Help choose SSDs for SPCR's test systems

Silencing hard drives, optical drives and other storage devices

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Copper
*Lifetime Patron*
Posts: 587
Joined: Sun Feb 22, 2004 8:43 am

Re: Help choose SSDs for SPCR's test systems

Post by Copper » Sat Dec 08, 2007 11:41 am

You'll want to pay close attention to random write speed. I have an inexpensive Transcend 8GB IDE SSD rated at ~20MB sequential read and ~10MB sequential write. Day to day use running apps and booting and shutting down all mimic the 5400 RPM drive performance you ask. I can even clone ~2GB from and restore ~2GB to the drive in as little as 3 minutes.

But! If you'll actually be installing Windows often you'll be bumping into what is for all but a few (or all?), the SSD weak spot: Exremely poor random write performance. I didn't actually time it, but the one time I installed on the SSD was all of two hours*. Windows updates were equally as brutal. Good luck finding published random write speeds. :(

*Whether it was an hour and a half or three hours is unimportant. It was noticably way out of proportion to an HDD. Hardly an issue if you install and go, but if installing is a common occorance it can't help but be an issue.
Last edited by Copper on Sat Dec 08, 2007 11:49 am, edited 1 time in total.

jojo4u
Posts: 806
Joined: Sat Dec 14, 2002 7:00 am
Location: Germany

Post by jojo4u » Sat Dec 08, 2007 11:48 am

from viewtopic.php?t=39278:

Speed
The installation of the USBoot filter driver ("driveguard") gave a write speed boost. Don't know why, though.
My sticks are a Buffalo Firestix-R 4 GB and a Buffalo Firestix-S 1 GB for all my applications which don't rely on a particular Windows installation.
Some basic benchmarks for the Firestix-R.
read access time: 0,5 ms
write access time: 17 ms
sequential read: 32 MiB/s
sequential write: 19 MiB/s
c:\program files -> read: ~12 MiB/s
c:\program files -> write: ~5 MiB/s
very small files (~1 KiB) read: 1,5 MiB/s
very small files (~1 KiB) write: ~1 MiB/s
The Firestix-S was examined here by xbitlabs.com.

cloneman
Posts: 448
Joined: Sat May 21, 2005 9:48 am

Post by cloneman » Sat Dec 08, 2007 12:59 pm

If you do decide to go for the CF card option, you should look into CF adapters that support DMA or CF to SATA adaptors. Not how much faster those two are, but certainly worth investigating.

Or you can get Gigabyte I-Ram. Too bad they don't make those with DDR2, it would be so cheap to run!

jojo4u
Posts: 806
Joined: Sat Dec 14, 2002 7:00 am
Location: Germany

Post by jojo4u » Wed Dec 12, 2007 12:34 pm

Ok I've found something new in the recent C't (26/07)
Addonics Quad-CF PCI Adapter

This beast let's you combine up to 4 CF cards in RAID 0/1/10. More than 2 fast cards in RAID-0 are pointless since the internal IDE channel gets saturated.
They used a Kingstom Ultimate 266x and got 41/24 MiB/s read/write. In RAID-0 they got 74/54 MiB/s read/write. Acess time was 0,4 ms, write access was longer with 22 ms.
Now for the really interesting stuff:
Windows XP worked after using a self-compiled F6 driver floppy disk. And they also tried a Sandisk Extreme-IV and switched it to fixed mode using atcfwchg. The card performed similar to the Kingston.

The big news for me ist that atcfwchg is said to work with the Sansdisk Extreme-IV cards. This could mean standby and more than one partition in Windows ;)

Copper
*Lifetime Patron*
Posts: 587
Joined: Sun Feb 22, 2004 8:43 am

Post by Copper » Wed Dec 12, 2007 12:49 pm

That thing is nice. I wonder if the high write latency is an indicator that the flash will have poor random write performance - if they are related?

jojo4u
Posts: 806
Joined: Sat Dec 14, 2002 7:00 am
Location: Germany

Post by jojo4u » Wed Dec 12, 2007 12:58 pm

Well some more good news from http://www.hjreggel.net/cardspeed/index_de.html
The mentioned Kingston Ultimate 266x seems to be fixed from factory. There are all UDMA CF cards listed and some tested. And some of the tested have the magic word "fixed" as ATA type.

