Using Compact Flash base IDE Drives Instead of Hard Disks

Message

jwcrim · #1 Post by **jwcrim** » Mon Dec 19, 2005 8:47 pm

I have some old Thinkpads I would like to replace the 3 to 6 GB drives with IDE Compact Flash based drives that would emulate the original drives (and not rquire a special BIOS).

If the new drive size is in this range will the BIOS handle it?

Is this possible?

The following table are CompactFlash IDE Drive CMOS Setup Parameters. These are some CHS parameters known to exist for various CompactFlash Cards.

Capacity Sectors/Card Heads Sectors/Track Cylinders
2 MB.........3,936....... 2.............. 16....................123
4 MB........ 7,872....... 2.............. 32............ ........123
6 MB........ 11,776...... 2.............. 32.....................184
8 MB........ 15,680...... 2.............. 32.....................245
10 MB....... 20,480...... 2.............. 32............ ........320
15 MB....... 29,312...... 2.............. 32.....................458
20 MB 39,168 2 32 612
24 MB 46,976 2 32 734
30 MB 58,752 6 32 306
40 MB 78,336 4 32 612
48 MB 93,952 4 32 734

cruzlite · #2 Post by **cruzlite** » Tue Dec 20, 2005 11:49 am

Very interesting idea...
Have been curious about it myself...
Probably only a matter of time till it is manufactured...
Certainly would make for durable, 'solid state' unit...

Sorry I cannot answer your question...
I'm betting somneone here can though...

jwcrim · #3 Post by **jwcrim** » Tue Dec 20, 2005 1:47 pm

cruzlite wrote:Very interesting idea...
Have been curious about it myself...
Probably only a matter of time till it is manufactured...
Certainly would make for durable, 'solid state' unit...

Sorry I cannot answer your question...
I'm betting somneone here can though...

It seems like it might create a second life for those great old Thinkpads. These days computers are used for audio applications like MP3 music, computer aided hearing aid user testing, etc.. Total silence is a key issue for many of these applications.

Flash memory is about to get a lot cheaper with Toshiba and Samsung both now investing at new multi-billion levels for dedicated Flash manufacturing facilities.

Given the cost/memory-size issue, these truly "solid state" silent notebooks may also tend to push apps toward Linux.

Besides all that, it gives us a new excuse to beat on these poor old workhorses.

emorphien · #4 Post by **emorphien** » Tue Dec 20, 2005 3:52 pm

I would be concerned with the longevity of the drives. Traditional hard drives handle the constant accessing and rewriting of a swap file without much trouble but a standard flash memory based drive wouldn't last very long under those conditions.

Unless it were similar memory technology to RAM (and thus would need a battery to maintain data when the system is off) I don't see this being a very good idea.

smugiri · #5 Post by **smugiri** » Tue Dec 20, 2005 7:04 pm

Have to disagree with you @emorphien: some unscientific analysis to support this ( no emperical data ) - speaking from the perspective of "this is the wave of the future":

Flash drives consume 5% of the power that regular hard drives do, curently come with SATA interfaces that are about 10x as fast as IDE/ATAPI with 80gb sustained data rates, a 2.4 million hour MTBF ( go to page 9 ) vs. 1.2 million for regular drives( I took SCSI commercial drives that offer the best performance ), it is no wonder that they are becoming the choice of the arme d force s,the most notorious early adopter of technologies known to man with their bottomless wallet ( section starting with "The ruggedness of Flash" which describes the advantages of flash over regular hard disks. ). Finally, this articledoes a direct comparison of solid state disks ( currently Flash memory or some variant thereof ) and magnetic drives. I think the section dealing with multiple read/write ops helps one make their own conclusions about swap file performance.

Another benefit of flash is longer battery life: assuming disk drains about 4w of ( table 1 on page 1 ) a 18w overall drain typical on todays laptops ( I got that little tidbit from somewhere but cannot for the life of me pull it up now .. ) moving to flash saves 3.95 W, a 20% power savings that translates to an extra hour of worktime on a T43p with a 5 hr battery life.

