flash. If you're just going to be using it for a cache, then on a desktop, RAM is the way to go.


Yes, but Ram drives are not convenient and just adding ram will not give you the same effect. You can install just about anything to a ram drive, even operating systems, but as soon as you reboot, you have to install everything all over again. I dont know if people still do it, but it used to be fairly common among hardcore PC users to install GLQuake to a ram drive for better frame rates, benchmarks ect... It seems to me like a few years ago there was talk of creating ram drives that would remain powered up even when the PC (we are talking PCs here) were turned off. I don't know what ever became of that idea. My guess is one of two things; either the drives were to volatile and easily corrupted, or the cost was too high for any real market viability. There are raid controllers that allow for copious amounts of ram installed directly on the controller board itself though.

[DHagan4755]

umm, how about Apple just worries about catching up to the competition and getting a notebook out the door with a Core2 Duo.

[demenas]

Slapping in another DRAM slot on a desktop is easier than on a notebook. DRAM is cheaper than NAND flash. If you're just going to be using it for a cache, then on a desktop, RAM is the way to go.

RAM would require battery backup.

Steve

[JeffDM]

Well that part I can agree with.*:lol:

Flash is so where it's at, only not quite today.* I'm totally getting a 16gb compact flash card to put in an IDE adapter for my PowerBook, whose hard drive has seen better days anyway.* Currently there's just not enough capacity in flash, but 16 gigs should handle Tiger / Leopard and my apps with some room for swap.* FireWire for the rest.* I'm a finnicky bugger for silence and fancy the experiment.* I dub thee "FlashBook".

Um but yes, the extended hard drive buffer idea of Intel's is sound.* The platforms for future generations of MacBook±Pro should be pretty interesting.

You don't want to swap with flash, that's what's going to kill it with limited write cycles unless the OS regularly moves the swap file location every time it writes, but if you have limited space, that wears it out much quicker. You'll want to max out the RAM to minimize swapping.

My Application directory is 9GB. I suppose I can cull it of lesser-used programs, but it's not going to carry a lot of media or working files for portable use.

I really don't have a noisy notebook drive (I can't even hear it clicking), so the reduced noise is not much benefit to me. Most of the faint noise that my notebook makes is the fan for cooling the hot CPU.

[Mr. Me]

The use of flash memory as cache doesn't make a lot of sense. Its use as a replacement for rotating disk platters sounds good until you do a little math. It is the technology of the future and probably always will be. Compared to conventional hard drives, flash memory is very expensive and very slow. Then there is the issue of read/write cycles. Virtual memory OSes access the hard drive a lot. When will the flash memory be able to handle the read/write cycles required of MacOS X?

[JeffDM]

DRAM is cheaper than NAND flash.

What type do USB pocket flash drives use? Those things can fit 4GB in a package the size of a small pack of gum and only cost $100. I have a hard time finding 1GB RAM sticks for much less than $100.

[Kickaha]

Yes, but Ram drives are not convenient and just adding ram will not give you the same effect. You can install just about anything to a ram drive, even operating systems, but as soon as you reboot, you have to install everything all over again.

RAM would require battery backup.

Steve

Yes, which is why I said specifically...

If you're just going to be using it for a cache,

... to qualify it if it being used for cache only. Obviously it has other applicability for persistent storage.

[melgross]

Yeah that, and with Nand there's a limited amount of times you can read/write to it.
/but they're working on that.

Yeah, I was going to post that as well. I haven't heard much talk about that aspect.

One can write about 100 thousand times to this, but can read more than a million times.

So, while I can see this used for information that won't be changed much, there are newer technologirs for FLASH that will be out that will solve the write problem.

Without that being solved, there is no way that it could be used for a HD replacement.

Check this out.

http://samsung.com/PressCenter/PressRelease/PressRelease.asp?seq=20060911_0000286481

With Apple being such a big customer of Samsung FLASH, they might be working with them on this already. Let's hope!

[melgross]

interesting, is it a big enough number to the point where it almost doesn't matter?

Nope. See my post above this one.

[melgross]

The math has been done for this *many* times. According to *their* math, it should last for quite some time. Much longer than you will probably retain the notebook. So I wouldn't worry too much.

100 thousand writes can be completed in a few months.

[Chucker]

One can write about 100 thousand times to this, but can read more than a million times.

The read cycles of NAND Flash aren't limited; only the write cycles are, and they're a lot higher these days than the often-cited 100,000 times.