I'll copy the table (a hacked to look nice)

Code: Select all

read							write
43.11MB/s 	294x 		46.30MB/s 	316x 		UDMA-5 		CF-Removable	SONY 300x UDMA CF 8GB
41.82MB/s 	285x 		44.99MB/s 	307x 		UDMA-5 		CF-Removable	Lexar Professional UDMA 300x CF 8GB  		
43.00MB/s 	293x 		43.65MB/s 	298x 		UDMA-4 		CF-Removable	SanDisk Extreme Ducati Edition CF 8GB
40.08MB/s 	273x 		38.97MB/s 	266x 		UDMA-4 		CF-Removable	SanDisk Extreme IV CF 4GB
44.12MB/s 	301x 		38.74MB/s 	264x 		UDMA-5 		Fixed + CF-Removable	Kingston Ultimate 266x CF 8GB
45.26MB/s 	309x 		36.60MB/s 	249x 		UDMA-4 		Fixed + CF-Removable	Transcend 266x High-Speed CF 8GB
42.90MB/s 	282x 		35.69MB/s 	243x 		UDMA-4 		Fixed + Standard	Transcend ULTRA Industrial CF 8GB
47.66MB/s 	325x 		29.29MB/s 	200x 		UDMA-4 		Fixed + CF-Removable	A-Data Turbo 266x CF 8GB
28.01MB/s 	191x 		21.54MB/s 	147x 		UDMA-4 		Fixed + CF-Nicht-Konform	TwinMOS Ultra-X CF 300X 4GB
34.26MB/s 	233x 		15.26MB/s 	104x 		UDMA-4 		Fixed + CF-Removable	Transcend 133x Ultra Speed CF 16GB
I have no idea wether Fixed + Standard or Fixed + CF-Nicht-Konform (not conforming) works with Windows, so just choose one Fixed + CF-Removable. You can go the URL above and look at single tests for each product.

jojo4u
Posts: 806
Joined: Sat Dec 14, 2002 7:00 am
Location: Germany

Post by jojo4u » Wed Dec 12, 2007 1:10 pm

Ok after further researching about the ATA type I can says that some cards just detect wether they are behind an ATA host and just switch the type. This means a card that has Fixed + CF-Removable sets itself to either one of them.

Sources:
http://hjreggel.de/cardspeed/speed-news.html
"It also supports automatic switching between ATA ID 848A (CF Removable) and 044A (ATA Fixed), so it might be a good choice for a homemade SSD."
http://www.dforum.net/showthread.php?p=1011058 (german)

ist.martin
Posts: 220
Joined: Fri Jul 18, 2003 11:59 am
Location: Vancouver, B.C.

HyperDrive

Post by ist.martin » Tue Dec 18, 2007 2:51 pm

And then there is always a HyperDrive - http://www.hyperossystems.co.uk/ - maybe they'll donate one, tell 'em that SPCR is the standard where high-performance and silencing meet :)

andyb
Patron of SPCR
Posts: 3307
Joined: Wed Dec 15, 2004 12:00 pm
Location: Essex, England

Post by andyb » Tue Dec 18, 2007 3:29 pm


PostPosted: Tue Dec 18, 2007 10:51 pm Post subject: HyperDrive
And then there is always a HyperDrive - http://www.hyperossystems.co.uk/ - maybe they'll donate one, tell 'em that SPCR is the standard where high-performance and silencing meet Smile
Donations welcome.

The only price of this thing I have seen (not looked) is enormous.

http://www.scan.co.uk/Products/Products ... bnails=yes


Andy

Elijah86
Posts: 41
Joined: Wed Aug 22, 2007 2:11 pm
Location: Plymouth MN

Post by Elijah86 » Thu Dec 20, 2007 1:56 pm

go with a mtron. plus in a 9 disk raid its kind of fast. http://www.nextlevelhardware.com/storage/battleship/

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Fri Dec 21, 2007 6:48 am

I noticed in that SSD review that power consumption is near 0W. Switching to SSDs will exacerbate the "Low power system not booting with high efficiency PSU?!" problem

I'm also not liking the slowness of these drives in general use. They have a long way to go before they can kill off usage of 3.5" hard drives for the average computer user.

At the least if you decide to use these you should be using Symantec Ghost or an equivalent to make drive images of the SSD to a traditional hard drive or two so that if the drive dies or you get a new SSD in and want to try it out you can do so in a hurry.

