Jo22 wrote on 2024-03-03, 08:36:

Good point, taken. Thanks. […]
Show full quote

kingcake wrote on 2024-03-03, 02:32:

Jo22 wrote on 2018-09-19, 06:10:

No issues with CF cards so far.. I'm using quality CF-IDE adapters with voltage regulators and cheap ones (with diodes added manually).

Diodes are not the solution. The forward voltage drop of a diode changes with forward current. The Vf in the datasheet is only valid at the one testing current. If you go further down in the datasheet you will find the graph of forward voltage drop vs forward current.

So as the CF card current consumption varies the voltage will also be going up and down.

Good point, taken. Thanks.

I've checked the interwebs and it seems that certain CF cards are more vulnerable to power fluctuations.

In addition, Compact Flash cards for the consumer market are often based on battery-powered systems. Batterie- powered systems s […]
Show full quote

In addition, Compact Flash cards for the consumer market are often based on battery-powered systems. Batterie-
powered systems such as notebooks, digital cameras, etc. mean that abrupt voltage drops do generally not occur.
This is not the case with industrial applications without UPS.
At the same time, most CF cards perform active “wear leveling“ in order to distribute the memory wear evenly on the
card.
If a voltage drop occurs during critical write phases, faults up to total failures may occur on many Compact Flash
cards.
For this reason, great value is placed on the Power OFF ruggedness when selecting CF cards.

https://cache.industry.siemens.com/dl/files/1 … lash_en-enu.pdf

By contrast, industrial cards seem to support various voltages.

Industrial CompactFlash (CF) cards are designed to operate in demanding and rugged environments. Key features that make them sui […]
Show full quote

Industrial CompactFlash (CF) cards are designed to operate in demanding and rugged environments. Key features that make them suitable for such conditions include:

1. **Temperature Range:** Industrial CF cards are built to withstand extreme temperature variations, both high and low, making them suitable for applications in harsh climates.

2. **Shock and Vibration Resistance:** These cards are engineered to resist shocks and vibrations commonly encountered in industrial settings, ensuring data integrity even in physically demanding environments.

3. **Durability:** Industrial CF cards are constructed with durable materials to withstand rough handling and challenging conditions, making them more robust compared to standard consumer-grade cards.

4. **Extended Lifecycle:** Industrial CF cards often have a longer lifecycle and are built with higher endurance to withstand frequent read/write cycles, making them suitable for continuous use in industrial applications.

5. **Error Correction:** Enhanced error correction mechanisms are implemented in industrial CF cards to ensure reliable data storage and prevent data corruption, especially in situations where the environment may introduce interference.

6. **Power Failure Protection:** Some industrial CF cards include power failure protection mechanisms to safeguard data during unexpected power outages, ensuring data integrity in critical applications.

7. **ECC (Error-Correcting Code):** Industrial CF cards commonly utilize advanced ECC algorithms to detect and correct errors, enhancing the overall reliability of data storage.

8. **Wide Operating Voltage Range:** To accommodate fluctuations in power supply, industrial CF cards often support a wide operating voltage range, ensuring stability even in situations where power levels may vary.

9. **Data Security:** Enhanced security features, such as write protection switches and secure erase functions, are often incorporated into industrial CF cards to prevent unauthorized access and protect sensitive data.

10. **Compliance with Industry Standards:** Industrial CF cards may comply with specific industry standards (e.g., MIL-STD-810G for military applications) to meet the requirements of various rugged environments.

These features collectively make industrial CF cards well-suited for applications where reliability, durability, and data integrity are critical, such as in aerospace, defense, industrial automation, and other harsh environments.

Source: https://www.quora.com/Do-industrial-CF-cards- … mer-grade-cards

I suppose it shouldn't hurt to attach a higher capacity electrolyte capacitor to the CF card adapter, to countermeasure short power sourges? Say, 470 µF?

Adding bulk capacitance can't really hurt anything. Just a standard electrolytic would be fine. No real need for low ESR in that scenario.

470uF is probably a little unnecessary given the current draw of a CF card. I'd probably bodge in something in the 22uF-120uF range. But if 470uF is what you already have then it shouldn't hurt.

I’m in the group that has not had compact flash devices spontaneously die, but I have had bad luck with some micro SD cards over the years, especially ones that I move between sockets/devices.

