write endurance - solid state flash drives, windows xp, virtualmemory
write endurance - solid state flash drives, windows xp, virtualmemory. Discuss write endurance - solid state flash drives, windows xp, virtualmemory, on Wireless Forums.
write endurance - solid state flash drives, windows xp, virtualmemory
I am interested in booting windows xp off a solid state drive.
The 8GB+ solid state flash drives are a bit expensive for me..
I am looking at 4GB and less(2Gb,1GB)
...
The compact flash cards may not have such good endurance.. - they may
wear out.
but there are companies making what some may market as IDE Flash
Modules . .(e.g. made by Emphase, Innodisk, Transcend.)
They have write endurance of 1 to 4 million cycles, and have a
controller utilising a wear levelling algorithm.
How long will they last running windows xp?
Would I -have- to disable virtual memory / set to no page file. I have
heard that virtual memory wears these things out.. But I wonder if
with such high endurance.
I have read that capacity works for it.. A larger capacity drive will
last longer. More memory cells.
So I wonder if a 2GB drive won't last so long..
BTW, they don't come with SMART data that gives how many more writes
they have / how the disk is.. do they ?
(though I have read that they are more likely to fail due to
controller error - picking a block that is already quite worn, wearing
it out completely and making the disk unusable or unreliable. )
It would be nice if the CF was 3.0, (even though more speed may mean
it wears out quicker!). But first and foremost, I want it to be
reliable. And not priced through the roof. I don't want to be
spending more than $80 / 40UKP. The big solid state drives, like 8GB,
+, are quite expensive.
I am interested in booting different operating systems e.g. FREEBSD..
without worrying about this writing issue.. and how for each OS, to
deal with it.
Re: write endurance - solid state flash drives, windows xp, virtual memory
On Thu, 29 May 2008 22:58:56 -0700 (PDT),
"jameshanley39@yahoo.co.uk" <jameshanley39@yahoo.co.uk>
wrote:
>I am interested in booting windows xp off a solid state drive.
>
>The 8GB+ solid state flash drives are a bit expensive for me..
>
>I am looking at 4GB and less(2Gb,1GB)
>..
>The compact flash cards may not have such good endurance.. - they may
>wear out.
Is there a limit on how many CF cards you can use? You can
get good performance and less concern for wear by using 3 of
them, all three using SLC flash chips (normally would then
be rated for over 10MB/s write speed) and CF4 spec (not just
"CF spec" as some list this but it's CF2.x or older, not
CF4.
Put two cards in a RAID0 for OS.
Put third card as pagefile, temp files for OS, browser temp
files, application scratch space if needed.
By doing this you have cut in half the writes to the OS
volume raided cards, and the 3rd card with highest writes
can just be swapped out in a few years when it wears out.
The SLC flash chip aspect is important not just for the
faster writes but possibly 10X higher write cycles.
If you only needed it to run for a couple years you might be
able to use almost anything, keeping in mind that since
flash cards do have wear leveling, the more free space you
have the much longer they will last as total data volume
becomes a smaller ratio compared to available space.
>
>but there are companies making what some may market as IDE Flash
>Modules . .(e.g. made by Emphase, Innodisk, Transcend.)
Aren't these even more expensive? Perhaps I'm not clear on
what you meant above about "8GB_ solid state flash drives
are a bit expensive", at first I had assumed you meant an
8GB CF card but now I think you mean an 8GB SSD instead.
>
>They have write endurance of 1 to 4 million cycles, and have a
>controller utilising a wear levelling algorithm.
Their endurance probably isn't much if any better than a CF
card using SLC chips... there are only so many chips out
there they might use.
>
>How long will they last running windows xp?
WinXP is only an OS, we still dont' know the use of the
system and even if you said "browse internet" it can vary a
lot from one person to the next how much writing that is.
You can also set up a ramdrive, there are some out there
that support at least 1GB (I dont' recall the max but I know
1GB is possible because I have 1GB set up with one). IIRC
the one I used was mentioned and d/l linked here:
If you were setting it up in a worst case scenario with only
one card, no ramdrive, not disabling pagefile, etc - I dont'
think anyone can really tell you because everyone seems to
think about their optimizations when they do it. I don't
remember finding anyone who has actually worn theirs out
yet, so if you were to use all possible optimizations
including 3 cards w/SLC chips of good excess capacity, a
ramdrive, and have a vaguely typical XP use, in theory you
could get a minimum of a few dozen years out of this config.
>
>Would I -have- to disable virtual memory / set to no page file. I have
>heard that virtual memory wears these things out.. But I wonder if
>with such high endurance.
I'd consider it more a question of whether you can disable
virtual memory, whether there is enough main system memory
available including consideration of whether your particular
apps reserve a lot which can vary quite a bit. On one
system I'd disabled virtual memory on, it ran fine like that
for whatever I wanted to do, until I kept getting a
mysterious crash in gaming. Finally after fiddling with
drivers and checking overheating/etc I finally turned
virtual memory back on and it clearly resolved the problem.
Also, how much you actually use virtual memory (regardless
of how much of it is allocated, since allocation <> actually
writing that much data to it) makes quite a difference. You
might have 500MB used as shown in Task Manager but
practically none of that written with data, or you might
have quite a bit written if the system doesn't have much
physical memory. In summary, whether you have pagefile
enabled or not matters a lot less than how much physical
memory you have vs the jobs ran.
I would do as mentioned above, leave virtual memory enabled
but set to use a different CF card than the one upon which
OS is installed.
>
>I have read that capacity works for it.. A larger capacity drive will
>last longer. More memory cells.
>So I wonder if a 2GB drive won't last so long..
Possibly. Consider the scenario where you might have 1.5GB
of space taken by the OS (though with XP, that would be a
pretty lean installation) and a 2GB card. Ignoring loss due
to formatting and definitions of a GB varying between binary
and decimal, let's say you have 512MB left empty. Now let's
say you wrote 1GB/day to this drive, it means you wrote
every area twice if the wear leveling was perfect.
Now contrast that with a 4GB card, which would have approx
2560MB free. Writing 1GB/day to this one you have written
to each about 0.4 times. 2/0.4 = 5, the 4GB card would last
5 times as long.
>
>BTW, they don't come with SMART data that gives how many more writes
>they have / how the disk is.. do they ?
>(though I have read that they are more likely to fail due to
>controller error - picking a block that is already quite worn, wearing
>it out completely and making the disk unusable or unreliable. )
If an area is unusable the OS should mark it as bad if the
controller didn't itself. I don't recall the exact
capabilities of the controller and don't recall ever seeing
any way to get statistics, I don't think Smart is supported.
>
>It would be nice if the CF was 3.0, (even though more speed may mean
>it wears out quicker!).
No, it wears out much slower if more speed is due to being
SLC chipped or wears about same speed if not since you're
still writing same amount per use.
>But first and foremost, I want it to be
>reliable. And not priced through the roof. I don't want to be
>spending more than $80 / 40UKP. The big solid state drives, like 8GB,
>+, are quite expensive.
We don't know how much capacity you need, nor the prices
there. With that budget I would look for at least two 4GB CF
cards.
>
>I am interested in booting different operating systems e.g. FREEBSD..
>without worrying about this writing issue.. and how for each OS, to
>deal with it.