It is also important to remember that swap files are primarily a relic of the fact that RAM was too expensive to stuff into a PC as one wished in the past. RAM prices are falling and soon, it will possible to have as much RAM as one wishes in their machine. I would actually like to argue that this already true today ... most new home and business PCs come with 256+MB of RAM and for the average user, this is more that enough to make sure that they never go to swap meaning that the swap file is the appendix of the modern computer - all dressed up for a party with no where to go. Power users and users with very specialized memory hungry applications also know to simply order/buy their RAM requirements upfront while getting the machine ( and with power machines, these come with 512+MB ) or upgrade when they need too again solving their swap problem. Seems to me that swap is going the way of the 640kb thingy with older machines ( anyone who used anything older than a PI knows what I am talking about ) just another problem solved by throwing hardware at it.

I want to argue that once the correct price / performance point is reached, flash disk adoption will be de ri·gueur in laptops almost in the same way that it is pas·sé to whip out a 1.44mb floppy for data transfer today ....

jwcrim · #6 Post by **jwcrim** » Tue Dec 20, 2005 8:59 pm

emorphien wrote:I would be concerned with the longevity of the drives. Traditional hard drives handle the constant accessing and rewriting of a swap file without much trouble but a standard flash memory based drive wouldn't last very long under those conditions.

Unless it were similar memory technology to RAM (and thus would need a battery to maintain data when the system is off) I don't see this being a very good idea.

Those are reasonable concerns. Here is some data from BitMicro:

http://www.bitmicro.com/...urces_flash_ssd_db2.php [bitmicro.com]

Example #2: Write Frequency in MB/sec

E-Disk® PB Size = 16 KB

I/O Block Size = 64 KB

Write Frequency = 6,016,204,800 KB per day (68 MB/sec)

E-Disk® capacity = 155,648 MB
Number of Flash chips = 608
Size of Flash chips = 2048 Mbit or 256 MB or 262,144 KB

Number of writes to Flash chip = 64 KB / 16 KB = 4
Total E-Disk® physical blocks = (262,114 / 16) x 608 = 9,961,472
Total max writes to E-Disk® drive = 9,961,472 x 1,000,000 = 9,961,472,000,000

Endurance (in days) = 9,961,472,000,000 / (4 X (6,016,204,800/64)) = 26,492 days
Endurance (in years) = 199,229 days / 365 = 72.59 years

I think they will get there.

smugiri · #7 Post by **smugiri** » Tue Dec 20, 2005 9:32 pm

OK, assuming that that last post was tongue-in-cheek:

compare with commercial hard drives:

MTBF is 1.2 million hours: this translates to ( again purely unscientific but we are not scientists spliting hairs here, right?, after all, whats 1 million hours difference in MTBF between friends ? )
1,200,000 hours / (24*365) years which app= 136 years.

So your data will be good for 70 years after you are dead .. ( assuming that you are 0 years old and that the typical human lifespan is 70 years ) .. do you want you Beatles/Commodores/Britney Spears mp3s/videos and/or Steve Ballmer podcasts to last that long?

I say be brave and go with the 70 year flash lifespan. I would hate to think that my data / media collection could outlive me ....

emorphien · #8 Post by **emorphien** » Tue Dec 20, 2005 10:07 pm

Your source didn't work for me, so I can't read and see what you are referring to. My main concern stems from the fact that most conventional flash memories (such as in thumb drives or memory cards for cameras) isn't designed for constant rewriting.

So unless you can show me something actually comparing the read/write duty cycle of your average compact flash card to a hard drive and showing it holds up favorably I don't know what the heck you are going on about. I know some new manufacturing techniques have come online in the past year, but I don't believe they significantly addressed the read/write longevity.