Not to underestimate your frustration or ability, because maybe you got really unlucky, but Occam’s razor suggests that when a wide variety of compact flash devices all die at a single location, the issue is specific to your location, not with the wide variety of compact flash devices.

douglar wrote on 2024-03-03, 13:53:

Not to underestimate your frustration or ability, because maybe you got really unlucky, but Occam’s razor suggests that when a wide variety of compact flash devices all die at a single location, the issue is specific to your location, not with the wide variety of compact flash devices.

IMHO, this is where it would be worth exploring the cause. If something about a user's environment, usage or something else is causing these failures, then determining the cause could lead to a preventative solution.

On the other hand it could just be bad luck. With a large enough sample size, there are going to be outliers.

I wanted to mention here that I have had problems with both CF->IDE adapters, as well as CF cards. The adapters seem to be entirely passive and foolproof but some machines don't like certain adapters. So it's good to have a small variety to learn what works for you. CF cards are also unpredictable. Some work. Some don't. If it has the removable bit set, it will now work as a fixed disk in some operating systems and you will be unable to configure your swap file.

Why does this CompactFlash to IDE adapter behave differently?

Using a CompactFlash adapter to workaround UDMA controller and BIOS bugs

Re: Is the Promise UltraATA133 TX2 (PDC20269) a CPU speed sensitive controller?

Kahenraz wrote on 2024-03-03, 14:44:

The adapters seem to be entirely passive and foolproof but some machines don't like certain adapters.

I've discovered some differences in the past 2 years for the 40 pin pata to 50 pin pata adapaters--

• Some have voltage regulators of various quality for cards that require 3.3V but the cheaper ones don't
• Most attempt to force "true ide mode" by grounding pin 9 but some don't
• Some provide CSEL via jumper but some don't
• In rare cases, some have shorts or weak/missing signals because of broken solder joints

It would be nice if the 40 pin female adapters altered pin 34 so that they could do UDMA >=3 without mods or funny cables, but I have not seen that as a feature.

Not to mention that CFs evolved from '94 through '14 (ATA-2 to ATA-8) , which is comparable to 40 pin PATA storage ( '85 to '05, ATA-0 to ATA-7 ), so they have a family tree of designs and firmware that's just about as complicated. I guess CF devices should always support LBA28 addressing, which simplifies some things, but the number of working mechanical drives made before 1994 is getting smaller by the year, so that's rarely much a distinction anymore.

I also forgot to mention that some motherboards detect the adapter as always being a 40-pin cable and throw a warning at the BIOS screen on boot.

compact flash was a dud for me performance-wise, but i never experienced corruption issues on industrial labeled cards. i moved on to sata and and msata using cheap chinese adapters.

Kahenraz wrote on 2024-03-03, 17:23:

I also forgot to mention that some motherboards detect the adapter as always being a 40-pin cable and throw a warning at the BIOS screen on boot.

That's because you have a device capable of UDMA 4 or faster plugged with pin 34 connected ( AKA you don't have an 80 wire cable )

If you short pin 34 to ground, the message goes away and you get faster transfers.

https://www.vogonswiki.com/index.php/Storage# … compatibilities

Re: UDMA 5 with DOM's & CF's

ykot wrote on 2024-03-02, 19:39:

Just wanted to add that my experience with CF cards have been exactly those as described by OP. I have purchased brand new compact flash cards from Amazon, multiple brands such as Kingston, Sandisk and Verbatim, with sizes ranging from 512Mb to 32Gb. Some of these cards were used just once, others were just opened from their box then stored in a sealed container. However, after certain time of being stored, they all stop working - do not get recognized in any of CF card readers, or CF-to-IDE adapters. The time required for each particular CF card seems random - for some it was less than a year, for others a couple of years, but eventually they all died, including completely new ones.

Can you post a photo of your CF adapter?

Perhaps your CF adapter has a malfunctioning voltage regulator.

Shponglefan wrote on 2024-03-03, 14:13:

IMHO, this is where it would be worth exploring the cause. If something about a user's environment, usage or something else is causing these failures, then determining the cause could lead to a preventative solution.