Kind of an open end issue, all you can do is look at where
it writes a lot and see if you can designate a different
location of not being able to turn it off.
You may be overly concerned about the write cycle
limitation, even if you used low end flash cards with the
least # of write cycles possible you can still get 16GB or
more of flash memory for $80... at least in the US they've
dropped to about $60 and up for the slower ones, but I feel
those might be a little too slow for many uses and would
suggest you get at least two 2-4GB cards with SLC chips and
CF4 spec. It might be a little over the budget but IMO the
best thing that comes close in the states would be a couple
of these: http://www.newegg.com/Product/Produc...82E16820208418
If you're sure you don't want to RAID a couple, you might
get one of each to come in around $80, putting OS & apps on
the faster one.
There's also a hack for XP where you can set up a write
filter from XPe so reduce or eliminate writes to the
HDD/CF-as-HDD. Google may find more info on that. Windows
For Legacy PCs seems to be a distant relative inbetween XP
and XPe (more like XPe but with most basic PC user features
enabled?) that might also allow turning off some features
that cause HDD writes, but I expect that if you use decent
SLC chipped flash cards of ample capacity that you won't
have it worn out in a few months of regular use, it should
be a few years at least, if the system isn't totally retired
by then.
I have got Acronis Disk Director booting of a CompactFlash device in an IDE
adaptor but it boots XP off a normal disk drive not directly off of the flash.
Works great.
> The 8GB+ solid state flash drives are a bit expensive for me..
Well, they're advertised on the UK eBay site starting at around £10 - if that's
too much then why bother at all ? I bought a pack of 3 CF to IDE adaptors for
£5 including postage from HK, arrived in 4 days.
> How long will they last running windows xp?
>
> Would I -have- to disable virtual memory / set to no page file. I have
> heard that virtual memory wears these things out.. But I wonder if
> with such high endurance.
I don't know what the endurance will actually be running XP directly off of one,
disabling the page file sounds like a good idea to me.
> BTW, they don't come with SMART data that gives how many more writes
> they have / how the disk is.. do they ?
No.
> It would be nice if the CF was 3.0, (even though more speed may mean
> it wears out quicker!). But first and foremost, I want it to be
> reliable. And not priced through the roof. I don't want to be
> spending more than $80 / 40UKP. The big solid state drives, like 8GB,
> +, are quite expensive.
Just buy a CF card and an adaptor, at least that way your expenditure is limited
even if the lifetime is short.
> I am interested in booting different operating systems e.g. FREEBSD..
> without worrying about this writing issue.. and how for each OS, to
> deal with it.
What's your particular interest in Fkash disks may I ask ?
Re: write endurance - solid state flash drives, windows xp, virtualmemory
On May 30, 10:40 pm, kony <s...@spam.com> wrote:
> On Thu, 29 May 2008 22:58:56 -0700 (PDT),
> "jameshanle...@yahoo.co.uk" <jameshanle...@yahoo.co.uk>
> wrote:
>
> >I am interested in booting windows xp off a solid state drive.
>
> >The 8GB+ solid stateflashdrives are a bit expensive for me..
>
> >I am looking at 4GB and less(2Gb,1GB)
> >..
> >The compactflashcards may not have such good endurance.. - they may
> >wear out.
>
> Is there a limit on how many CF cards you can use? You can
> get good performance and less concern for wear by using 3 of
> them, all three using SLCflashchips (normally would then
> be rated for over 10MB/s write speed) and CF4 spec (not just
> "CF spec" as some list this but it's CF2.x or older, not
> CF4.
<snip>
Your post was brilliant, full of great advice.. I have many comments,
but at the moment, my main is ,
Does a 133x card (20MB/s) imply CF standard of > 2.0 ?
I am a bit puzzled as to CF standards..
According to wikipedia,
wikipedia
according to the CompactFlash Association CFA
Revision 2.0 - supports up to 16 MB/s data transfer
Revision 3.0 supports up to a 66 MB/s data transfer rate, along with a
number of other features.
Revision 4.0 supports IDE Ultra DMA 133 for a maximum data transfer
rate of 133 MB/s
According to www.logicsupply.com - who have good info in compact flash cards.
Transcend industrial compact flash card
TS1GCF100I 100x 1GB - 2 million writes,
CF 4.1, 25-40MB/s, 100x, UDMA 0-4
(transcend have crap website, takes ages to load.. logicsupply is
better!) http://ec.transcendusa.com/product/I...mID=TS1GCF100I
what does 25MB-40MB actually mean? read? write? a range?
thing is, 100x is 15MB/s, so CF 2.0 would handle that. And 15MB/s is
alot less than 25-40MB/s
Even if it does 40MB/s, that is within CF 3.0
What advantage is in it being CF 4.1
can the speed ever imply anything about the compact flash
specification version number?
Re: write endurance - solid state flash drives, windows xp, virtual memory
On Sun, 1 Jun 2008 15:32:40 -0700 (PDT),
"jameshanley39@yahoo.co.uk" <jameshanley39@yahoo.co.uk>
wrote:
>On May 30, 10:40 pm, kony <s...@spam.com> wrote:
>> On Thu, 29 May 2008 22:58:56 -0700 (PDT),
>> "jameshanle...@yahoo.co.uk" <jameshanle...@yahoo.co.uk>
>> wrote:
>>
>> >I am interested in booting windows xp off a solid state drive.
>>
>> >The 8GB+ solid stateflashdrives are a bit expensive for me..
>>
>> >I am looking at 4GB and less(2Gb,1GB)
>> >..
>> >The compactflashcards may not have such good endurance.. - they may
>> >wear out.
>>
>> Is there a limit on how many CF cards you can use? You can
>> get good performance and less concern for wear by using 3 of
>> them, all three using SLCflashchips (normally would then
>> be rated for over 10MB/s write speed) and CF4 spec (not just
>> "CF spec" as some list this but it's CF2.x or older, not
>> CF4.
><snip>
>
>Your post was brilliant, full of great advice.. I have many comments,
>but at the moment, my main is ,
>
>Does a 133x card (20MB/s) imply CF standard of > 2.0 ?
No. It could be, it is possible they would mate a modern
CF3 or 4 controller with older flash chips to arrive at such
a product but it is less likely - most often they will use a
premium controller with premium flash chips which are faster
than 133X. You have no assurance of CF3 or 4 unless it
states that clearly (not just "CF spec" or similar but
actually includes the "3" or "4"), OR states UDMA support.
>
>I am a bit puzzled as to CF standards..
>
>According to wikipedia,
>wikipedia
>according to the CompactFlash Association CFA
>
>Revision 2.0 - supports up to 16 MB/s data transfer
>Revision 3.0 supports up to a 66 MB/s data transfer rate, along with a
>number of other features.
>Revision 4.0 supports IDE Ultra DMA 133 for a maximum data transfer
>rate of 133 MB/s
I don't see what is puzzling, these are upper limits, like
what interface a hard drive would use. What is also
important is the "UDMA" as it means very low CPU utilization
instead of very high using PIO mode. Also, cards adhering
to CF2 may not get anywhere near 16MB/s even if rated for
that, I have had quite a few that were capped at 6MB/s due
to the controller used when connected to CF-IDE adapter,
though these were faster using USB card reader IIRC.