That said:

Without that being solved, there is no way that it could be used for a HD replacement.

Very true.

[melgross]

The read cycles of NAND Flash aren't limited; only the write cycles are, and they're a lot higher these days than the often-cited 100,000 times.

That said:



Very true.

Read cycles are limited. Both might be a bit higher than I mentioned, but not much. Those figures are from earlier this year.

[jimbo123]

I remembr you could create a ram disk on the Amiga...many years a go.

[Chucker]

Read cycles are limited. Both might be a bit higher than I mentioned, but not much. Those figures are from earlier this year.

I'm genuinely curious; please find me a source that states that NAND flash read cycles are limited.

I remembr you could create a ram disk on the Amiga...many years a go.

You can still create a RAM disk in OS X, but since OS X aggressively caches frequently-accessed data in RAM anyway, it's not as useful any more.

[MacGregor]

I think that this is a great item for Microsoft, since they have to reboot so often. For me, I reboot when a new security fix requires it. Otherwise, my iBook (my MBP arrives tomorrow) just shuttles from home to work with the lid closed.

Perhaps I am missing a key value to this. Anyone else think that improving the boot time of a MB or MBP is not much of an issue for OSX users? Perhaps Bootcamp users want this!
Well regarding MS, I thought I saw a demo of a laptop running Vista where Vista allowed pda-like operations to take place in a small window on the lid of the laptop. This ran some palm-like apps that could be accessible without even having to open or awaken the computer itself.

I have thought for a while that Apple should be making their laptops more instant on, but they don't need to do it for everything. I don't care that Photoshop takes 5 minutes to load b/c after I start it up, I keep it going for quite a while and over the course of the day the boot time is insignificant. However the fact that iCal takes almost as long to load as Photoshop is exxasperating!!! I want to know within 5 seconds whether I'm late or not.

I think Apple's productivity iApps are perfect to be sitting in a NAND memory buffer so that iCal, AddressBook, Mail, Stickies and Widgets are absolutely available with no waiting. That would make the OS as a whole seem instantaneous. Who cares if Illustrator takes a while, it is an excuse to get another coffee.

[melgross]

I'm genuinely curious; please find me a source that states that NAND flash read cycles are limited.

This is one. I'll quote from it, as well as give the link, as it's a long scroll 'till you get to this. It's interesting though, so you might want to read the whole thing. Where he says 1m read/writes, he really means 1m reads.

"But besides cost, there are other obstacles for Flash memory replacing hard drives. Flash memory is typically guaranteed for around 1m read/write cycles, which sounds a lot, but in the context of the working life of a PC (which is continually writing, erasing or rewriting data) is not. So designers have to use techniques known as wear-levelling to shift data around the Flash memory block to reduce the risk of data error. "We can design around the problem," says Walsh."

http://technology.guardian.co.uk/weekly/story/0,,1856508,00.html

Wiki has this, which is, surprisingly correct, but as usual, not totally complete. Look down to where it says EPROM etc. lifetime.

The write life is complete. The read life is correct as far as it goes, in that they do say that it is limited, but they don't say that that limitation is considered to be about 1 million cycles. It isn't cycle limited, it's true, but the charge leakage gets worse with time, just as it does in most capacitors, and that charge drop ends up in a life limitation which usually comes out to 1 million cycles.

http://en.wikipedia.org/wiki/Read-only_memory

[Chucker]

I don't think a random newspaper that doesn't even focus on technology is a particularly good source. As for Wikipedia, it is self-contradictory, as it also states:
When compared to a hard disk drive, a further limitation is the fact that flash memory has a finite number of erase-write cycles (most commercially available flash products are guaranteed to withstand 1 million programming cycles) so that care has to be taken when moving hard-drive based applications, such as operating systems, to flash-memory based devices such as CompactFlash.

This implies that the limitation only applies to write cycles.

[melgross]

I don't think a random newspaper that doesn't even focus on technology is a particularly good source. As for Wikipedia, it is self-contradictory, as it also states:


This implies that the limitation only applies to write cycles.

There are quite a few other sources.

You can go and look if you want to see this that badly. They all say about the same thing.

In fact, most Flash cards have only about 10,000 write cycles.

Most information is concerned with writes, because it is so poor.

Leveling does help. But mostly for static data, such as pictures.

