dragonkn wrote on 2022-01-22, 12:22:

I want to connect CF Card to Ide - Intel Pentium 2 system […]
Show full quote

I want to connect CF Card to Ide - Intel Pentium 2 system

Could I just ignore it or do i need to connect psu power?
Should i set 3.3v (default) or 5v?
I want to use it with generic 1gb kingston cf card 33x speed

If its a 44 pin IDE cable IIRC it can get 3.3v from that which is the default setting, if its a 40 pin IDE setup then Yes, the 4o pin IDE cable doesnt supply power at all so the adaptor needs a supply of 3.3v/5v from the floppy to power the CF card.

IF you are using this in a PC then youll need to attach power to it.

44 pin IDE is Laptop IDE
40 pin is PC IDE - this one needs the floppy power connector

The 3.3v/5v setting is used if you are using a CF card that need 5v to run, not seen one like this myself so best to leave it on 3.3v. (Turns out CF can run at either voltage, 3.3v is default)

Connect a floppy drive power cable and set it to 5V. All compact flash cards can run on 5V, using 3.3V just enables a pointless voltage regulator (the 5V IDE signals still get passed directly to the CF card...).

I have a adapter with diodes, I had problems with some cards (especially hitachi microdrive), I replaced some with jumpers (5V only).

TrashPanda wrote on 2022-01-22, 12:25:

The 3.3v/5v setting is used if you are using a CF card that need 5v to run, not seen one like this myself so best to leave it on 3.3v. (Turns out CF can run at either voltage, 3.3v is default)

I think 16 or 32 Megabytes was HUGE for a 5v era card, and 2 or 4MB more common, so I kinda doubt many people would be using 5V cards now.

Some 40pin connectors do carry 5v on the key pin (if present), if memory serves..

As for the voltage.. 3.3v is the proper way to go.

Don't get me wrong, I love to tinker with 5v.
It's one of the traditional voltages (+/-12v, +/-5v) and used to be used by TTL.

But let's get realistic - 5v is outdated since the Pentium era.
Even floppy drives from the mid-90s did use 3.3v power and 3.3v signals.

So yes, virtually all CF cards are made with 3.3v in mind.
They can handle 5v, of course. But it's not exactly healthy for these cards.

Unfortunately, cheaper adapters don't incorporate a voltage regulator.
They don't even have caps. They are purely mechanical adapters.

Edit: The data lines may carry 5v signals from/to IDE, yes.
Especially, if installed in 8086-486 era systems.
Not sure how CF cards do handle these internally.
If pull-up/pull-down resistors or tri-state pins are used,
they may not directly get in touch with the voltage on the data lines.
Instead, they get their signals through the VCC voltage. Which is 3.3v, if a voltage regulator or a pair of diodes (in series) is installed.

Jo22 wrote on 2022-01-23, 09:10:

So yes, virtually all CF cards are made with 3.3v in mind.
They can handle 5v, of course. But it's not exactly healthy for these cards.

They handle 5v perfectly fine because the compact flash mechanical specification says they must. Saying it isn't "healthy" is rubbish - especially when they are operated in "true IDE" mode (by grounding a pin, which is done by the IDE adaptor) and IDE uses 5 volts by design.

Even when using 40pin.

Some connectors do provide VCC on pin 20 and many CF cards happily power the card from pin 20 if available.

But it is more likely than not that you have to use external power. I modified some of my IDE controllers to supply +5V on pin 20.

jmarsh wrote on 2022-01-23, 14:18:

Jo22 wrote on 2022-01-23, 09:10:

So yes, virtually all CF cards are made with 3.3v in mind.
They can handle 5v, of course. But it's not exactly healthy for these cards.

They handle 5v perfectly fine because the compact flash mechanical specification says they must. Saying it isn't "healthy" is rubbish - especially when they are operated in "true IDE" mode (by grounding a pin, which is done by the IDE adaptor) and IDE uses 5 volts by design.

I beg to differ. TTL with 5v is old school.
Modern ICs are designed with 3.3v or less in mind, legacy specifications or not.
(Old IDE specs can't be changed afterwards, after all. It's always an extension.
Old stuff like PIO-0, CHS or 5v can be 'deprecated', at best).