In my case, the following applies:

1. All CF cards were purchased new between 2014 and 2022, most from Amazon (store, not third-party vendors), a couple from Mouser and few from other retail stores that I picked up directly from shelves.
2. Some were written to, others just opened from their original box and then stored.
3. Once a certain CF card is prepared, for example, by installing DOS/Windows and copying relevant files, it is stored outside of the machine. A couple were left inside the machine in the adapter.
4. The cards remained stored at room temperature 14-18C and humidity < 40%. For what it's worth, the altitude is 1800m.
5. Each of the CF card was working perfectly fine, before it was stored.
6. After some undeterminate time, between six months and a couple of years, when I tried to use previously stored CF again, it was no longer being detected: both USB CF reader and CF-to-IDE adapter would simply behave as if the card is not there. Alternatively, either CF reader or CF-to-IDE adapter would "hang" indefinitely trying to access the card.
7. A couple of cards were more resilient than others in sense that they "kind of" worked intermittently. But after some indeterminate time, once I tried them again, they no longer worked.
8. Eventually, each and every single one CF card "died" in aforementioned fashion.

I threw a big portion of these "dead" cards already. Just for the sake of curiosity, yesterday I've took one of "dead" Kingston 8Gb CF cards that I still have, from a 486 machine, which had a couple of FAT32 partitions and Windows 98 installed (no hard disk overlay installed, because the card is within 8Gb BIOS limit). The original machine "hangs" trying to detect it as I've explained above, same thing happens to CF reader, as well as few other machines, including a modern Ryzen-based build with PCI-E IDE adapter card. However, using the same CF-to-IDE adapter from original 486 machine but on a different build - a Mendocino-based Celeron, the BIOS actually recognized the card! However, the card appears unformatted and when dumping contents from the beginning using "dd" command on Linux, appears to have gibberrish. I've filled it with zeroes and could install Windows 2000 successfully. After having done that, the card now gets recognized on other machines as well! However, subjectively speaking, it feels to be much slower than when I first bought it, especially for writes.

I've tried to apply the same technique to a pair of other "dead" 512Mb and 32Gb CF cards respectively, but, unfortunately, these aren't recognized on that same Celeron machine. I'll keep trying with my other builds, however, and will report if can revive them somewhere.

So it looks like NAND flash data retention issue could be a real problem. I wonder if CF controller stores any book-keeping information on the card itself, so when corrupted, the card becomes unusable.

rasz_pl wrote on 2024-03-03, 03:50:

I should probably word it differently, "losing charge". There is still some time between less reliable read and lost data. Good example was samsung 840 https://www.extremetech.com/computing/190746- … -fix-on-the-way samsung finally released a "Performance Restoration Tool" and firmware update forcing those drives to aggressively refresh old data https://pcper.com/2014/10/samsung-updates-840 … storation-tool/
850 EVO same problems despite 3D V-NAND https://goughlui.com/2016/11/08/note-samsung- … ce-degradation/
It can also go horribly wrong like Chinese attempts at fabbing modern silicon https://goughlui.com/2023/10/10/psa-ssds-with … heck-your-ssds/

How depressing, I actually happen to have two Samsung 850 EVO SSDs and they both exhibit the problem of "slow reads", which I thought was a firmware problem.

As a counter argument, I've found this:
https://www.pcworld.com/article/427602/debunk … -after-all.html

However, for past 2 years, I've been struggling with the issue that my main working system running Arch Linux with full disk encryption stored on a Kingston 120Gb SSD would eventually start to have very slow reads with constantly high iowait. My workaround was to buy another exactly the same SSD, clone the data to it and use it, which resolved the problem for another 6 months, after which I have to buy yet another SSD. If I re-use previous SSD by first issuing TRIM and then cloning the data back, the problem only goes away for a short while, maybe a day or so. Again, I attributed this to a bad firmware, which I hoped would get fixed one day. The data on these SSDs was written once and then never changed, there was no swap file as workstation has 128Gb of RAM. I knew that SSDs and other flash-based storage would lose the information eventually, but I thought it would take decades, not days.