[palegolas]

The best would be to replace the harddrive completely with.. say, 8x8GB flash memories connected to act like one volume. And then you could have a free slot around the battery or something for insertion of secondary flash drive.

[orge]

I don't think a random newspaper that doesn't even focus on technology is a particularly good source. As for Wikipedia, it is self-contradictory, as it also states:


This implies that the limitation only applies to write cycles.

This paper might explain the contradiction/issue with read and write cycles:

http://snipurl.com/xkma

It's not a fully published paper (and it's a couple of years old), but the authors work in engineering for space applications. Check out the second paragraph of section 2:

"FRAM data is read using a switched charge that determines the state of the ferroelectric memory. In the process of reading the data, the data is changed; therefore, additional circuitry must be used to re-write the original data back to the memory cell."

This suggests that reading from FLASH involves a write operation?

J

[Chucker]

I appreciate the effort, but FRAM is actually quite different from NAND Flash memory, so that quote doesn't really apply at all. Furthermore, while I merely skimmed through the paper, the terms "NAND" and "NOR" don't appear at all, and when they say "FLASH" (which they write in upper case for no good reason), it's unclear what type of it they refer to. Much of the information appears to have been written several years ago, whereas this market is one that develops very quickly.

[sunilraman]

Bootup will be the main target. Because that can handle enough writes when installing OS'es over the say max 5 year lifespan of the machine. It would also handle the reads on bootup.

Hibernate will be the secondary target but Apple Mac Sleep does that pretty well.

I suggest we'll see this tech for Bootups introduced alongside Crapsta (Vista) and gain traction towards the end of 2007.

Apple will use this towards the end of 2007 for superfast Bootups. Hell, 2GB Flash for Booting up the OS is more than enough. EFI on, Boot up off NAND, dump to OS kernel and processes, etc, into RAM, you're ready to rock. Yeah, maybe frequently used applications but that eats into the lifetime of the NAND so I'm keeping to OS installs as the write-functions we need to be concerned about at this stage.

Other companies will have some inane obvious marketing name like "Insta-Boot" technology but Apple will quietly update their laptops with this feature, with a sexy name like (work with me here peoples):

Wake
Demand
One
On Demand
Wakeup
Start (heh. Windows "Start" menu)
Abracadabra
OpenSesame
Steve
Startup
ClickStart
SnapStart <--- okay this is the best one so far

"Featuring the new MacBook 13" glossy widescreen with 1GB of RAM and SnapStart"

[Chucker]

SnapStart seems a reasonably good name.

[minderbinder]

If nand on the motherboard makes things faster, then why wouldn't they be interested in doing it with the desktops too?

Because many people never reboot desktops, or even put them in hibernation like laptops. My desktop is always available almost instantly, if I wait it's usually for the screen.

And for apps, you can just add more ram and have the OS completely load them into ram and have the same thing, but better.

umm, how about Apple just worries about catching up to the competition and getting a notebook out the door with a Core2 Duo.

They'd probably be more worried if their notebooks weren't flying off the shelves faster than they can make them already. It will happen one of these days, if you really need merom just wait for it.

[melgross]

The best would be to replace the harddrive completely with.. say, 8x8GB flash memories connected to act like one volume. And then you could have a free slot around the battery or something for insertion of secondary flash drive.

In the future, when it is reliable enough.

[clockwork-master]

Honestly speaking on this issue, has anyone posting here heard about M-Systems IDE on Flash or even RAM modules which retains the memory even when the power is off. The Technology is not new I think. What I suspect here is that there would be a new form factor for this technology which would find themselves into laptops and desktops. The advantage would be the faster access time or some addressing innovation. Nuff said.

[cool11]

I am not sure I have understood the whole thing.

Is this a matter of hard disk technology only, or general update of motherboards, chipsets etc?

If it is all about hard disk, everyone can buy one!

[Transcendental Octothorpe]

Just FYI,

I make my living in the NAND testing industry, and AFAIK, the write-cycle problem is getting WORSE, not better.

The methods they are using to acheive increased NAND density (bigger for less $$$) are almost always connected with smaller program cycling life. (smaller litho, multi-bit cells)

I've asked around, and even posed this question at one of these "developer conferences" and I have never heard a satisfying answer.

Frankly, I am begining to suspect it is a case of planned obsolescence. :\


Oh, and NAND doesn't need to be re-written after every read. (DRAM, however, does work that way, as does FRAM, apparently). The cycles we are talking about are changing the data. Reading the same data is only subject to the devices data-retention span (which I beleive is spec'd around 10 years).