All that being said I'm very glad to see that you are looking to make this transition in your test bed. Reducing noise of non reviewed components in a test bed is exactly the thing I'd expect you to do even if it isn't easy. Your efforts and sharing of information are greatly appreciated.

I'm looking forward to the day the industry as a whole make it so that even non reviewers would see this sort of transition as a reasonable undertaking.

ist.martin
Posts: 220
Joined: Fri Jul 18, 2003 11:59 am
Location: Vancouver, B.C.

Not convinced SSD is too slow for general usage

Post by ist.martin » Fri Dec 21, 2007 9:13 am

I'm also not liking the slowness of these drives in general use.
I have been using a SSD drive in a 0 dB VIA Epia system for the past 3 1/2 years. For the first two years, I had an 800 MB Simple Tech drive. Lately I've used a 2 GB PQI drive. The drive houses 98 SE and all the apps I need for general usage. I use the system for at least 50 hours a week, using it for my home-based work.

Lately, my system is definitely too slow for general usage, but I don't blame it on the SSD (it worked fine for the first few years). I think the VIA CPU is too underpowered to deal with modern web browsing.

I have one of these on the way, http://www.cappuccinopc.com/slimpro-sp635f.asp , Celeron M 1.4 with their 'industrial' 8 GB SSD (not sure who makes it). It will have XP on it, which I'll strip down with XPLite. I'll report back how quick it is for general usage.

halcyon
Patron of SPCR
Posts: 1115
Joined: Wed Mar 26, 2003 3:52 am
Location: EU

Post by halcyon » Fri Dec 21, 2007 1:44 pm

Just two quick points:

USB Flash drives, try to get OCZ ATV Turbo (turbo being the defining word here) in either 4GB or 8GB. That is the highest performing flash USB drive (consistently from test to test) that I've been able to find. Beats anything/everything by Corsair, Buffalo, Patriot, etc. For both read and write speeds. I just did a very extensive research on this, not more than week ago.

CF cards. Try to get a lexar UDMA 300x card (UDMA being the defining word here). This should have highest controller bandwidth, although in USB readers the Lexar 133x Professional line is almost on par. Don't know how it handles when attached to PATA. There might be something even faster on the CF market, I'm not 100% up-to-date.

As for SSD drives variety is the key here. Maybe you can even score a pre-production unit from Toshiba on their upcoming models.

http://techon.nikkeibp.co.jp/english/NE ... 12/144128/

BTW, thanks for all the great reviews, esp. the recent WD GP review, which was just brilliant and not because the drive itself was good :)

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Fri Dec 21, 2007 2:08 pm

halcyon -- you're welcome.

And thanks for your investigation with USB flash drives, andyb.

Corsair supplied a couple of 8gb Flash Voyager GT drives, which appear to be decently fast. I haven't had any time to try the www.usboot.org/ that jojo4u used to load windows on his flash drive. Not sure it'll happen during this crazy season...

jojo4u
Posts: 806
Joined: Sat Dec 14, 2002 7:00 am
Location: Germany

Post by jojo4u » Fri Dec 21, 2007 3:27 pm

halcyon wrote: USB Flash drives, try to get OCZ ATV Turbo (turbo being the defining word here) in either 4GB or 8GB. That is the highest performing flash USB drive (consistently from test to test) that I've been able to find. Beats anything/everything by Corsair, Buffalo, Patriot, etc. For both read and write speeds. I just did a very extensive research on this, not more than week ago.
Funny enough the only test I know of this stick does not praise it very much. A HDTach test showed 33 MiB/s read and only 14 MiB/s write. A Corsair Flash Voyager GT got 33/26.
Look it up here: http://www.hardtecs4u.com/reviews/2007/ ... index3.php

halcyon
Patron of SPCR
Posts: 1115
Joined: Wed Mar 26, 2003 3:52 am
Location: EU

Post by halcyon » Sat Dec 22, 2007 2:32 am

jojo4u,

thanks for that link, I had missed that test altogether (probably not the only one). That is indeed different from the other tests I've read, so it's good you brought it up.

Here are the test I was referring to:

http://www.hardwarecanucks.com/forum/ha ... iew-5.html
http://www.matbe.com/articles/lire/511/ ... ir-vs-ocz/
http://www.everythingusb.com/ocz_atv_tu ... 13894.html
http://www.mikhailtech.com/content/view/277/28/1/1/
http://www.allround-pc.com/index.php?re ... ericht.htm
http://www.cdrinfo.com/Sections/Reviews ... 4&PageId=1

I have removed most of the tests that didn't test Voyager GT in the same test settings.