>
>According to
>www.logicsupply.com - who have good info in compact flash cards.
>
>Transcend industrial compact flash card
>TS1GCF100I 100x 1GB - 2 million writes,
>CF 4.1, 25-40MB/s, 100x, UDMA 0-4
>(transcend have crap website, takes ages to load.. logicsupply is
>better!)
>http://ec.transcendusa.com/product/I...mID=TS1GCF100I
>
>what does 25MB-40MB actually mean? read? write? a range?
Since they don't clarify I can only speculate that it is a
peak read speed.
>
>thing is, 100x is 15MB/s, so CF 2.0 would handle that.
CF2 is not just about max speed, the important part is
whether it supports UDMA mode in your use on a CF-IDE
adapter and that requires CF3 or 4, unless your motherboard
and the OS don't mind running from a CF card in a card
reader. That combination is possible, I have card readers
that will allow booting a CF card from USB and others that
won't - on the same motherboard, though windows won't
install to one marked as removable instead of fixed disk,
IIRC, which is part of the advantage (maybe one of the only
real advantages of) having an industrial card.
>And 15MB/s is
>alot less than 25-40MB/s
>
>Even if it does 40MB/s, that is within CF 3.0
>What advantage is in it being CF 4.1
You have a point there, if two cards had the exact same
flash chips and their controllers were the same except for
the difference of one supporting CF3 and the other CF4, I
wold expect the same performance from both.
However, like all things flash chips tend to get faster over
time. It would tend to be that a modern CF4 controller is
faster and using newer faster flash chips even if both had
the same peak speed, since we know there is more to speed
than just a momentary peak but they can't very well rate
sustained speed except if it were an isolated test (and they
wouldn't want to anyway since it would be a signficantly
lower speed than the peak).
>
>can the speed ever imply anything about the compact flash
>specification version number?
I have not studied the specs enough to say definitively as a
rule, would assume any card not claiming CF3, 4, or UDMA
support did not support it. These are prominent,
value-adding features of a card so it would not make sense
for the manufacturer to leave out the detail that justifies
the higher price they command, at least in a spec sheet or
webpage if the seller didn't provide full specs.
The only safe bet is buying a card clearly spec'd at
manufacturer's site, unless you find a customer review that
you can place enough confidence in (sometimes hard to do on
the internet).
Re: write endurance - solid state flash drives, windows xp, virtualmemory
On May 30, 10:40 pm, kony <s...@spam.com> wrote:
> On Thu, 29 May 2008 22:58:56 -0700 (PDT),
> "jameshanle...@yahoo.co.uk" <jameshanle...@yahoo.co.uk>
> wrote:
>
<snip>
>
> Put two cards in a RAID0 for OS.
> Put third card as pagefile, temp files for OS, browser temp
> files, application scratch space if needed.
>
this is excellent advice, thanks
I notice some high end (DMI supporting) compact flash to IDE adaptors
have slots for master and slave. So I can put 3, even 4 in, if 2 IDE
connectors.
of course, USB slots are more freely available and more numerous.
Since I am not booting an OS for the 3rd card. Would a USB flash drive
be OK for the "third card" ?
I guess the RAID is expensive if RAIDing 2 fast cards.. so I might do
your other suggestion of 2 cards not RAIDed, but with OSnApps on the
faster one.
In this case, the partitioning, rather, I mean, separate modules,
would be for preservation of data..
Actually, if I was RAIDing, wouldn't I need 4 cards not 3 ?
2 for the OSnApps. 1 for scratch. One for Data.
Your suggestion missed out the data one.
BTW, out of interest, it sounds like you've done this before.. Whta
was your reason.. I know you don't have a problem with fan noise, so
that's not it.
<snip>
> >but there are companies making what some may market as IDEFlash
> >Modules . .(e.g. made by Emphase, Innodisk, Transcend.)
>
> Aren't these even more expensive? Perhaps I'm not clear on
> what you meant above about "8GB_ solid stateflashdrives
> are a bit expensive", at first I had assumed you meant an
> 8GB CF card but now I think you mean an 8GB SSD instead.
>
>
CF cards are fine..
You're right, the IDE Flash modules are more expensive.. a bit more.
I think maybe the marketting is such that you have smaller capacity
ones called IDE Flash Modules, that prob aren't 2.5" or 3.5". And
Bigger, capacity and size, are solid disk drives. All flash drives
though, and all solid state.
I was just looking at the IDE Flash modules because they tend to have
better write endurance. But the high end CF cards, sometimes marketted
as industrial, have write endurance to match them. like 1-4 million.
I'll look at CF cards now.
I notice there are SD cards too.. still Flash. Alot cheaper than CF. I
guess they prob don't do UDMA.
>
> >They have write endurance of 1 to 4 million cycles, and have a
> >controller utilising a wear levelling algorithm.
>
> Their endurance probably isn't much if any better than a CF
> card using SLC chips... there are only so many chips out
> there they might use.
>
>
>
> >How long will they last running windows xp?
>
> WinXP is only an OS, we still dont' know the use of the
> system and even if you said "browse internet" it can vary a
> lot from one person to the next how much writing that is.
>
> You can also set up a ramdrive, there are some out there
> that support at least 1GB (I dont' recall the max but I know
> 1GB is possible because I have 1GB set up with one). IIRC
> the one I used was mentioned and d/l linked here:
>
> http://www.mydigitallife.info/2007/0...or-windows-vis...
>
this is fantastic..
On a related note, I also found this..
Gigabyte iRAM - It's A physical RAM disk
PCI Card, takes RAM modules.. And it seems, actually stores the data..
(dunno how it does that, without losing it)
Somebody posted that he can boot windows xp in 5 seconds with it!!!
It wouldn't have the write endurance problem either.
A bit expensive, but not bad.. One of these cards, and alot of RAM
lying about, could go a long way.
They claim limited support, just Gigabyte motherboards that use one of
a bunch of certain supported southbridges..
I haven't spoken to anybody that has one,
But I have heard , from those that don't have one, that it should
work on any motherboard with a PCI slot.. Because it is essentially a
SATA drive, that just gets its 3.3V of power from PCI, and transfers
data through a connected SATA cable.
> If you were setting it up in a worst case scenario with only
> one card, no ramdrive, not disabling pagefile, etc - I dont'
> think anyone can really tell you because everyone seems to
> think about their optimizations when they do it. I don't
> remember finding anyone who has actually worn theirs out
> yet, so if you were to use all possible optimizations
> including 3 cards w/SLC chips of good excess capacity, a
> ramdrive, and have a vaguely typical XP use, in theory you
> could get a minimum of a few dozen years out of this config.
>
>
>
> >Would I -have- to disable virtual memory / set to no page file. I have
> >heard that virtual memory wears these things out.. But I wonder if
> >with such high endurance.
>
> I'd consider it more a question of whether you can disable
> virtual memory, whether there is enough main system memory
> available including consideration of whether your particular
> apps reserve a lot which can vary quite a bit. On one
> system I'd disabled virtual memory on, it ran fine like that
> for whatever I wanted to do, until I kept getting a
> mysterious crash in gaming. Finally after fiddling with
> drivers and checking overheating/etc I finally turned
> virtual memory back on and it clearly resolved the problem.