And I'm not concerned about power consumption or impact resistance. We may however be talking about different "flash" memories as far as the duty cycle or number of read/writes the media can withstand.

jwcrim · #9 Post by **jwcrim** » Tue Dec 20, 2005 10:45 pm

smugiri wrote:OK, assuming that that last post was tongue-in-cheek:

compare with commercial hard drives:

MTBF is 1.2 million hours: this translates to ( again purely unscientific but we are not scientists spliting hairs here, right?, after all, whats 1 million hours difference in MTBF between friends ? )
1,200,000 hours / (24*365) years which app= 136 years.

So your data will be good for 70 years after you are dead .. ( assuming that you are 0 years old and that the typical human lifespan is 70 years ) .. do you want you Beatles/Commodores/Britney Spears mp3s/videos and/or Steve Ballmer podcasts to last that long?

I say be brave and go with the 70 year flash lifespan. I would hate to think that my data / media collection could outlive me ....

Those are sobering thoughts.

On the other hand I still have my grandfather's 1903 78 RPM record of Harry Lauder singing "Oh how I hate to get up in the morning." Caruso lovers must have sneered.

My grandfather died before I was born but a century later I'm playing his record (also to the sneers of fellow inhabitants).

Like you, he may well have preferred a truncated 70 year media lifespan. On the other hand he might rest easier if he knew that I like getting up in the morning no better than he and Lauder did.

smugiri · #10 Post by **smugiri** » Tue Dec 20, 2005 10:47 pm

@emorphien:

I think you may have missed my point.

I am not doing a scientific study of the emperical characteristics of CF vs. magnetic disk media and I hope that you did not get the impression that I was trying to.

My argument is that we are talking about using CF drives in one's home or even work computer. In this situation, I am arguing that we are not trying to archive our personal or work data for eternity - thats what the smithsonianand iron mountainare for .... and even it is the case, by your calculations , data recorded on CF will outlive the typical human being. Think about this, you told us that CF data has a life of about 70 years which is the average lifespan of a human being if we consider anyone on the planet, not just someone in the western world where lifespans are longer ( you can check if you are average here). Assuming this is true, even if one begins to amass data the moment they are born, the average CF card with the performance characteristics of CF cards you described which are on the market today will hold the data of this average person from the moment they are born until the moment they die.

And here is one of the fundermental flaws in your argument, you calculations assume that each and every data block of the CF cards will be being written onto at "Write Frequency = 6,016,204,800 KB per day (68 MB/sec) " for each and every day of the 70 years for which they are in use. Sounds like this may not be correct to me. If you think about it, most data on disk is WORM ( write once read many) in nature ( my Beatles/Commodores/Britney Spears mp3s/videos and/or Steve Ballmer podcasts as well as my personal email that is useless to anyone but me, all my tax returns, useless to anyone but me and archives of my forum.thinkpads.com posts, again, of doubtful value to anyone but me ).

This is the reason that I thought that your post was tongue in cheek.

smugiri · #11 Post by **smugiri** » Tue Dec 20, 2005 10:54 pm

@jwcrim:

do you have an mp3 of that (Harry Lauder singing "Oh how I hate to get up in the morning.")?

I would love to take a listen ....

and maybe add it to my mp3 collection ..

who knows, maybe 70 years from now, people will be sneering at my grandson as he plays Lauder then Britney then Ballmer from my Thinkpad using CF media ( no hard drives for me, I am dropping these magnetic suckers as soon as I can get CF that my HD caddy will take )...

jwcrim · #12 Post by **jwcrim** » Tue Dec 20, 2005 10:55 pm

emorphien wrote:Your source didn't work for me, so I can't read and see what you are referring to. My main concern stems from the fact that most conventional flash memories (such as in thumb drives or memory cards for cameras) isn't designed for constant rewriting.

So unless you can show me something actually comparing the read/write duty cycle of your average compact flash card to a hard drive and showing it holds up favorably I don't know what the heck you are going on about. I know some new manufacturing techniques have come online in the past year, but I don't believe they significantly addressed the read/write longevity.