This is very sad, but it has been both informative and revealing, thanks!

ykot wrote on 2024-03-03, 18:53:

In my case, the following applies: […]
Show full quote

In my case, the following applies:

1. All CF cards were purchased new between 2014 and 2022, most from Amazon (store, not third-party vendors), a couple from Mouser and few from other retail stores that I picked up directly from shelves.
2. Some were written to, others just opened from their original box and then stored.
3. Once a certain CF card is prepared, for example, by installing DOS/Windows and copying relevant files, it is stored outside of the machine. A couple were left inside the machine in the adapter.
4. The cards remained stored at room temperature 14-18C and humidity < 40%. For what it's worth, the altitude is 1800m.
5. Each of the CF card was working perfectly fine, before it was stored.
6. After some undeterminate time, between six months and a couple of years, when I tried to use previously stored CF again, it was no longer being detected: both USB CF reader and CF-to-IDE adapter would simply behave as if the card is not there. Alternatively, either CF reader or CF-to-IDE adapter would "hang" indefinitely trying to access the card.
7. A couple of cards were more resilient than others in sense that they "kind of" worked intermittently. But after some indeterminate time, once I tried them again, they no longer worked.
8. Eventually, each and every single one CF card "died" in aforementioned fashion.

Are you able to get any of the S.M.A.R.T. diagnostic data off the cards?

https://en.wikipedia.org/wiki/Self-Monitoring … ting_Technology
https://en.wikipedia.org/wiki/Comparison_of_S.M.A.R.T._tools

If you can't get any S.M.A.R.T. diagnostic info back, I'd suspect that the controller chips are dying, not the flash cells.

douglar wrote on 2024-03-03, 21:54:

Are you able to get any of the S.M.A.R.T. diagnostic data off the cards? […]
Show full quote

ykot wrote on 2024-03-03, 18:53:

In my case, the following applies: […]
Show full quote

In my case, the following applies:

1. All CF cards were purchased new between 2014 and 2022, most from Amazon (store, not third-party vendors), a couple from Mouser and few from other retail stores that I picked up directly from shelves.
2. Some were written to, others just opened from their original box and then stored.
3. Once a certain CF card is prepared, for example, by installing DOS/Windows and copying relevant files, it is stored outside of the machine. A couple were left inside the machine in the adapter.
4. The cards remained stored at room temperature 14-18C and humidity < 40%. For what it's worth, the altitude is 1800m.
5. Each of the CF card was working perfectly fine, before it was stored.
6. After some undeterminate time, between six months and a couple of years, when I tried to use previously stored CF again, it was no longer being detected: both USB CF reader and CF-to-IDE adapter would simply behave as if the card is not there. Alternatively, either CF reader or CF-to-IDE adapter would "hang" indefinitely trying to access the card.
7. A couple of cards were more resilient than others in sense that they "kind of" worked intermittently. But after some indeterminate time, once I tried them again, they no longer worked.
8. Eventually, each and every single one CF card "died" in aforementioned fashion.

Are you able to get any of the S.M.A.R.T. diagnostic data off the cards?

https://en.wikipedia.org/wiki/Self-Monitoring … ting_Technology
https://en.wikipedia.org/wiki/Comparison_of_S.M.A.R.T._tools

If you can't get any S.M.A.R.T. diagnostic info back, I'd suspect that the controller chips are dying, not the flash cells.

AFAIK most consumer-grade CF cards do not have SMART. Some cards do have SMART but only a select few not so important parameters are exposed.

Unless the card is meant to be an actual SSD (usually true IDE), manufacturers don't really make SMART available in the firmware.

An exception would be if the card has been MP'd to a fixed disk. A while ago I successfully MP'd a Toshiba CF card (SM2236-based) and it's now a fixed disk with all SMART parameters available. So far I'm only able to MP one out of the many cards, as I don't know what controller those other cards use... and will definitely do so if possible...

douglar wrote on 2024-03-03, 18:16:

Can you post a photo of your CF adapter?
Perhaps your CF adapter has a malfunctioning voltage regulator.

As I've said earlier, among multiple adapters, I use StarTech 35BAYCF2IDE, which doesn't have any voltage regulating circuitry. I also got a couple of adapters that do seem to have voltage regulator, but they are set to 5V, which bypasses the regulator and IT IS the voltage you should use according to IDE specification. I'm using high-tier power supplies and I really don't think it's related to voltage in any way because we are talking about multiple systems each with its own power supply and adapter, and there doesn't seem to be any connection. Also, according to CF specification, Vcc can go up to 6.5V max, which gives plenty of room as far as voltage is concerned.

douglar wrote on 2024-03-03, 21:54:

Are you able to get any of the S.M.A.R.T. diagnostic data off the cards?