Of course, modern CF cards do *handle* 5v in practice, too. That's what I did explicitly mention. 😀
In a same manner as an 7805 regulator *could* be operated with a 24v source, I guess. 🤷‍♂️

It's just not healthy on the long run. Heat or change in temperature (up/down/up..) makes things brittle.
Turning less power/voltage into heat is thus favorable.
Let's just think of modern CPUs, were +/-0.2v make a big difference..
9v-12v vs 24v-30v would make an 7805 happier, too.

That's why there's such a big controversy going on about using flash carts in vintage systems, by the way.

The modern SRAMs and other parts are designed for 3.3v, so people do argue, but are forced to interface with 5v logic.
Some flash cart designs use diodes - which works fine practice, but critics still complain about the heat dissipation.. Diodes are not meant to dissipate heat, they say.

Example:
https://www.youtube.com/watch?v=vv4TeyLcSSw

Anyway, I don't mean to make a big fuss about such things.
In practice, things might be much more forgiving..
Let's think about ESD.. Old TTL chips were too clunky to takw damage, whereas modern CMOS contains built-in ESD-protection.
It merely was young CMOS (SRAMs etc) thst was very sensitive to ESD.
Nevertheless, it's better if users use things with caution.

Edit: Typo fixed.

Edit: Or let's put it this way: Assuming that pull-ups/downs are used on the i/o pins, then the VCC defines the i/o voltage, too.:
Pull-up/pull-down circuits work similar to switches based on traditional mechanical relays.
The incoming voltage (say, 5v via IDE) doesn't directly interface with the chip - it merely powers a switch. Which connects a data line either to plus (VCC) or minus (ground).
Such a switch in turn connects the internal data pin to the chip's power source - which can be eithet 5v or 3.3v - depending on whether a voltage regulator is installed/activated on that CF adapter.
Or let's pretend that switch acts like a telegraph key that sends morse code (on/off; TTL) into the chip.
Anyway, that's merely a simple analogy. I'be never dissambled such a chip myself. Not sure if that's a common way of interfacing, thus.🤷‍♂️

Edit: Modified.

Jo22 wrote on 2022-01-23, 21:33:

I beg to differ. TTL with 5v is old school. Modern ICs are designed with 3.3v or less in mind, legacy specifications or not. (Ol […]
Show full quote

jmarsh wrote on 2022-01-23, 14:18:

Jo22 wrote on 2022-01-23, 09:10:

So yes, virtually all CF cards are made with 3.3v in mind.
They can handle 5v, of course. But it's not exactly healthy for these cards.

They handle 5v perfectly fine because the compact flash mechanical specification says they must. Saying it isn't "healthy" is rubbish - especially when they are operated in "true IDE" mode (by grounding a pin, which is done by the IDE adaptor) and IDE uses 5 volts by design.

I beg to differ. TTL with 5v is old school.
Modern ICs are designed with 3.3v or less in mind, legacy specifications or not.
(Old IDE specs can't be changed afterwards, after all. It's always an extension.
Old stuff like PIO-0, CHS or 5v can be 'deprecated', at best).

You are wrong. The CF specs say CF cards have to comply with PATA specs, and work with both 3.3V and 5V. The easiest link to understand is https://en.wikipedia.org/wiki/CompactFlash (unless you want to google more technical documents, of which there are plenty online).

guys - spolier alert!
I will make another thread for an experiment!
From many topics it turns out that the CF card is not recommended for Windows 95/98
I'll NOT turn off the virtual memory or defragmentation - the windows 95 will be in the default configuration
I'll be using a 'evil' 5v
Also i will restore a whole system backup to CF from time to time
How long will it last? A day, week, year? We'll see

It's New Old Stock Kingston 1gb CF ( no industrial) 33x speed
experiment will start in a few days

dragonkn wrote on 2022-01-23, 22:42:

From many topics it turns out that the CF card is not recommended for Windows 95/98

Is that because of the swap file? I've been thinking about using them for 9x builds. Possibly even 2K. I have a couple of revo drives for XP, and plenty of SSDs for SATA builds.

debs3759 wrote on 2022-01-24, 02:17:

dragonkn wrote on 2022-01-23, 22:42:

From many topics it turns out that the CF card is not recommended for Windows 95/98

Is that because of the swap file? I've been thinking about using them for 9x builds. Possibly even 2K. I have a couple of revo drives for XP, and plenty of SSDs for SATA builds.