>
> Also, how much you actually use virtual memory (regardless
> of how much of it is allocated, since allocation <> actually
> writing that much data to it) makes quite a difference. You
> might have 500MB used as shown in Task Manager but
> practically none of that written with data, or you might
> have quite a bit written if the system doesn't have much
> physical memory. In summary, whether you have pagefile
> enabled or not matters a lot less than how much physical
> memory you have vs the jobs ran.
ok.. so for win xp. "PF Usage" in task manager..performance, would be
of practical relevance to that - if I wanted to monitor it a bit.
Besides the XPe option. And the windows xp for legacy pcs option you
mentioned.
You mentioned earlier about moving the location of the temporary
folder used by applications.. on a per application basis. I found the
program "Process Monitor"(a combination of those old sysinternals
programs, filemon and regmon). It intercepts all registry changes and
file writes. It's prob clearer though. I can filter by paths
containing C:\ or whatever drive, and thus see all the file writes.
Regarding moving the Windows "Temp" directory.. (I think a good idea
is to have %TMP% and %TEMP% pointing to te same place)..
I find IE writes not just to
c:\docu..\user\Local Settings\"Temporary Internet Files",
but within Local Settings, it also writes to LS\Cookies, and some,
maybe IE too, writes to "LS\Application Data".
It may be better to just move the whole Documents And Settings
directory..
I found one or two links on that. http://www.tech-recipes.com/windows_tips1409.html
Of course.. it seems with XPe at least, I won't have to work on an
application basis.. Any writes go to HDD gets intercepted and
redirected - it seems - and go to RAM.
>
> I would do as mentioned above, leave virtual memory enabled
> but set to use a different CF card than the one upon which
> OS is installed.
>
That would wear out fast..
But
I guess it's important that that should be CF3.0/4.0 So, UDMA too?
'cos Windows would be accessing that alot..
So I guess that doesn't save on price.. It just preserves the card
holding my data.
>
>
> >I have read that capacity works for it.. A larger capacity drive will
> >last longer. More memory cells.
> >So I wonder if a 2GB drive won't last so long..
>
> Possibly. Consider the scenario where you might have 1.5GB
> of space taken by the OS (though with XP, that would be a
> pretty lean installation) and a 2GB card. Ignoring loss due
> to formatting and definitions of a GB varying between binary
> and decimal, let's say you have 512MB left empty. Now let's
> say you wrote 1GB/day to this drive, it means you wrote
> every area twice if the wear leveling was perfect.
>
> Now contrast that with a 4GB card, which would have approx
> 2560MB free. Writing 1GB/day to this one you have written
> to each about 0.4 times. 2/0.4 = 5, the 4GB card would last
> 5 times as long.
>
>
>
> >BTW, they don't come with SMART data that gives how many more writes
> >they have / how the disk is.. do they ?
> >(though I have read that they are more likely to fail due to
> >controller error - picking a block that is already quite worn, wearing
> >it out completely and making the disk unusable or unreliable. )
>
> If an area is unusable the OS should mark it as bad if the
> controller didn't itself. I don't recall the exact
> capabilities of the controller and don't recall ever seeing
> any way to get statistics, I don't think Smart is supported.
>
>
>
> >It would be nice if the CF was 3.0, (even though more speed may mean
> >it wears out quicker!).
>
> No, it wears out much slower if more speed is due to being
> SLC chipped or wears about same speed if not since you're
> still writing same amount per use.
>
> >But first and foremost, I want it to be
> >reliable. And not priced through the roof. I don't want to be
> >spending more than $80 / 40UKP. The big solid state drives, like 8GB,
> >+, are quite expensive.
>
It seems that the ones with good endurance, are transcend, emphase,
innodisk..
where they advertise the endurance.. and ideally, the version of spec.
sandisk - prob because of their end user market- do not advertise it.
So even newegg do not list it in their specification. But I notice the
prices of their CF cards (not their SD cards).. Are the same as the
big names..
Approximately $12.5 per GB. $25 for 2GB. 50 for 4GB ..
> I
> We don't know how much capacity you need, nor the prices
> there. With that budget I would look for at least two 4GB CF
> cards.
>
That would be fine.
>
>
> >I am interested in booting different operating systems e.g. FREEBSD..
> >without worrying about this writing issue.. and how for each OS, to
> >deal with it.
>
> Kind of an open end issue, all you can do is look at where
> it writes a lot and see if you can designate a different
> location of not being able to turn it off.
>
> You may be overly concerned about the write cycle
> limitation, even if you used low endflashcards with the
> least # of write cycles possible you can still get 16GB or
> more offlashmemory for $80... at least in the US they've
> dropped to about $60 and up for the slower ones, but I feel
> those might be a little too slow for many uses and would
> suggest you get at least two 2-4GB cards with SLC chips and
> CF4 spec. It might be a little over the budget but IMO the
> best thing that comes close in the states would be a couple
> of these:http://www.newegg.com/Product/Produc...82E16820208418
>
> For a little less but half the capacity and a speed
> reduction,http://www.newegg.com/Product/Produc...82E16820208296
>
nice.. I notice they are transcend also.. 266x and 300x. 2GB and 4GB
amazon sell some cheap transcend. I notice they are SD - secure
digital though. Not CF.
And no doubt no UDMA.
TS4GSDHC6 4GB
10UKP inc PnP (like $20)
TS2GSDC 2GB
5UKP inc PnP (like $10)
> If you're sure you don't want to RAID a couple, you might
> get one of each to come in around $80, putting OS & apps on
> the faster one.
>
> There's also a hack for XP where you can set up a write
> filter from XPe so reduce or eliminate writes to the
> HDD/CF-as-HDD. Google may find more info on that. Windows
> For Legacy PCs seems to be a distant relative inbetween XP
> and XPe (more like XPe but with most basic PC user features
> enabled?) that might also allow turning off some features
> that cause HDD writes, but I expect that if you use decent
> SLC chippedflashcards of ample capacity that you won't
> have it worn out in a few months of regular use, it should
> be a few years at least, if the system isn't totally retired
> by then.
yeah.. I guess if using flash, dealing with this on an OS by OS basis,
I may as well.
are there laptops shipped with SSD drives using windows xp with
virtual memory and eating the drives? The average end user would not
have done any of these optimisations to protect the drive.
Re: write endurance - solid state flash drives, windows xp, virtual memory
On Mon, 02 Jun 2008 07:11:30 -0500, Joel <Joel@NoSpam.com>
wrote:
>"jameshanley39@yahoo.co.uk" <jameshanley39@yahoo.co.uk> wrote:
>
><snip>
>> Your post was brilliant, full of great advice.. I have many comments,
>> but at the moment, my main is ,
>>
>> Does a 133x card (20MB/s) imply CF standard of > 2.0 ?
>>
>> I am a bit puzzled as to CF standards..
>>
>> According to wikipedia,
>> wikipedia
>> according to the CompactFlash Association CFA
>>
>> Revision 2.0 - supports up to 16 MB/s data transfer
>> Revision 3.0 supports up to a 66 MB/s data transfer rate, along with a
>> number of other features.
>> Revision 4.0 supports IDE Ultra DMA 133 for a maximum data transfer
>> rate of 133 MB/s
>
> Sometime we just learn to make thing as simplest as we can understand to
>be able to understand and to make life a little easier to live... unless we
>really need detailed technical information to write a book or something then
>that's another story.