So, for a digital camera, on a 1GB card, you can be taking a few 1000 pics a day for many years before having a problem. Having an OS swap virtual memory during normal use would burn through 100,000 cycles in no time.

So, as mentioned above, this seems feasable for boot, and maybe even app/dll storage, but I still don't understand how the SSHDD (solid-state HDD, replaceing a mechanical HDD with NAND) will fly. It seems that OS's would have to be written with extensive planning to minimize the HDD writes.

[melgross]

Just FYI,

I make my living in the NAND testing industry, and AFAIK, the write-cycle problem is getting WORSE, not better.

The methods they are using to acheive increased NAND density (bigger for less $$$) are almost always connected with smaller program cycling life. (smaller litho, multi-bit cells)

I've asked around, and even posed this question at one of these "developer conferences" and I have never heard a satisfying answer.

Frankly, I am begining to suspect it is a case of planned obsolescence. :\


Oh, and NAND doesn't need to be re-written after every read. (DRAM, however, does work that way, as does FRAM, apparently). The cycles we are talking about are changing the data. Reading the same data is only subject to the devices data-retention span (which I beleive is spec'd around 10 years).

So, for a digital camera, on a 1GB card, you can be taking a few 1000 pics a day for many years before having a problem. Having an OS swap virtual memory during normal use would burn through 100,000 cycles in no time.

So, as mentioned above, this seems feasable for boot, and maybe even app/dll storage, but I still don't understand how the SSHDD (solid-state HDD, replaceing a mechanical HDD with NAND) will fly. It seems that OS's would have to be written with extensive planning to minimize the HDD writes.

It won't work. Not yet.

[a_greer]

If nand on the motherboard makes things faster, then why wouldn't they be interested in doing it with the desktops too?
Desktops use faster drives so the gain is less notable and this is as much if not more about saving battery life than speed.

[melgross]

Desktops use faster drives so the gain is less notable and this is as much if not more about saving battery life than speed.

That's right. NAND is slower than laptop drives, but MUCH slower than desktop drives.

Battery power is a major consideration.

But the Samsung technology, when available, will make desktop use far more feasable.

[Programmer]

Other companies will have some inane obvious marketing name like "Insta-Boot" technology but Apple will quietly update their laptops with this feature, with a sexy name like (work with me here peoples):

Wake
Demand
One
On Demand
Wakeup
Start (heh. Windows "Start" menu)
Abracadabra
OpenSesame
Steve
Startup
ClickStart
SnapStart <--- okay this is the best one so far

"Featuring the new MacBook 13" glossy widescreen with 1GB of RAM and SnapStart"

Perhaps Apple's partnership with Nike will take it to the next level: "Just Use It"

[Lust]

Desktops use faster drives so the gain is less notable and this is as much if not more about saving battery life than speed.
This is mainly for booting up. Faster drive or not, the hard drive still has to start rotation and then actually seek the data. Flash drives do not move, so it just seeks, cutting down the time.

[DuckofPrey]

How fast are NAND flash memory in terms of access speed? Compared to RAMs I mean... I'm thinking... one of the big advantages of going to a 64bit system is the availbility of truly stupendously large memory addresses. At the moment, getting more then a few gigs of ram is not viable in terms of cost. If these flash memories can provide the same level of access speed, couldnt NANDs effectively replace rams? If so, the NAND enabled systems would go hand in hand with a 64bit system... me thinks...

J

[JeffDM]

How fast are NAND flash memory in terms of access speed? Compared to RAMs I mean... I'm thinking... one of the big advantages of going to a 64bit system is the availbility of truly stupendously large memory addresses. At the moment, getting more then a few gigs of ram is not viable in terms of cost. If these flash memories can provide the same level of access speed, couldnt NANDs effectively replace rams? If so, the NAND enabled systems would go hand in hand with a 64bit system... me thinks...

NAND cannot replace RAM, not with anything like what is available now. It's considerably slower, and the problems with rewrites with flash drives gets WORSE with RAM because memory gets rewritten very often. I am not certain, but NAND might require you to rewrite in large chunks of memory, RAM lets you change much smaller chunks, like a few bytes rather than a few kilobytes.