How can I read SMART from the cards if they are not recognized at all or the system hangs while trying to detect them? I'm going to get SMART from the card that I've partially revived though.

ykot wrote on 2024-03-03, 18:53:

This is very sad, but it has been both informative and revealing, thanks!

The whole situation with industry pushing for flash + tape and slowly phasing out magnetic HDDs is quite depressing. It means not to long in the future users will be forced to Cloud storage for long term preservation 🙁

douglar wrote on 2024-03-03, 21:54:

Are you able to get any of the S.M.A.R.T. diagnostic data off the cards?

Attached is SMART from a partly revived CF card. By the way, it seems to be recognized as fixed disk.

Edit: added another CF that I could "revive" using aforementioned technique, it's exactly the same capacity and brand as the other one. This one had some partitioning left that could be recognized, but still I couldn't mount its partitions. I've reformatted it. Curiously, it has more SMART information.

rasz_pl wrote on 2024-03-04, 02:07:

ykot wrote on 2024-03-03, 18:53:

This is very sad, but it has been both informative and revealing, thanks!

The whole situation with industry pushing for flash + tape and slowly phasing out magnetic HDDs is quite depressing. It means not to long in the future users will be forced to Cloud storage for long term preservation 🙁

I've a few 60-120GB SSDs I bought at least a decade ago and they're still good to be used.

I personally don't consider SSDs unfit for long-term storage, however. Given adequate care they'll continue to be usable for quite a while.

Mechanical HDDs are not perfect anyway. They need to be carefully handled, and from my past experiences, they can fail in all kinds of unexpected ways.

I think the problem with SSDs is that the NAND market can be a bit messy. Flash memory chips deemed of inferior quality are often recycled for use with lower-end SSD models as well as by less-known brands which get sold cheaper than average.

If your SSD needs to be powered on every week to not lose data then you have a bad SSD. That's not "the state of things" as you say. That is 100% abnormal.

Mechanical drives also fail if left unpowered for a few years. The ball bearings in the spindle motor will sink into the nylon races due to gravity and being in the same position for years. That causes a bearing failure, and since the platters move down you get a head crash when you power on.

ykot wrote on 2024-03-04, 03:37:

douglar wrote on 2024-03-03, 21:54:

Are you able to get any of the S.M.A.R.T. diagnostic data off the cards?

Attached is SMART from a partly revived CF card. By the way, it seems to be recognized as fixed disk.

Edit: added another CF that I could "revive" using aforementioned technique, it's exactly the same capacity and brand as the other one. This one had some partitioning left that could be recognized, but still I couldn't mount its partitions. I've reformatted it. Curiously, it has more SMART information.

I didn't find a match for the SMART attributes from the device with firmware = Ver6.04K . I looked in these spots:
https://www.data-medics.com/forum/threads/lis … -specific.1476/
https://media.kingston.com/support/downloads/ … T_attribute.pdf
https://en.wikipedia.org/wiki/Self-Monitoring … ting_Technology

If I had to guess,
E8 is probably Endurance Remaining 99%
E9 and EC might be block counts of some sort
EB is 100%, whatever that is

kingcake wrote on 2024-03-04, 10:11:

If your SSD needs to be powered on every week to not lose data then you have a bad SSD. That's not "the state of things" as you say. That is 100% abnormal.

not weekly, but at least once a year after few thousand cycles, the closer you are to end of life the shorter it gets
https://www.vencoel.com/wp-content/uploads/20 … bility-Note.pdf and this is the absolute leader in NAND flash, former Toshiba - they invented this stuff

rasz_pl wrote on 2024-03-04, 18:21:

kingcake wrote on 2024-03-04, 10:11:

If your SSD needs to be powered on every week to not lose data then you have a bad SSD. That's not "the state of things" as you say. That is 100% abnormal.

not weekly, but at least once a year after few thousand cycles, the closer you are to end of life the shorter it gets
https://www.vencoel.com/wp-content/uploads/20 … bility-Note.pdf and this is the absolute leader in NAND flash, former Toshiba - they invented this stuff

I was responding to the guy that said his SSD needed to be refreshed every week.

Of course nand needs to be refreshed now and then. I didn't say it never needs to be refreshed.

No offense, but is English not your native language? You seem to follow me around misunderstanding my posts.