>
> And these are my guessing .. cuz I never care to know more to confuse me.
>
>133X = 133 times faster than whatever the original speed they use to compare
>to. Same with the speed of CD/DVD writer, the CPU speed etc.. my first CD
>burner was 1X and now I believe it's around 52X or so, and it seems like the
>4X DVD is faster than 4X CD-R
The speed they compare to is a single speed CD rate of
150KB/s, keeping in mind that the speed is only seen with
certain kinds of large transfers in an otherwise ideal
environment. IOW, there may be other bottlenecks as well as
the card before achieving that speed.
>
>Revision - I just think of it as firmware, generation, version of the
>hardware etc..
>
> Me, I don't care about the thing I need not to know but the basic thing
>like "size", "speed", "price" and read the feedback's to see how it compares
>to others carrying the higher price tag.
Revision is very important for the intended use as a
replacement for a hard drive on a Compact Flash - IDE
adapter because it determines whether the card will operate
in UDMA mode. Some cards rated for 150X (which is about
22MB/s) can't even exceed 6MB/s when connected to a CF-IDE
adapter in PIO mode, and use quite a lot of CPU time to do
it while UDMA mode doesn't. It makes a night and day
difference in the performance, changing an extremely
sluggish system into one that's fairly snappy so long as the
file transfers aren't very large or huge numbers of tiny
files.
Re: write endurance - solid state flash drives, windows xp, virtual memory
On Mon, 2 Jun 2008 19:37:32 -0700 (PDT),
"jameshanley39@yahoo.co.uk" <jameshanley39@yahoo.co.uk>
wrote:
>On May 30, 10:40 pm, kony <s...@spam.com> wrote:
>> On Thu, 29 May 2008 22:58:56 -0700 (PDT),
>> "jameshanle...@yahoo.co.uk" <jameshanle...@yahoo.co.uk>
>> wrote:
>>
><snip>
>
>>
>> Put two cards in a RAID0 for OS.
>> Put third card as pagefile, temp files for OS, browser temp
>> files, application scratch space if needed.
>>
>
>this is excellent advice, thanks
>
>I notice some high end (DMI supporting) compact flash to IDE adaptors
>have slots for master and slave. So I can put 3, even 4 in, if 2 IDE
>connectors.
Yes, it is more space, and often cost, effective.
>
>of course, USB slots are more freely available and more numerous.
>Since I am not booting an OS for the 3rd card. Would a USB flash drive
>be OK for the "third card" ?
I don't recall if windows(?) allows using removable media
for pagefile, temp files, etc. which was the use I had
suggested for the 3rd card. If it does allow placement on
removable media, that would be an acceptible option except
that with a fast card it will be slower over USB2 than
running in UDMA mode on a PATA controller.
>
>I guess the RAID is expensive if RAIDing 2 fast cards.. so I might do
>your other suggestion of 2 cards not RAIDed, but with OSnApps on the
>faster one.
It may be a bit more expensive because you have to buy the
raid controller card if you dont' have one (about $15 in the
US) but otherwise no so expensive. Remember that if you
RAID0 two cards, you can (for example) use 2 x 2GB cards to
achieve 4GB, instead of one 4GB card to achieve 4GB. Since
to a certain extent a smaller card may only cost half as
much the cost comes out pretty close, but definitely quite a
bit faster raided than not.
>
>In this case, the partitioning, rather, I mean, separate modules,
>would be for preservation of data..
>
>Actually, if I was RAIDing, wouldn't I need 4 cards not 3 ?
>2 for the OSnApps. 1 for scratch. One for Data.
>Your suggestion missed out the data one.
I suggested RAID1 of two cards for OS and apps, which have
the most I/O, and a single card in non-raid mode for the
data which is usually accessed less often (but of course it
would depend on the use of the system, you wouldn't want
that if you were continually working with large pictures in
photoshop or video editing for example, but for practical
purposes there isn't enough capacity for video editing.
Lastly, the suggestion of only 3 cards was to stay nearer
the proposed $80 budget. Certainly a 4th card to allow 2 x
RAID0 arrays would be faster for the data, and might be
reasonable still given what I wrote above, that you can get
cards half the size for about half the price so it really
isn't much if any more expensive if you're using a raid
controller card already and have the free slot for the 4th
card.
I'd love to see a low profile PCI and PCIe versions of a
RAID card with 4 CF slots on it. Really I mean a cheap one,
there may be some out there somewhere but they're likely to
be quite high priced - I'm wanting one like the $15
controller but add maybe $1/CF-slot so it's a $19 card.
>
>BTW, out of interest, it sounds like you've done this before.. Whta
>was your reason.. I know you don't have a problem with fan noise, so
>that's not it.
I have a lot of systems, most of which don't need everything
but the kitchen sink installed. It is nice to cut HDD
noise, power, have it more reliable when these systems are
mostly sitting around idle. Setting them to spin down a
mechanical drive introduces lag every time the drive spins
up again. Plus, some of these are older systems and if the
drive ages to the point of failure or becoming a liability
(expected to fail some month or year soon) it can be cheaper
to just buy a 2-4GB CF card for it than a new HDD. Often
these are the cheap slow cards, but I've also put together a
very low power browsing and email station with a Via
processor - which is sluggish because of the low
performance of the processor but nice that it uses so little
power and is so low maintenance with minimal dust buildup
from one very low RPM fan - some people like to build these
with only passive cooling but it gets a bit warm and I'd
like it to last longer than I'd expect it to passively
cooled.
Plus, having a CF card that boots dos with some utilities
and Windows installation files on it is very handy for
plopping into a system to run various things or install
windows, quicker than having to mount a 2nd hard drive or
dig out and suffer latency of an optical drive, and
especially on old systems I can't assume they'd be able to
boot USB even if they're supposedly able to... then with the
really old systems they can't support larger hard drives so
a CF card within the capacity of the drive sizes they can
support seems a perfect match.
>
><snip>
>> >but there are companies making what some may market as IDEFlash
>> >Modules . .(e.g. made by Emphase, Innodisk, Transcend.)
>>
>> Aren't these even more expensive? Perhaps I'm not clear on
>> what you meant above about "8GB_ solid stateflashdrives
>> are a bit expensive", at first I had assumed you meant an
>> 8GB CF card but now I think you mean an 8GB SSD instead.
>>
>>
>
>CF cards are fine..
>
>You're right, the IDE Flash modules are more expensive.. a bit more.
>
>http://linitx.com/viewcategory.php?catid=126&pp=126
>
It's a shame they're still priced so high only due to being
a niche product, the extra traces and plastic connector
shouldn't add but a couple dollars to the total price.
>I think maybe the marketting is such that you have smaller capacity
>ones called IDE Flash Modules, that prob aren't 2.5" or 3.5". And
>Bigger, capacity and size, are solid disk drives. All flash drives
>though, and all solid state.
yes different market, the flash modules are generally
intended for SFF system or industrial uses while the SSDs
are trying to break into laptop and desktop PC territory.
>
>I was just looking at the IDE Flash modules because they tend to have
>better write endurance. But the high end CF cards, sometimes marketted
>as industrial, have write endurance to match them. like 1-4 million.