Many tests are showing over 30 MB/S for OCZ AVT Turbo write speed (in HDtach 3), although some are showing below.

I think the speed might also be affected by size of the memory. At least I think 8GB Voyager GT performs slightly worse than the 4GB variant in writing. Also the 8GB ATV Turbo variant seems to perform worse on write tests:

http://www.legionhardware.com/document.php?id=695&p=2

Another thing that seems to affect the speed (esp. for writes?) is the USB port (chipset) variant. Otherwise I cannot explain the more than 30% difference in write speed from test to test (when the size of the memory, test software and settings are the same).

Write speed is clearly affected by the workload as well (i.e. the software and settings performing the write test). Nothing new there.

Then again, now that I did a new search round up, I found also these:

http://napedy.idg.pl/news/129125.html
http://www.hardware.no/tester/lagring/o ... gb/45426/3

So, I think it's a crapshoot. Some tests show ATV Turbo some show Voyager GT on the top.

So, would it be safe to say that either of those would be good and MikeC already has the Voyager GT :)

EDIT PS. Oops. I thought this was a one page thread. Hadn't even noticed all the other posts. Well, to live and learn :)

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Sat Dec 22, 2007 1:21 pm

Had a go at www.usboot.org this morning. Tried several different ways of configuring the system (in steps 2 & 3) -- unfortunately no go every time. I get a msg saying "OS missing". One thread at usboot.org suggests this is a hardware issue -- that the motherboard might not support booting from the USB flash drive. The BIOS lets me choose between the HDD and the Corsair flash drive as the first boot drive, tho. Anyway, I am stymied for now. May have to read through the instructions again very carefully and maybe give it a last go before giving up. Also, see if I can get WinXP to boot on one of these Corsair drives on another motherboard.

frostedflakes
Posts: 1608
Joined: Tue Jan 04, 2005 4:02 pm
Location: United States

Post by frostedflakes » Sat Dec 22, 2007 1:44 pm

Oddly enough I've had this issue as well with my Flash Voyager GT, even with different operating systems (I believe I tried Windows XP, Damn Small Linux, and PuppyLinux). I had always assumed it was something to do with my motherboard, but perhaps it's an issue with the Corsair drives. Is a shame, because they have a lot of potential as OS drives.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Sat Dec 22, 2007 1:57 pm

frostedflakes wrote:Oddly enough I've had this issue as well with my Flash Voyager GT, even with different operating systems (I believe I tried Windows XP, Damn Small Linux, and PuppyLinux). I had always assumed it was something to do with my motherboard, but perhaps it's an issue with the Corsair drives. Is a shame, because they have a lot of potential as OS drives.
:!: :? :|
uh-oh... That's bad! Anyone succeed booting/running an OS w/Flash Voyager GT? If there's even one success story, then there's hope.

Felger Carbon
Posts: 2049
Joined: Thu Dec 15, 2005 11:06 am
Location: Klamath Falls, OR

Post by Felger Carbon » Sat Dec 22, 2007 3:07 pm

halcyon wrote:Another thing that seems to affect the speed (esp. for writes?) is the USB port (chipset) variant. Otherwise I cannot explain the more than 30% difference in write speed from test to test (when the size of the memory, test software and settings are the same).
In any flash memory write test, the size of the files being written is absolutely critical. This is not unique to Flash Voyager; it's a property of all flash drives.

If you want to see the very highest test numbers, write a Ghost image file of 2GB. ;-)

The very worst is writing a large number of 1KB files such as the Cookies folder and the like. Most write tests will tell you the size of the files being written for that reason.

DepSPC
Posts: 1
Joined: Sat Dec 22, 2007 9:31 pm

Re: Help choose SSDs for SPCR's test systems

Post by DepSPC » Sat Dec 22, 2007 9:52 pm

Copper wrote:You'll want to pay close attention to random write speed. I have an inexpensive Transcend 8GB IDE SSD rated at ~20MB sequential read and ~10MB sequential write. Day to day use running apps and booting and shutting down all mimic the 5400 RPM drive performance you ask. I can even clone ~2GB from and restore ~2GB to the drive in as little as 3 minutes.
[/i]
Hi, I'm a bit new to this and apologise if you've covered this. I want a basic solution for an SSD in a p3/p4 class laptop on parallel ATA- PATA IDE, not SATA. The machine also has some PCMCIA slots.

