Well, to be frank - you were asking for trouble a bit, there.

I mean, off label media * exotic adapters * notebook environment basically equals trouble³.

Cheap CF cards are dead slow.
(Guess the old digital cameras did not mind too much as they were busy compressing the data before saving...)
They still make sense for really old machines like 486 and before since their speed still matches the PIO mode of the PC or close to that and they don't have any access time to speak of.
(and no noise...)
At least in some data sheet, that CF card will have a "NNNx" rating like 66x. AFAIK, for some weird reasons, that's in "CD-ROM speed", namely, N times 150 kByte per second.

Furthermore,
- win95 rtm/osr1 have no fat32 support so that means 8 fat16 partitions to fully utilize the storage
...well, duh. That's why everyone uses OSR2 ("95B"?) and 98 on anything from a 386 up.
- the usb adapter i'm using only presents the first partition on the card
...why a USB adapter? If you have USB, you could use a USB stick with any number of advantages?
- laptop pcmcia boot option doesn't support ata devices
...PCMCIA, afaik, is almost the same connection as CF and should work pretty "native"?
- cardsoft dos drivers 'appear' to support ata devices, but crash the system instantly upon access
...adding drivers to a drive elevates the story to trouble⁴.
- ridiculously priced ataenab appears to be the only viable dos alternative for now, but it crashes under 386 memory managers, especially when using noems switch
...that's proprietary Toshiba tech? No idea what trouble level that is.
- even in pcmcia ata mode, nt (and later) operating systems restrict access to only the first partition on the card by design
...Well, next good reason to go fat32 then.

Weird idea: That thing has a normal 2.5" HDD bay? Why don't you just slap in a huge pseudo-SSD in there via some 44pin to CF or SD card adapter? Am I missing something?

the laptop has a 810mb hdd which i think is mostly adequate, especially since i don't have to deal with drive overlay software.

the initial idea of the cf card was to transfer files to & from the laptop (which has no pci bus). the only other viable options would be 802.11b wireless or laplink cable (which is actually already in place for softmpu use). i thought the cf card would be an instant winner, but i guess not so much.

the usb cf card reader is for my other systems i'm transferring the files from (most of which don't have ide interfaces anymore). the hot-swapability was also front of mind when i selected compact flash.

i like how windows 95 rtm runs on my laptop. my other car is windows 11.

Well, the 810 MB paper weight in there is already more than 504 MB and since it's a Pentium system, I keep forgetting those rules, but it should happily take 4, 8, if not 32 GB of primary hard disk.
In 2 GB FAT16 increments, if that really can't be helped.
So, I'd seriously consider keeping the CF card, buying a 44pin adapter and losing the conventional HDD. Having C through J of normal hard disk partitions with some speed and no noise still seems like a step forward.

File transfer: 802.11b sounds great. The interface probably won't agree with the encryption of your main wifi, so you'd have to set up a different network for legacy use, but that's a good idea, anyway. It should work. Also, what kind of weird old notebook does have wifi but no ethernet? There are ethernet cards for PCMCIA.
If sneakernet, consider a PCMCIA to SD-card adapter, might give you less grief than CF, for some reason. Have one of those for my old Compaq TC1000, but used it only with XP and linux, so can't tell about Win9x. CF port is native, also whithout a hitch.

How attached are you to the RTM installation on this laptop? You should have enough storage on the HDD to store the Win95 OSR2.5 setup and cab files (I’m thinking about 150mb?), copy them over once then install it. Doesn’t fix the speed situation but you will immediately benefit from FAT32.

The CF card itself should not be a problem. I can achieve around 45 MB/s with it, and the theoretical specs go even a bit further. I assume you have the same “industrial” card from Transcend as they only have one kind of the size of 16 GB.

To be frank your bad experience is trying to use the CF card through the PCMCIA adapter and not plugging it in to the IDE. Since it gives you more trouble than you like/are willing to accept, I suggest looking to network and ftp to transfer files.

you know, it might help if i actually looked at the physical card since there is an actual model number with the complete specs listed online.

holy hell! that's quite some comedown in listed read speed from 87mbs to 550kbps. unless this card is a fake i am bottlenecking like there's no tomorrow. my pcmcia and usb compact flash readers are quite shite.