Any CF card using SLC chips and wear leveling (practically
all) will have about the same write endurance. It's mainly
the ones with MLC chips that you want to avoid regardless of
whether it is a CF card, flash module, or SSD. It is a bit
disgusting to me that some SSDs are using MLC chips but
rated like they are, it's as if they were intentionally
vague or misleading citing only the spec for their lower
capacity part(s) that used the SLC chips instead of MLC.
>
>I'll look at CF cards now.
>
>I notice there are SD cards too.. still Flash. Alot cheaper than CF. I
>guess they prob don't do UDMA.
Correct, they are nowhere near as fast as a higher end CF
card, and you can't get them in the highest capacities with
SLC chips because the chip density is just too high to get #
of chips needed into such a small space. I have a
multi-in-one USB card reader that I can and have booted to a
SD card before, but mostly just for utilities or bios
updates... think I did actually install Ubuntu onto one and
it was very slow performing, though it wasn't a performance
SD card either, maybe 150X rated.
>> You can also set up a ramdrive, there are some out there
>> that support at least 1GB (I dont' recall the max but I know
>> 1GB is possible because I have 1GB set up with one). IIRC
>> the one I used was mentioned and d/l linked here:
>>
>> http://www.mydigitallife.info/2007/0...or-windows-vis...
>>
>
>this is fantastic..
>
>On a related note, I also found this..
>Gigabyte iRAM - It's A physical RAM disk
>PCI Card, takes RAM modules.. And it seems, actually stores the data..
>(dunno how it does that, without losing it)
>
>Somebody posted that he can boot windows xp in 5 seconds with it!!!
yeah but that's fairly expensive isn't it? IIRC those were
about $130 last time I checked, or maybe even more as I
can't recall for certain. I don't care much how long it
takes windows to boot cold, I'd just hibernate if it
mattered and with today's large and cheap DDR2 memory
modules it doesn't at all seem necessary to me to have a
dedicated battery backed module instead of just using a
ramdrive, except of course that the OS would run quite a bit
faster from it until the reads from the CF module were
cached into main memory - not the ramdrive just the windows
filecaching.
>
>It wouldn't have the write endurance problem either.
>A bit expensive, but not bad.. One of these cards, and alot of RAM
>lying about, could go a long way.
True but I also dislike the size of it, I figure if going
for low power and quiet then a system could be quite small
and thin which makes the ramdrive cards look pretty wide
instead of low profile. Plus, I don't like buying things
with batteries that would tend to fail or at least lose a
lot of capacity within 5 years as I tend to build systems I
intend to last for at least 10, even if not in the original
use or by the original owner all that time.
>
>They claim limited support, just Gigabyte motherboards that use one of
>a bunch of certain supported southbridges..
>I haven't spoken to anybody that has one,
>But I have heard , from those that don't have one, that it should
>work on any motherboard with a PCI slot.. Because it is essentially a
>SATA drive, that just gets its 3.3V of power from PCI, and transfers
>data through a connected SATA cable.
That's essentially what I've heard too but come to think of
it, don't they use DDR(1) memory instead of DDR2? Seems a
shame since it costs so much more.
>> Also, how much you actually use virtual memory (regardless
>> of how much of it is allocated, since allocation <> actually
>> writing that much data to it) makes quite a difference. You
>> might have 500MB used as shown in Task Manager but
>> practically none of that written with data, or you might
>> have quite a bit written if the system doesn't have much
>> physical memory. In summary, whether you have pagefile
>> enabled or not matters a lot less than how much physical
>> memory you have vs the jobs ran.
>
>ok.. so for win xp. "PF Usage" in task manager..performance, would be
>of practical relevance to that - if I wanted to monitor it a bit.
No that isn't telling you how much data is written to the
pagefile. For example right now mine says 1.4GB but it's
not paging out any data, just marking reserved space if the
system were to ever run out of main memory which isn't too
likely at the moment as same system has almost 2GB free and
that's not even counting a 1GB ramdrive.
>
>Besides the XPe option. And the windows xp for legacy pcs option you
>mentioned.
>
>You mentioned earlier about moving the location of the temporary
>folder used by applications.. on a per application basis.
yes, for each application there may be settings in it's
menus to pick where the temp files go. In addition, set
windows global "temp" variable to point to the ramdrive as
well, if not the 3rd CF card as mentioned previously.
> I found the
>program "Process Monitor"(a combination of those old sysinternals
>programs, filemon and regmon). It intercepts all registry changes and
>file writes. It's prob clearer though. I can filter by paths
>containing C:\ or whatever drive, and thus see all the file writes.
>Regarding moving the Windows "Temp" directory.. (I think a good idea
>is to have %TMP% and %TEMP% pointing to te same place)..
yes, and all user's temp directories to the same place or at
least the ramdrive too, although in theory it is possible
this can cause a problem, sometimes things installed want
you to reboot and could potentially then finish up their
installation from the files they'd put in the temp
directory, files that were then lost with a reboot. In such
cases you have to temporarily reassign the temp folder to a
non-volatile place like the CF card.
>I find IE writes not just to
>c:\docu..\user\Local Settings\"Temporary Internet Files",
>but within Local Settings, it also writes to LS\Cookies, and some,
>maybe IE too, writes to "LS\Application Data".
>It may be better to just move the whole Documents And Settings
>directory..
>I found one or two links on that.
>http://www.tech-recipes.com/windows_tips1409.html
Yes there are lots of little things windows incessantly
writes, but it may not be necessary to divert all of them
and with the example of cookies, personally I would rather
have some of those persistant instead of losing them.
>
>
>Of course.. it seems with XPe at least, I won't have to work on an
>application basis.. Any writes go to HDD gets intercepted and
>redirected - it seems - and go to RAM.
>
>It might also be possible to get a write filter like the one on XPe,
>onto Win XP Pro. I think I saw it mentioned on this forum
>http://www.mp3car.com/vbulletin/mp3car-technical/
Yes, that was what I was suggesting, to use the XPe filter
on XP home or pro, not using XPe. I don't find XPe
advantageous for a *PC* with typical desktop uses, might as
well just install plain XP or XP FLP (for legacy PCs).
>> I would do as mentioned above, leave virtual memory enabled
>> but set to use a different CF card than the one upon which
>> OS is installed.
>>
>
>That would wear out fast..
Based on what? Suppose it has 100,000 writes a day that are
small, like marking virtual memory reserved, not actually
paging out data which is the scenario if you have ample
system memory. Suppose those 10,000 writes average 100K
each and you have 1GB of free space.
That's 10,000 * 100K = 9765MB, filling the 1GB of space 9.5
times every day.
Now even with a cheap MLC card that has a mere 100,000 write
cycle endurance, that's almost 29 years till it wore out.
>But
>I guess it's important that that should be CF3.0/4.0 So, UDMA too?
>'cos Windows would be accessing that alot..
Yes, for a system a human is actively using like a regular
desktop that would be pretty important to maintain a
reasonable level of performance.
>> >But first and foremost, I want it to be
>> >reliable. And not priced through the roof. I don't want to be
>> >spending more than $80 / 40UKP. The big solid state drives, like 8GB,
>> >+, are quite expensive.
>>
>
>
>It seems that the ones with good endurance, are transcend, emphase,
>innodisk..
>where they advertise the endurance.. and ideally, the version of spec.