[freakboy]

i saw a review of a flash based harddrive, and it was incredibly fast in real world use, even though the actual throughput wasn't faster than a fast HD. the reason was that the seek times are like 0.02 ms instead of 10 ms. this makes a ton of operations that happen on the machine all the act much more instantaneous. any application that makes a lot of reads and writes to files will be seriously boosted. I'm not sure what apps those would be...

this is it.. btw, this will just work in a lappy.

http://www.tomshardware.com/2006/09/20/conventional_hard_drive_obsoletism/

[melgross]

This is mainly for booting up. Faster drive or not, the hard drive still has to start rotation and then actually seek the data. Flash drives do not move, so it just seeks, cutting down the time.

Drives in desktops often continue to rotate. Laptops have power considerations that desktops don't.

[melgross]

i saw a review of a flash based harddrive, and it was incredibly fast in real world use, even though the actual throughput wasn't faster than a fast HD. the reason was that the seek times are like 0.02 ms instead of 10 ms. this makes a ton of operations that happen on the machine all the act much more instantaneous. any application that makes a lot of reads and writes to files will be seriously boosted. I'm not sure what apps those would be...

It depends on what you are doing. If use use a database where you have many small files, then you will benefit fron reduced seek times, if available. But when using larger files, the seek times become less important, as long as you have enough room on your drive so that most files are not severely fragmented.

Most drives these days have seek times in the low milliseconds, so, even there, seek times aren't usually much of a problem.

The computer spends most of its time waiting for input from us. Most of the time, the file is found as soon as your finger is lifted from the key that selected it.

Reading the data from the drive can take an order of magnitude longer than finding it. And, that's if it's a small file. a long file can take several orders of magnitude longer.

[sunilraman]

A thought came while I was jogging today. This technology could go into desktops within the next few years as NAND capacity and price drops.

For example, you could have a generic motherboard, and a PCIExpress expansion card that has an array of slots for NAND chips, much like slots for RAM.

The key is to view the NAND as a consumable, in the sense of somewhere between a DVDR and a Hard Disk.

The NAND Drive Array Manager would be essential for SnapStart(tm) of, well, whatever you choose. Booting Windows, firstly. Next, for example, your favourite applications. Then, for PC Gamers, their favourite games. Starting a game and level load times will be TehUltraSnappy(tm).

The first key is to have the NAND Drive Array Manager as an expansion card to slot into the machine, allowing you to add NAND modules as needed, and remove it as needed when the write cycles are consumed. Yes, as you may have thought, the second key is for the NAND Drive Array Manager software to report on read and write cycles left (in a sexy Aqua-esque usage bar thingy) so you as a power user or desktop support/ deployment IT dude can tailor your NAND Drive Array usage accordingly.

Budget enthusiasts and mid-size enterprise would use say 2GB to 4GB to have Windows and most used apps on NAND with mostly-read operations only once the initial install of Windows and apps are done. Adding apps to the NAND shouldn't be too painful. The key is that Windows and the apps use the hard disk and RAM for memory access and swapping, not the NAND. In this case the NAND modules need not be changed in less than 2 years if at all.

Harder enthusiasts and maybe mid-to-high-size enterprises would use bigger blocks on the NAND Drive Array, say an 8 x 2GB array - in this case, the Gamers would have their favourite games loaded up on the NAND, and the enterprise could have major file servers serving stuff off the NAND rather than hard disk. In this case, managing the usage (consumption) of NAND modules would be monitored and considered carefully, a balance of cost of modules, speed of loading stuff, and how frequently you can add and change stuff on the NAND modules.

Intel has pioneered this with Santa Rosa, but I think there are many an American and/or Taiwan and/or Korean company that can get a foot in the door - there are many applications for the consumer through to the high-level enterprise markets. The keys are:

1. Flexible modular addition of NAND modules
2. High capacity per cost of NAND modules
3a. Highly flexible and usable software management of NAND Drive Array
3b. Highly accurate reporting on write and read cycles left for each NAND module
4. Improvements in write and read cycle capacity of NAND modules
5. Load-into-RAM speeds far ahead of the desk/laptop hard disk speed growth curve
6. Adoption of hard disk RAID configurations applied to NAND modules
7. High-speed channel of NAND Drive Array to Motherboard BUS.
Remember this channel can be asynchronous as READS are more important.

Just to reiterate, the insight I want to add is to say, let's not futz around arguing about how fast and how long NAND modules last. The key is a system of hardware and software that allows modular expandability and close monitoring and management of how the NAND is performing and how the NAND is consumed . This system should also be scalable as NAND technology marches on.