There is a cheapish transcend 32 GB on ebay and stores for IDE. Reports of slowness. I want a moderately useable machine web browsing and email. I hear these SSDs are fast on startup, but infer slow maybe on some apps.

There are also 1.8 SSDs Samsung on ebay. I think these somehow go in a PCMCIA slot with some sort of an adapter. I am wondering if in a 500 or 900 mhz machine these are ok or not?Any help or way to get other types into those inputs would be much appreciated. Many I have seen now are SATA so I can't use. Thx

JazzJackRabbit
Posts: 1386
Joined: Fri Jun 18, 2004 6:53 pm

Post by JazzJackRabbit » Sun Dec 23, 2007 9:14 pm

I haven't read every single post and I'm sure it has been said already, but why not go with 2.5" drive in a scythe enclosure? You could actually experiment first with enclosing drive in 2.5" enclosure, and then enclosing 2.5" enclosure in a regular scythe quiet drive enclosure. I don't know if it will work for sure, but if it does you'll have zero noise coming from the hard drive. Flash based drives are too expensive at the moment.

Regarding longevity, buy Acronis True Image. It has the ability to recreate hard drive contents from an image. I've used it to clone Server 2003 OS drive and it worked perfectly. Just create a backup copy and use it should occasion arise. It's only $50 regular price, pity you live in Canada, Dell USA is currently having it on sale for only $10 after MIR.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Sun Dec 23, 2007 9:48 pm

Actually, I've cloned the OS to a Samsung 40GB PATA 2.5", which is still one of the quietest ever, and suspended it in a closed cell foam box, mostly sealed, with a thick layer of open cell foam on the top. The PSU is a fanless 300W Fortron Zen. So in fact, noisewise, it is now silent, even quieter than the 18~19 dBA/1m VGA test platform.

But the fact of the HDD having moving parts, and the need to handle the motherboard and its connections quite a lot, and often, to install and uninstall HS still makes this a bit touchy. Yes, there is now a backup image of the drive on another system, so it's relatively easy to replace, but I will still aim to transition to a SSD some time in the future as long as we're continuing to use a real system (rather than a heater block simulator like those used by FrostyTech, Overclockers.com and others).

Why not move to a heater block? The arguments for them are strong -- convenient, robust, repeatable -- but not nearly as good a test of the mounting mechanism as a real board. I've always maintained that the mounting mechanism is an integral part of the HS, and the HS must be tested in this context. I have yet to see a heater block simulator that really replicates a 775 board.
Last edited by MikeC on Mon Dec 24, 2007 12:08 am, edited 1 time in total.

Moose
Posts: 3
Joined: Wed Nov 29, 2006 10:56 am
Location: UK

USB Flash Drives

Post by Moose » Sun Dec 23, 2007 11:45 pm

There has been some mention of using USB Flash drives instead of SSDs, however when using Ordinary flash drives to load an OS there are serious lifespan limitations as compared to SSDs, especially using NTFS or more sophisticated file systems. SSDs have special wear levelling algorithms which increase lifespan compared to ordinary flash drives. Flash drives have limited write cycles before they die so without sophisticated wear levelling (which usb flash drives don't tend to have) any write intensive operation can write too much to the same cells wearing them out and dramatically reducing their life.

halcyon
Patron of SPCR
Posts: 1115
Joined: Wed Mar 26, 2003 3:52 am
Location: EU

Post by halcyon » Mon Dec 24, 2007 4:31 am

Felger Carbon wrote: In any flash memory write test, the size of the files being written is absolutely critical. This is not unique to Flash Voyager; it's a property of all flash drives.
I'm in perfect agreement and this goes without saying. I may have been unclear above, but this is exactly what I tried to write.

However, just as I wrote above, results for exact same USB Flash memory drive vary, even when the program, settings & test load is identical.