Any decent modern card has wear leveling so the main issue
is whether it is SLC chipped. All SLC chipped have roughly
10X the endurance, but see the calculation I did above, that
it might take many years even with an MLC chipped card and
it then leads to the conclusion that the improved write
performance is more important.
>
>sandisk - prob because of their end user market- do not advertise it.
>So even newegg do not list it in their specification. But I notice the
>prices of their CF cards (not their SD cards).. Are the same as the
>big names..
? Sandisk is one of the big names
Often their prices are slightly higher for the same
performance, I'd never buy an unrated Sandisk, nor Kingston
product for that matter unless performance didn't matter at
all for the application.
>nice.. I notice they are transcend also.. 266x and 300x. 2GB and 4GB
The 300X being dual channel, "Might" have the potential to
be faster than the numbers suggest, I mean it migh retain
more of it's performance in some situations like writing
with smaller files which cases a slowdown with other cards.
It is only speculation since I don't have one of those.
>are there laptops shipped with SSD drives using windows xp with
>virtual memory and eating the drives? The average end user would not
>have done any of these optimisations to protect the drive.
You might visit the Dell forms to see if they haven't
deleted any threads about it. There was a controversy I
read about recently that said something like 20% of Dells
SSD based laptops were failing or returned - or something,
but Dell denied it. I wish I could remember where I read
that.
The thing is, even if there were a lot of the laptops
failing, it doesn't necessarily doom other SSDs and doesn't
tell us anything about the exact failure mode which wouldn't
have to be a write cycle limitation.
Re: write endurance - solid state flash drives, windows xp, virtualmemory
On 4 Jun, 00:32, kony <s...@spam.com> wrote:
<snip>
> Revision is very important for the intended use as a
> replacement for *a hard drive on a Compact Flash - IDE
> adapter because it determines whether the card will operate
> in UDMA mode. *Some cards rated for 150X (which is about
> 22MB/s) can't even exceed 6MB/s when connected to a CF-IDE
> adapter in PIO mode, and use quite a lot of CPU time to do
> it while UDMA mode doesn't. *It makes a night and day
> difference in the performance, changing an extremely
> sluggish system into one that's fairly snappy so long as the
> file transfers aren't very large or huge numbers of tiny
> files.
I see how PIO is bad because of its CPU utilization. But in terms of
speed.. UDMA is generally faster. But there are fast PIO and slow
UDMA..
Howcome you keep referring to PIO as 6MB/s. Why wouldn't PIO cards go
faster, like 16.7MB/s. Is it because you are saying the cheap cards
are PIO mode 1.
I suppose ATA(ATA/IDE connector) doesn't support Modes 5,6. Since that
is specific to CF. But it would support Mode 4 PIO I would think.
16.7MB/s
Another oddity is that sandisk Extreme IV cards say UDMA mode 4. But
have a max R/W of up to 40MB/s (266x), I guess they adverise as 266x,
so it's a max. The oddity though, is they may as well be on UDMA mode
3. !!
And if that's the max speed, I wonder what the minimum be like
Re: write endurance - solid state flash drives, windows xp, virtual memory
On Wed, 4 Jun 2008 17:38:36 -0700 (PDT),
"jameshanley39@yahoo.co.uk" <jameshanley39@yahoo.co.uk>
wrote:
>On 4 Jun, 00:32, kony <s...@spam.com> wrote:
><snip>
>> Revision is very important for the intended use as a
>> replacement for *a hard drive on a Compact Flash - IDE
>> adapter because it determines whether the card will operate
>> in UDMA mode. *Some cards rated for 150X (which is about
>> 22MB/s) can't even exceed 6MB/s when connected to a CF-IDE
>> adapter in PIO mode, and use quite a lot of CPU time to do
>> it while UDMA mode doesn't. *It makes a night and day
>> difference in the performance, changing an extremely
>> sluggish system into one that's fairly snappy so long as the
>> file transfers aren't very large or huge numbers of tiny
>> files.
>
>I see how PIO is bad because of its CPU utilization. But in terms of
>speed.. UDMA is generally faster. But there are fast PIO and slow
>UDMA..
There's no UDMA (0) slower than the fastest PIO. Mainly the
point is, AFAIK nobody would pair up the low end cheap flash
chips that are slow, with the premium UDMA capable
controller, though they'll be more likely to pair the
fastest chips up with the premium controller. This also
tends to be true because of customer demand, higher
performance more sophisticated equipment would be the most
likely to use UDMA mode instead of a camera, though I
suppose these days that generalization is changing, a lot of
purpose built ICs do far more than they used to.
Generally we can ignore all these issues and just remember
one thing - you absolutely must have a card with CF3 or 4
spec operating in UDMA mode if you want it more than
sluggish.
Keep in mind, any modern flash chips can achieve over 27MB/s
read speed, including the low-end, if only they have the
right controller and for your purposes that means it can't
operate in PIO mode even if what I was about to write next
wasn't true.
>Howcome you keep referring to PIO as 6MB/s. Why wouldn't PIO cards go
>faster, like 16.7MB/s. Is it because you are saying the cheap cards
>are PIO mode 1.
I'm saying that I didn't care enough to find out why they
only achieved 6MB/s, apparently they only supported a lower
PIO mode - on several different motherboards so it would
tend to implicate the cards. These were mostly 512MB to 2GB
cards I had leftover from an old camera.
>I suppose ATA(ATA/IDE connector) doesn't support Modes 5,6. Since that
>is specific to CF. But it would support Mode 4 PIO I would think.
>16.7MB/s
I don't understand why you are spending time pondering the
very cards you don't want to use if performance is
important. It wouldn't be surprising, in the low end
cards, for the manufacturer to only support the bare minimum
it takes to qualify as a "Compact Flash" card, particularly
since many cameras aren't using them in PIO mode. That
could be seen as a good thing, as today you can get a 2GB
low end card almost for nothing, $15 or less generally, but
it might be the equivalent of comparing a zip drive to a
hard drive when it comes to performance running a large(r)
OS from it on an ATA controller. Even Win98 using Win98
shell a la 98Lite, an installation taking under 200MB drive
space and about a couple dozen MB memory, is a bit sluggish
running from a low end CF card in PIO mode.
>
>Another oddity is that sandisk Extreme IV cards say UDMA mode 4. But
>have a max R/W of up to 40MB/s (266x), I guess they adverise as 266x,
>so it's a max. The oddity though, is they may as well be on UDMA mode
>3. !!
This is the same as with any other drive, you want the
interface bandwidth to be as high as possible - modern spec,
no matter whether the actual performance could've been done
with some slower interface. It wold serve no useful purpose
for it to be CF3 instead, CF4 is the modern spec and thus
any modern card pretending to be high performance should
adhere to that spec.
I'm going to back up and say again what you seemed to miss.