.........................
.........................

[Lust]

Drives in desktops often continue to rotate. Laptops have power considerations that desktops don't.
Continue to rotate when the power is *off*? I said booting up not waking up.

[slughead]

If nand on the motherboard makes things faster, then why wouldn't they be interested in doing it with the desktops too?


Desktop hard drives are fast enough where the performance boost would not be as noticable (if at all).

[melgross]

Continue to rotate when the power is *off*? I said booting up not waking up.

What I meant was that during computer USE, the laptop drives stop spinning, unless they are being accessed, while desktop drives don't.

Obviously, when the machine is off, they aren't spinning. That was hardly worth your commenting on. :lol:

You should have known what I meant.

[Lust]

What I meant was that during computer USE, the laptop drives stop spinning, unless they are being accessed, while desktop drives don't.

Obviously, when the machine is off, they aren't spinning. That was hardly worth your commenting on. :lol:

You should have known what I meant.
Regardless of whether it was worth it or not, you're totally missing the point. This story is about Robson, which is used mainly, if not exclusively, for booting. Nothing more, nothing less. Why is the speed of the drive once it starts spinning relevant when this is used only for booting up and not for storage of normal files and/or programs?

[melgross]

Regardless of whether it was worth it or not, you're totally missing the point. This story is about Robson, which is used mainly, if not exclusively, for booting. Nothing more, nothing less. Why is the speed of the drive once it starts spinning relevant when this is used only for booting up and not for storage of normal files and/or programs?

Because, if you didn't notice, the discussion, as always, moves sideways from the original article to cover the larger issues involved. We do that. You have to get used to it.

[Chucker]

This story is about Robson, which is used mainly, if not exclusively, for booting.

That is incorrect, sir.

[sunilraman]

Because, if you didn't notice, the discussion, as always, moves sideways from the original article to cover the larger issues involved. We do that. You have to get used to it.

Well the discussion kinda always splits - one alongside the original article with various usual comments. Then one part branches of to a bit of silliness, name calling, and insulting. Then the third part sort of expands rapidly to cover the much much larger ramifications of any AppleInsider article, going far (too) deep into every single dribble of information. :D 8)

So anyway is there a head-to-head comparison with booting up off say a 512MB or 1GB or 2GB Flash vs a 7200rpm or 15,000rpm SATA2 desktop hard disk? I'd love to see soe benches on this.

[sunilraman]

Here's one, I'll have to read it later. Going for lunch. Some sashimi sounds good :D
http://www.tomshardware.com/2006/09/20/conventional_hard_drive_obsoletism/

[melgross]

Here's one, I'll have to read it later. Going for lunch. Some sashimi sounds good :D
http://www.tomshardware.com/2006/09/20/conventional_hard_drive_obsoletism/

I had seen that. The FLASH used in this won't be used in cheaper solutions. They also didn't bother to do any lifetime testing.

[sunilraman]

I had seen that. The FLASH used in this won't be used in cheaper solutions. They also didn't bother to do any lifetime testing.

Hmm... There are some interesting points out of that but yeah,

1. As you mentioned this Samsung FLASHdrive, how does it compare to, if I used my suggested FlashDrive Array PCIExpress/ SATA2 card that takes say current 2GB modules...

2. They use UtraATA/66 for the Samsung FLASHdrive, that's problematic in the some sense, but the read speed was frickin FAST. Again, current NAND modules on a SATA2 or PCIExpress channel? Speed may be faster?

http://images.tomshardware.com/2006/09/20/conventional_hard_drive_obsoletism/image004.gif

[sunilraman]

Writes are problematic because yeah, there is no lifetime data but it seems writes are slowish. However, for my SnapStart FlashDriveArray strategy you wouldn't be writing so much, because it chews up the write cycles and the life of the NAND module. So writes would be more to prep things for FAST READS.

http://images.tomshardware.com/2006/09/20/conventional_hard_drive_obsoletism/image005.gif

[sunilraman]

In this case the Samsung FlashDrive halves boot-up speeds. Intel's demo using let's assume newer, fast, NAND modules (maybe somewhat different or could be similar to Samsing FlashDrive) shows quick boot-up speeds.