Possible explanations: driver/chipset.

kocoman
Posts: 2
Joined: Mon Jan 15, 2007 7:03 am

Post by kocoman » Fri Dec 28, 2007 5:01 pm

Is there any website that shows which SATA or IDE or IDE DMA to CF adapter performs the best? I looked at hjreggel's cardspeed site but it doesn't have any, only USB, FW, ..etc

walkingjohn
Posts: 29
Joined: Sun Mar 18, 2007 9:51 am
Location: Ontario, Canada

Post by walkingjohn » Sat Dec 29, 2007 5:35 pm

Felger Carbon wrote:If you want to see the very highest test numbers, write a Ghost image file of 2GB. ;-)

The very worst is writing a large number of 1KB files [....]
Can you help me understand why that is? I gather one of the advantages of ss is tiny access times, which would tend to speed up random writes. Is it a limitation of flash or something to do with the interface?

michaelb
Posts: 53
Joined: Mon Nov 03, 2003 3:35 pm
Location: Texas

Post by michaelb » Sun Dec 30, 2007 1:17 am

kocoman wrote:Is there any website that shows which SATA or IDE or IDE DMA to CF adapter performs the best? I looked at hjreggel's cardspeed site but it doesn't have any, only USB, FW, ..etc
PATA IDE to CF I think is just connecting the right pins to the right places. So the adapter is passive. Have read that some old, cheap ones didn't connect all pins needed for current DMA, but hopefully newer ones do. I'm guessing it's a good sign if the card says "DMA" on it, like some of the $4 ones on ebay do, but I'm not experienced or expert in this.

I'd like to see a comparison of SATA-CF adapters, if there is one.
Have seen some cheap on ebay, but don't know if maker/model info is available. Addonics sells one in notebook and desktop format. Desktop version was reviewed on Tom's Hardware: http://www.tomshardware.com/2005/08/18/ ... act_flash/
Unfortunately, they used a slow CF card. They do report very quick access times though.

I suppose it might be possible to use a $4 CF to IDE adapter, and then just connect that to an IDE-SATA bridge?

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Sun Dec 30, 2007 7:42 am

walkingjohn wrote:
Felger Carbon wrote:If you want to see the very highest test numbers, write a Ghost image file of 2GB. ;-)

The very worst is writing a large number of 1KB files [....]
Can you help me understand why that is? I gather one of the advantages of ss is tiny access times, which would tend to speed up random writes. Is it a limitation of flash or something to do with the interface?
Writing to a single cell in a flash module involves charging, clearing, writing, charging an entire row of cells (or something similarly byzantine). It's not as simple as the write process on a hard drive.

This is also why I'd never do RAID 5 on SSDs. RAID 5 has a heavy write penalty so you are just mulitplying the write delay by another write delay. Raid 1 or 1+0 would make more sense if you are wanting to avoid the risk of RAID 0.
One limitation of flash memory is that although it can be read or programmed a byte or a word at a time in a random access fashion, it must be erased a "block" at a time. This generally sets all bits in the block to 1. Starting with a freshly erased block, any location within that block can be programmed. However, once a bit has been set to 0, only by erasing the entire block can it be changed back to 1. In other words, flash memory (specifically NOR flash) offers random-access read and programming operations, but cannot offer arbitrary random-access rewrite or erase operations. A location can, however, be rewritten as long as the new value's 0 bits are a superset of the over-written value's. For example, a nibble value may be erased to 1111, then written as 1110. Successive writes to that nibble can change it to 1010, then 0010, and finally 0000. Although data structures in flash memory can not be updated in completely general ways, this allows members to be "removed" by marking them as invalid. This technique must be modified somewhat for multi-level devices, where one memory cell holds more than one bit.

michaelb
Posts: 53
Joined: Mon Nov 03, 2003 3:35 pm
Location: Texas

Post by michaelb » Sun Dec 30, 2007 11:20 am

Good point about Raid 5.


I plan to put page swap on a completely separate CF drive, so if those writes kill a CF someday, it will only kill the swap file. With Linux and/or a lot of RAM, page swap may not even be needed.

I keep most my data on a NAS across the room from me. So, writing of most data won't touch CFs. I do plan to keep Firefox data local, but anyone concerned could put the cache (and other similar data often written to) onto a separate CF drive from OS and programs.

Also, I've read that most modern CFs are much longer lasting than early ones, handling 10x as many write cycles. And, on manufacturer web pages for CF models I've looked at, I've seen claims of wear-leveling algorithms.

The larger the CF the more the wear-leveling should help, so using big CFs is another option to improve longevity.

Post Reply