And I'm not concerned about power consumption or impact resistance. We may however be talking about different "flash" memories as far as the duty cycle or number of read/writes the media can withstand.

Try this link:
http://www.bitmicro.com/press_resources ... sd_db2.php

Toshiba, Samsung, Intel and Micron are suddenly throwing huge bucks at this. Bitmicro is a current high-end low-volume player who will probably not figure in the next round of advances.

smugiri · #13 Post by **smugiri** » Tue Dec 20, 2005 11:33 pm

ok, caught myself in a major whoopsie here ....

just realized that I got posters a little mixed up on this one.

@jwcrim posted the math and not @emorphien so the calculation was definitely tongue in cheek ....

please excuse my bad @jwcrim / @emorphien ....

emorphien · #14 Post by **emorphien** » Wed Dec 21, 2005 1:45 am

Looks like slightly different flash tech. I'll read it later when I've had some rest.

If the OP is considering these newer forms of flash memory then it would work presumably, but that's not what I was thinking of.

For example if you wanted to boot windows off a thumb drive (and I know someone who has been tinkering with this) you probably shouldn't expect it to last.

jwcrim · #15 Post by **jwcrim** » Wed Dec 21, 2005 9:29 am

smugiri wrote:@jwcrim:

do you have an mp3 of that (Harry Lauder singing "Oh how I hate to get up in the morning.")?

I would love to take a listen ....

and maybe add it to my mp3 collection ..

who knows, maybe 70 years from now, people will be sneering at my grandson as he plays Lauder then Britney then Ballmer from my Thinkpad using CF media ( no hard drives for me, I am dropping these magnetic suckers as soon as I can get CF that my HD caddy will take )...

Here's a source:

http://cylinders.library.ucsb.edu/searc ... rtOrder=ia

smugiri · #16 Post by **smugiri** » Wed Dec 21, 2005 10:12 am

Thanks for the link to http://cylinders.library.ucsb.edu/ : this is quite the esoteric site: how did you manage to dig this up? I'm sure I shall spend tons of time exploring this site.

@emorphien:
people have been booting OS'es off of flash media for a while now, just not Windows. Check out this link, this link and this link among many thousands of others.

smugiri · #17 Post by **smugiri** » Wed Dec 21, 2005 10:48 am

Found a real media version of "Oh How I Hate to get up in the Morning "by Arthur Fields here from a 1918 hand cranked Victrola, no less ....

Esben · #18 Post by **Esben** » Thu Dec 22, 2005 7:19 pm

Correct me if this calculation is off-base.

The example I've come up with, is based on a Kingston Elite Pro SLC-CompactFlash card, with 4 GB capacity, employing wear-leveling technology. That means the same logical sector isn't written to the same physical one. It endures 100,000 write cycles. Performance is 8 MB/sec write.

4 GB rewritten 100,000 times = 400,000,000 MB
400,000,000 divided by 8 MB/s write = 50,000,000 seconds of writing
50,000,000 / 60 / 60 / 24 = 578 days of continious writing.

I hardly think non-stop writing will be the case, assuming a more civilian 10 % write duty cycle, it means we have a lifetime of 15 years.

stangri · #19 Post by **stangri** » Wed Dec 28, 2005 7:11 am

You guys have any idea how much these solid state drives cost? I've had a corporate estimate from BitMicro (I believe), they're about $5k each if you buy a few dozen.

So yeah, maybe a few years down the road, hopefully they'll fix the leongevity issue by then.

whizkid · #20 Post by **whizkid** » Wed Dec 28, 2005 10:41 am

You can get a 1GB CF card for $50, or 2GB for $95. A CF card pretty closely emulates an ATA interface, so a simple circuit board should be able to make a CF card appear as an ATA drive.