Yeah I think a swap file would quickly trash an older CF card but the newer larger ones shouldn't have a huge issue with it, the alternative is to keep at least one spinning rust device in the PC to act as a windows swap drive and general dump drive for unzipping files and other stuff that requires a lot of read / writes.

Sandisk wrote a white paper on CF wearing if anyone is interested, it seems that a modern CF card will likely outlive the machine its put into if subjected to normal use, since being used as HDD replacement isn't exactly normal use you can expect it'll die a bit sooner but its still a long time, likely longer than the retro machine its in will survive before being rebuilt and having the CF card replaced. Larger CF cards also wear slower and so have a large life span, a 64gb CF card could quite reasonably be expected to outlive several retro machines.

https://web.archive.org/web/20150326122100/ht … arLevelv1.0.pdf

If you can avoid the tiny CF cards and use large ones then it'll likely be better in the long run even if it costs a little more to start with.

mpe wrote on 2022-01-23, 16:59:

Even when using 40pin.

Some connectors do provide VCC on pin 20 and many CF cards happily power the card from pin 20 if available.

But it is more likely than not that you have to use external power. I modified some of my IDE controllers to supply +5V on pin 20.

This.

Another thing to keep in mind is that some 40pin adapters don't route pin 20 to the card. So even though the hole is there and your IDE connector supplies power, you'll still need to use the floppy power connector.

debs3759 wrote on 2022-01-23, 22:16:

Jo22 wrote on 2022-01-23, 21:33:

I beg to differ. TTL with 5v is old school. Modern ICs are designed with 3.3v or less in mind, legacy specifications or not. (Ol […]
Show full quote

jmarsh wrote on 2022-01-23, 14:18:

They handle 5v perfectly fine because the compact flash mechanical specification says they must. Saying it isn't "healthy" is rubbish - especially when they are operated in "true IDE" mode (by grounding a pin, which is done by the IDE adaptor) and IDE uses 5 volts by design.

I beg to differ. TTL with 5v is old school.
Modern ICs are designed with 3.3v or less in mind, legacy specifications or not.
(Old IDE specs can't be changed afterwards, after all. It's always an extension.
Old stuff like PIO-0, CHS or 5v can be 'deprecated', at best).

You are wrong. The CF specs say CF cards have to comply with PATA specs, and work with both 3.3V and 5V. The easiest link to understand is https://en.wikipedia.org/wiki/CompactFlash (unless you want to google more technical documents, of which there are plenty online).

We talk past each other, I'm afraid. 😒

I said CF cards are 5v tolerant. I never said that they get destroyed by 5v whatsoever.
I just meant to say that the 5v to 3.3v regulator isn't without purpose.
That's why I used the term "healthy".

Lifespan of flash media in general is very debatable, I guess. 🤷‍♂️
What comes to mind: I had SSDs that started to act funny after a few years of use,
despite an expert friend stating that SSDs last usually for ages.

Same goes for LED lamps, I guess. I remember, my father had an OSRAM light bulb in the bed room that lasted for 20 years until it broke.
Now he replaces LED bulbs about three times a year.
The culprit usually is the power supply. The missing of caps. Short voltage spikes (overvoltage) often kill the rectifying diode(s).
The white/blue LEDs themselves aren't broken.
Anyway, that's not really on-topic. It just came to mind (old vs new).

Attached is a screenshot to what I was referring to.
The complete PDF is available below.

https://www.google.de/url?q=https://www.mouse … xdnE7EZ62-DOtd-

dragonkn wrote on 2022-01-23, 22:42:

From many topics it turns out that the CF card is not recommended for Windows 95/98
I'll NOT turn off the virtual memory or defragmentation - the windows 95 will be in the default configuration

Good luck! 🙂

Last time I used Win 98SE/Me on a CF card (4GB or 8GB Transcend?), it worked fine.