Clearly booting up is a case where access times extremely favour the virtually instantaneous file addressing of NAND. Coupled with fast read speeds unhampered by something stupidly obsolete like UltraATA/66 and starting Windows, OSX, applications and games could be TehUltraSnappy(tm).

http://images.tomshardware.com/2006/09/20/conventional_hard_drive_obsoletism/image012.gif

[sunilraman]

Again we go back to, I suggest, a system of hardware and software that manages a scalable NAND drive array which has a high speed channel into the system BUS. Not something idiotic like UltraATA/66 but SATA2(300MBytes/sec) and PCIExpress (250MBytes/sec per lane).

This system with newer NAND modules could be more favourably used in Bootup and loads of OSes, Apps, and Games.

At say USD $40 per GB of NAND retail, you could have the interface array card at say USD $150, then say 8GB (4x2GB modules) for $320.

A bit tricky here because a WD 74gb 15,000rpm Raptor Hard Disk is USD $160. A pair of two at RAID 0 is $320. I know RAID 0 read/writes have been debatable in terms of improvements in OS, application, and game load times.

But here's an interesting tidbit from HardMac.Com:
http://hardmac.com/news/2006-09-28/#5971

They show a mind-blowing almost 300Mbytes/sec read and write speeds, and that seems to be AVERAGE (!) for a four 7200rpm SATA(2?) RAID 0.

It will be interesting to see NAND and HardDisks duke it out for commonly-read blocks of data which do not really change much over time. Throw in access time differences, power consumption, heat, noise, speeds, read/write cycle limitations, no moving parts vs moving parts, it seems like an interesting fight scene laid out over the next few years. As NAND and HardDisks dazzle us for our love. But the way I like it is all ways so we could see a mix of NAND and HardDisks even in desktops.

Looks like Intel's Santa Rosa is going straight for the juicy low-hanging fruit with speeding up boot times using an on-motherboard NAND chippy.

My FlashDriveArray is something a bit more ambitious but we could see that - anyway it was just an idea floating around in my head.

The TomsHardware article provides some more grist for the rumour mill and future "Intel-Apple" genius. ....?!

[melgross]

Writes are problematic because yeah, there is no lifetime data but it seems writes are slowish. However, for my SnapStart FlashDriveArray strategy you wouldn't be writing so much, because it chews up the write cycles and the life of the NAND module. So writes would be more to prep things for FAST READS.

http://images.tomshardware.com/2006/09/20/conventional_hard_drive_obsoletism/image005.gif

I'm trying to get it straight, because sometimes I mix them up. But NAND and NOR perform differently. One has faster reads, and one has faster writes.

[melgross]

Again we go back to, I suggest, a system of hardware and software that manages a scalable NAND drive array which has a high speed channel into the system BUS. Not something idiotic like UltraATA/66 but SATA2(300MBytes/sec) and PCIExpress (250MBytes/sec per lane).

This system with newer NAND modules could be more favourably used in Bootup and loads of OSes, Apps, and Games.

At say USD $40 per GB of NAND retail, you could have the interface array card at say USD $150, then say 8GB (4x2GB modules) for $320.

A bit tricky here because a WD 74gb 15,000rpm Raptor Hard Disk is USD $160. A pair of two at RAID 0 is $320. I know RAID 0 read/writes have been debatable in terms of improvements in OS, application, and game load times.

But here's an interesting tidbit from HardMac.Com:
http://hardmac.com/news/2006-09-28/#5971

They show a mind-blowing almost 300Mbytes/sec read and write speeds, and that seems to be AVERAGE (!) for a four 7200rpm SATA(2?) RAID 0.

It will be interesting to see NAND and HardDisks duke it out for commonly-read blocks of data which do not really change much over time. Throw in access time differences, power consumption, heat, noise, speeds, read/write cycle limitations, no moving parts vs moving parts, it seems like an interesting fight scene laid out over the next few years. As NAND and HardDisks dazzle us for our love. But the way I like it is all ways so we could see a mix of NAND and HardDisks even in desktops.

Looks like Intel's Santa Rosa is going straight for the juicy low-hanging fruit with speeding up boot times using an on-motherboard NAND chippy.

My FlashDriveArray is something a bit more ambitious but we could see that - anyway it was just an idea floating around in my head.

The TomsHardware article provides some more grist for the rumour mill and future "Intel-Apple" genius. ....?!

I'll post this again, as it's the way the future for FLASH will be heading. It will be much more effective than this current stuff.

http://samsung.com/PressCenter/PressRelease/PressRelease.asp?seq=20060911_0000286481