Someone should be able to sell these boards for $25. Hmmm. Time to fire up the CAD software!

stangri · #21 Post by **stangri** » Wed Dec 28, 2005 12:35 pm

whizkid - I also bet that you can get the actual memory modules from Samsung for half as much (if not fraction of that), however we aren't seeing the cheap solid state HDDs flooding the market, are we?

whizkid · #22 Post by **whizkid** » Wed Dec 28, 2005 12:42 pm

Well we kind of are, actually. A 1GB USB drive is nearly the same price as just a memory card. I bought a USB SD card reader for under $10.

I just think there isn't much demand for a CF to ATA device. Most people want more storage capacity over silence. When $50 will buy you a 1GB card or a 40GB drive for your laptop, not too many folks will go with the card.

jwcrim · #23 Post by **jwcrim** » Fri Dec 30, 2005 11:57 pm

whizkid wrote:Well we kind of are, actually. A 1GB USB drive is nearly the same price as just a memory card. I bought a USB SD card reader for under $10.

I just think there isn't much demand for a CF to ATA device. Most people want more storage capacity over silence. When $50 will buy you a 1GB card or a 40GB drive for your laptop, not too many folks will go with the card.

Here are some IDE/ATA Flash adapter cards:
http://www.pcengines.ch/cflash.htm
http://www.acscontrol.com/Index_ACS.asp ... dapter.htm

With Toshiba, Samsung, Intel and Micron all suddenly throwing billions at making NAND FLASH memory, the economics for this isn't likely to remain at present levels.

BruisedQuasar · #24 Post by **BruisedQuasar** » Mon Jan 02, 2006 11:30 am

This hardiness of Hard Disk Drives versus Compact Flash drives is a central question among serious true mobile PC officianados.

The conclusion repeatedly is that at this time Flash Drives cannot replace a mechanical hard drive as HHD is accessed hundreds of times an hour, which would quickly wear out flash drives.

The majority of CE device enthusiasts are convinced Flash technology as it is now could not replace a hard disk. There are some of us however who are not convinced. The fact is not even the most obsessive Sharp handheld user has managed to yet wear out flash in a Zaurus and like all true mobile devices the Zaurus line of mobile PCs have no moving parts. The newest Zaurus works around a 4MB flash drive.

Furthermore, enterprise handhelds such as the NEC 720 through 900c and the HP Jornada 680 through 728 even when used heavily for four years by the Dutch Railways and by hospitals did not wear out wither internal flash or CF card flash.

But pessimists point out mobile devices run highly efficient, stripped down versions of software & the devices are not constantly read-writing like a Windows or Mac desktop or lug-a-bout PC does.

jwcrim · #25 Post by **jwcrim** » Mon Jan 02, 2006 3:40 pm

BruisedQuasar wrote:This hardiness of Hard Disk Drives versus Compact Flash drives is a central question among serious true mobile PC officianados.

The conclusion repeatedly is that at this time Flash Drives cannot replace a mechanical hard drive as HHD is accessed hundreds of times an hour, which would quickly wear out flash drives.

The majority of CE device enthusiasts are convinced Flash technology as it is now could not replace a hard disk. There are some of us however who are not convinced. The fact is not even the most obsessive Sharp handheld user has managed to yet wear out flash in a Zaurus and like all true mobile devices the Zaurus line of mobile PCs have no moving parts. The newest Zaurus works around a 4MB flash drive.

Furthermore, enterprise handhelds such as the NEC 720 through 900c and the HP Jornada 680 through 728 even when used heavily for four years by the Dutch Railways and by hospitals did not wear out wither internal flash or CF card flash.

But pessimists point out mobile devices run highly efficient, stripped down versions of software & the devices are not constantly read-writing like a Windows or Mac desktop or lug-a-bout PC does.

I haven't looked into this but you can imagine things that would reduce the erasures such as buffering a block in a small battery backed RAM buffer so that many pages and pointers could be changed before a whole NAND block is rewritten. Then as was mentioned the controller can rotate the use of blocks so that the memory wears out uniformly.