Don't think about these things. Think about only:
CF4 spec
SLC chips
Max read speed the budget can afford
(in this order of importance)
>And if that's the max speed, I wonder what the minimum be like
>
>ref-
>
>PIO
>Mode 6 - 25MB/s <--- CF2
>Mode 5 - 20MB/s <-- CF2
>Mode 4 - 16.7MB/s <-ATA
>Mode 3 – 11.1MB/s <-ATA
>Mode 2 – 8.3MB/s <-ATA
>Mode 1 – 5.2MB/s <-ATA
>Mode 0 - 3.3MB/s <-- ATA
>
>UDMA
>Mode 5 - 100MB/s <-- ATA/ATAPI
>Mode 4 - 66.7 MB/s <-- ATA/ATAPI
>Mode 3 – 44.4MB/s <-- ATA/ATAPI
>Mode 2 – 33.3MB/s <-- ATA/ATAPI
>Mode 1 - 25MB/s <-- ATA/ATAPI
>Mode 0 - 16.7MB/s <-- ATA/ATAPI
>
>http://en.wikipedia.org/wiki/Programmed_input/output
>(better than pcguide PIO link. which is missing mode 0 and modes 5,6
>
>http://www.pcguide.com/ref/hdd/if/ide/modes_UDMA.htm
>pcguide UDMA link.
You have no guarantee about any mode any card uses unless it
clearly specifies it. However, what we can assume is that
something that makes a CF card more valuable is something
the manufacturer will not overlook mentioning, ie - if it is
CF3 or 4 spec. and the inherant support those mandate.
Beyond that, your choices are really quite limited since you
want not to spend a lot, with the higher cost being the
primary thing keeping many from moving to flash based
storage.
Re: write endurance - solid state flash drives, windows xp, virtualmemory
On Jun 5, 4:24*am, kony <s...@spam.com> wrote:
> On Wed, 4 Jun 2008 17:38:36 -0700 (PDT),
> "jameshanle...@yahoo.co.uk" <jameshanle...@yahoo.co.uk>
> wrote:
>
>
>
>
>
> >On 4 Jun, 00:32, kony <s...@spam.com> wrote:
> ><snip>
> >> Revision is very important for the intended use as a
> >> replacement for *a hard drive on a Compact Flash - IDE
> >> adapter because it determines whether the card will operate
> >> in UDMA mode. *Some cards rated for 150X (which is about
> >> 22MB/s) can't even exceed 6MB/s when connected to a CF-IDE
> >> adapter in PIO mode, and use quite a lot of CPU time to do
> >> it while UDMA mode doesn't. *It makes a night and day
> >> difference in the performance, changing an extremely
> >> sluggish system into one that's fairly snappy so long as the
> >> file transfers aren't very large or huge numbers of tiny
> >> files.
>
> >I see how PIO is bad because of its CPU utilization. But in terms of
> >speed.. UDMA is generally faster. But there are fast PIO and slow
> >UDMA..
>
> There's no UDMA (0) slower than the fastest PIO. *Mainly the
> point is, AFAIK nobody would pair up the low end cheap flash
> chips that are slow, with the premium UDMA capable
> controller, though they'll be more likely to pair the
> fastest chips up with the premium controller. *This also
> tends to be true because of customer demand, higher
> performance more sophisticated equipment would be the most
> likely to use UDMA mode instead of a camera, though I
> suppose these days that generalization is changing, a lot of
> purpose built ICs do far more than they used to.
>
> Generally we can ignore all these issues and just remember
> one thing - you absolutely must have a card with CF3 or 4
> spec operating in UDMA mode if you want it more than
> sluggish. *
>
> Keep in mind, any modern flash chips can achieve over 27MB/s
> read speed, including the low-end, if only they have the
> right controller and for your purposes that means it can't
> operate in PIO mode even if what I was about to write next
> wasn't true.
>
> >Howcome you keep referring to PIO as 6MB/s. Why wouldn't PIO cards go
> >faster, like 16.7MB/s. *Is it because you are saying the cheap cards
> >are PIO mode 1.
>
> I'm saying that I didn't care *enough to find out why they
> only achieved 6MB/s, apparently they only supported a lower
> PIO mode - on several different motherboards so it would
> tend to implicate the cards. *These were mostly 512MB to 2GB
> cards I had leftover from an old camera.
>
> >I suppose ATA(ATA/IDE connector) doesn't support Modes 5,6. Since that
> >is specific to CF. But it would support Mode 4 PIO I would think.
> >16.7MB/s
>
> I don't understand why you are spending time pondering the
> very cards you don't want to use if performance is
> important. * *It wouldn't be surprising, in the low end
> cards, for the manufacturer to only support the bare minimum
> it takes to qualify as a "Compact Flash" card, particularly
> since many cameras aren't using them in PIO mode. *That
> could be seen as a good thing, as today you can get a 2GB
> low end card almost for nothing, $15 or less generally, but
> it might be the equivalent of comparing a zip drive to a
> hard drive when it comes to performance running a large(r)
> OS from it on an ATA controller. *Even Win98 using Win98
> shell a la 98Lite, an installation taking under 200MB drive
> space and about a couple dozen MB memory, is a bit sluggish
> running from a low end CF card in PIO mode.
>
>
>
> >Another oddity is that sandisk Extreme IV cards say UDMA mode 4. But
> >have a max R/W of up to 40MB/s (266x), I guess they adverise as 266x,
> >so it's a max. The oddity though, is they may as well be on UDMA mode
> >3. !!
>
> This is the same as with any other drive, you want the
> interface bandwidth to be as high as possible - modern spec,
> no matter whether the actual performance could've been done
> with some slower interface. *It wold serve no useful purpose
> for it to be CF3 instead, CF4 is the modern spec and thus
> any modern card pretending to be high performance should
> adhere to that spec.
>
> I'm going to back up and say again what you seemed to miss.
> Don't think about these things. *Think about only:
>
> CF4 spec
> SLC chips
> Max read speed the budget can afford
> (in this order of importance)
>
>
>
>
>
> >And if that's the max speed, I wonder what the minimum *be like
>
> >ref-
>
> >PIO
> >Mode 6 - 25MB/s <--- CF2
> >Mode 5 - 20MB/s <-- CF2
> >Mode 4 - 16.7MB/s <-ATA
> >Mode 3 – 11.1MB/s <-ATA
> >Mode 2 – 8.3MB/s <-ATA
> >Mode 1 – 5.2MB/s <-ATA
> >Mode 0 - 3.3MB/s <-- ATA
>
> >UDMA
> >Mode 5 - 100MB/s <-- ATA/ATAPI
> >Mode 4 - 66.7 MB/s <-- ATA/ATAPI
> >Mode 3 – 44.4MB/s <-- ATA/ATAPI
> >Mode 2 – 33.3MB/s <-- ATA/ATAPI
> >Mode 1 - 25MB/s <-- ATA/ATAPI
> >Mode 0 - 16.7MB/s <-- ATA/ATAPI
>
> >http://en.wikipedia.org/wiki/Programmed_input/output
> >(better than pcguide PIO link. which is missing mode 0 and modes 5,6
>
> >http://www.pcguide.com/ref/hdd/if/ide/modes_UDMA.htm
> >pcguide UDMA link.
>
> You have no guarantee about any mode any card uses unless it
> clearly specifies it. *However, what we can assume is that
> something that makes a CF card more valuable is something
> the manufacturer will not overlook mentioning, ie - if it is
> CF3 or 4 spec. and the inherant support those mandate.
> Beyond that, your choices are really quite limited since you
> want not to spend a lot, with the higher cost being the
> primary thing keeping many from moving to flash based
> storage.-
write endurance - solid state flash drives, windows xp, virtualmemory 
Linear Mode