Except.. Behaviour of Windows was "jerky". Windows or Windows Explorer sometimes froze for ~half a second when the CF card was busy performing multiple read/write requests at once.

Using DMA helped making things smoother. Disabling swap file, even more so.

System was an HP thin client with 64MB RAM or so.

I played around with a lot of different CF devices in retro PCs in the last two years.

I suspect that some day there will be a new CF that doesn't like 5v, but that has not happened yet.

The only CF device that I've encountered that required 3v to work was an SD -->CF adapter.

It's essentially a Sinitechi SD-->IDE adapter in a different form factor. Same internal chip, similar firmware versions, same 25MB/s max SD read speed.

Jo22 wrote on 2022-01-24, 10:32:

Last time I used Win 98SE/Me on a CF card (4GB or 8GB Transcend?), it worked fine. Except.. Behaviour of Windows was "jerky". […]
Show full quote

Last time I used Win 98SE/Me on a CF card (4GB or 8GB Transcend?), it worked fine.

Except.. Behaviour of Windows was "jerky". Windows or Windows Explorer sometimes froze for ~half a second when the CF card was busy performing multiple read/write requests at once.

Using DMA helped making things smoother. Disabling swap file, even more so.

System was an HP thin client with 64MB RAM or so.

Yes, enabling DMA I/O helps a lot. Using PIO transfers on a 486 or pentium system with a storage device that is unimaginably fast for the time period can cause a CPU load that makes Win98se less responsive to user input and feel jerky.

And yes, the default Win98se swap file behavior didn't age well. By default, Win98se attempts to proactively write inactive mermory pages to the swap file before they need to be paged, so when it comes time to page, there is less work to do. That extra background I/O didn't so hurt much when the disk was slow and it made things work more smoothly when many systems only had 16MB of RAM to page. Now we have retro systems with maxed out ram & flash storage. That makes the unnecessary swapfile writes undesirable. The background IO to a superfast device causes unwanted spikes of CPU load & bus contention, and the constant drip of unnecessary I/O is not good for the life span of flash storage.

Disabling the swap file is one way to avoid that. Another way is to add this entry to your system.ini file under [386enh] :

1ConservativeSwapfileUsage=1

ConservativeSwapfileUsage = 1 tells Win98se to only write pages to the swap file when they need to be swapped out.

douglar wrote on 2022-01-25, 15:26:

Disabling the swap file is one way to avoid that. Another way is to add this entry to your system.ini file under [386enh] : […]
Show full quote

Disabling the swap file is one way to avoid that. Another way is to add this entry to your system.ini file under [386enh] :

Copy code to clipboard

1ConservativeSwapfileUsage=1

ConservativeSwapfileUsage = 1 tells Win98se to only write pages to the swap file when they need to be swapped out.

Oh jeez! Where were you in the days of my youth (1997) ?! 😀)

Does this feature also works for Win'95?

froller wrote on 2022-01-25, 18:12:

Does this feature also works for Win'95?

"ConservativeSwapfileUsage = 1" makes Windows 98 work like Windows 95. If you see unwanted paging in Windows 95, your only options are add more ram & disable the swap file.

https://www.betaarchive.com/wiki/index.php/Mi … _Archive/223294

Windows 98 added a new feature, PageFile_Call_Async_Manager, that allows the Memory Manager to asynchronously write out page file (swap file) buffers during periods of time when VFAT file system activity is not busy.

This feature can affect the behavior of VxDs that monitor and/or otherwise intercept PageFile VxD functions. This article is applicable to you if your VxD hooks PageFile_Read_Or_Write, and you discover that you are not seeing all the page file traffic when using Windows 98.
...
When the above Boolean is set to TRUE as shown, PageFile_Call_Async_Manager is never called. If this entry is absent from System.ini, the default setting for ConservativeSwapfileUsage is 1 for Windows 95, and 0 (zero) for Windows 98.