The other thing to note is that when you do wear out a block the result is immediate and local, it is not a catastrophic event like a hard disk crash. It doesn't happen to data that was previously written and checked. Error correction catches it immediately and the controller gets to move things around and rope off the worn-out block. The result is a gradual and graceful reduction in memory size that you can be warned about and have a chance to alleviate.

Paul Unger · #26 Post by **Paul Unger** » Wed Jan 04, 2006 2:44 pm

Here's one project trying to do what the original 'poster' asked about:

http://www.explan.co.uk/solo/index.shtml

I've written the developers, and one of their main goals is to build a machine with no moving parts (e.g., spinning HD) for use in harsh environments (extremes of heat, humidity, dust, etc.). To that end they've used CF (and MicroDrives; so much for "no moving parts"

) for ROM and RAM.

Obviously, I'm interested enough in the project to have written them for more information. I think it's a fantastic idea! Don't get me wrong--I LOVE MY T42--but I have colleagues who could never afford such a well designed machine as the Thinkpad line (which has held up quite well in this "harsh environment"), and I've seen lesser machines drop like flies out here! A 'solid state' machine with low power consumption is what a lot of the world needs (if indeed they need a computer . . . )

Anyway, that's my two bits. Thanks for the enjoyable reading!

Paul

Website · #27 Post by **tfflivemb2** » Fri Jan 06, 2006 8:41 pm

How about this:

http://forum.thinkpads.com/viewtopic.php?t=19034

BruisedQuasar · #28 Post by **BruisedQuasar** » Sat Jan 14, 2006 3:33 pm

[/quote] I haven't looked into this but you can imagine things that would reduce the erasures The other thing to note is that when you do wear out a block the result is immediate and local, it is not a catastrophic event like a hard disk crash. It doesn't happen to data that was previously written and checked. Error correction catches it immediately and the controller gets to move things around and rope off the worn-out block. The result is a gradual and graceful reduction in memory size that you can be warned about and have a chance to alleviate.[/quote]

Believe it or not the poor and verbose design of Microsoft O/S from Windows based on DOS core o/s, NT X.X and Wndows Server not only leaves Windows O/S highly vulnerable to attack but creates most of the excessive hard disk calls and saves and Hard Drive problems that call for users to waste time and money on Hard Drive utilities such as defrag and various repairs. UNIX variants such as Linux Core O/S and MAC O/S do not present these time wasting maintenance issues. In fact, I have a PC running Linux that has not been restarted in two years and I have another LINUX PC that children shut down and restart all the time by the power button (for a year now) without any adverse effect.

Any full fledged PC that has no moving parts would have to be a PC that runs a much better designed O/S than anything Microsoft pushes, which would be a relief as the programming standards would also tighten. For instance, there would be an end to software bloat?

stangri · #29 Post by **stangri** » Sun Jan 15, 2006 3:51 pm

Did you personally do any tests to prove that Windows Server causes "excessive" hard disk drive usage comparing to linux/masocx or are you just going by your opinion, unsubstantiated by any proven data?

Could you also elaborate on the specific features of the NTFS which make it problematic and the details on the HPFS/linuxfs implementations which make them problem-free?

You made a statement about the internals of the OS, so I would expect your response to be technical enough for the OS core developer.

smugiri · #30 Post by **smugiri** » Sun Jan 15, 2006 9:38 pm

@stangri

Did you personally do any tests to prove that Windows Server causes "excessive" hard disk drive usage comparing to linux/masocx or are you just going by your opinion, unsubstantiated by any proven data?

Could you also elaborate on the specific features of the NTFS which make it problematic and the details on the HPFS/linuxfs implementations which make them problem-free?

You made a statement about the internals of the OS, so I would expect your response to be technical enough for the OS core developer.

Dude!!! You are bringing out the big guns.
I agree though. If you talk the talk, you gotta walk the walk. @BruisedQuasar, its time to pay the piper.

Using Compact Flash base IDE Drives Instead of Hard Disks

Using Compact Flash base IDE Drives Instead of Hard Disks

i beg to differ ...

wow ....

Who is online