"The CompactFlash card operates in one of three modes.
The function of some of the pins change with each mode.
The three modes are PC Card ATA using I/O mode, PC Card ATA using Memory mode, and True IDE mode compatible with most IDE drives. "

http://www.interfacebus.com/CompactFlash_Memo … ule_pinout.html

I'm using old HDTune and an USB 3.0 card reader for benchmarking flash media.
Because, personally, I've decided not to be a diehard and so I gave up upon using anything pre-USB 3.0 for testing storage media. 😉
USB 1.x/2.x readers for example do a poor job doing most CF cards justice, from my experience.
The program also has a nice surface scanner, by the way, and a bargraph for displaying any breakdowns during benchmark.
Access time should be a a couple of ms or less.

Edit:
"The PCMCIA 'PC Card' standard [discussed on this page] should not be used as an up-grade because it's based on the slower ISA bus standard.
Use the Cardbus standard of PCMCIA instead, which is based on the PCI bus specification. The PCI implementation in Cardbus is much faster, and a newer bus standard.
I've heard that version 2 [1991] of the PCMCIA standard was not widely followed or was not well defined.
Version 2.1 [1993] addressed some of the standards issues, but may still have issues with standardization.
The PCMCIA 'PC Card' bus is obsolete, and should not be used for new designs.
At a minimum use the PCI version of the bus discussed above.
Removable memory cards are produced in a number of different formats "

http://www.interfacebus.com/Design_Connector_PCMCIA.html

Even generic PCMCIA->CF adapters (which are practically passive since 16bit pcmcia is more or less a direct extension of isa bus and CF cards are native ATA devices and native PCMCIA cards) are good for old laptops which do not support cardbus ATA devices and/or (U)DMA. Keep in mind, however, that for example on Pentium 133 MMX maximum possible PIO performance is around 2MB/s due to CPU bottleneck.

Anyway, nothing beats cardbus ethernet for file transfer on such notebooks. On the same laptop with 133MMX I am able to fully saturate 100mbit ethernet (using 3com 575CT, transfer on linux to /dev/null 😀 ). Writing this data to disk, however, is a completely different story (internal HDD is on ISA bus controller without DMA, who knows why Toshiba implemented PCI onboard yet used ISA for HDD controller)

Moreover, CF cards were notoriously commonly faked in the past. I somehow happened to buy fakes even through official retailers (fortunately got free replacements back then). USB readers are also hit and miss.

It should be possible to access all partitions regardless. I'd guess partitioning was done in a weird fashion - MSDOS / Win9x often have trouble with more than 1 primary partition (even though they should support 4 of them). Also when partitioning using linux tools for example it is easy to forget "lba" flag which causes weird issues in DOS/Win if they expect LBA access.

EDIT:
Forgot to mention that "PCMCIA" bios boot settings commonly referred to PCMCIA floppies and only later laptops detected anything else.
Also, I have a working dos cardsoft setup for regular PCMCIA on my Toshiba SatPro 440CDT: it detects CF cards in dumb adapters without problem. For them to work, however, I always have to force PCMCIA 16 bit mode in bios, IIRC it would lock up otherwise.

chiveicrook wrote on 2022-05-06, 09:21:

It should be possible to access all partitions regardless.

There is an issue with partitions and removable devices (=CF card on usb card reader). Windows, at least until some version of windows 10, only display the first.

konc wrote on 2022-05-06, 09:57:

chiveicrook wrote on 2022-05-06, 09:21:

It should be possible to access all partitions regardless.

There is an issue with partitions and removable devices (=CF card on usb card reader). Windows, at least until some version of windows 10, only display the first.

Yup, that issue used to be normal for all Windows NT derivatives.
Ironically, Windows 9x doesn't seem to have this issue, if memory serves.

The multi partition support thing was introduced "somewhen" in a later version of Windows 10.
- I'm sorry, I lost track since MS used awkward build numbers instead of proper version system. 😅

Other systems like MacOS 9 or OS X didn't have this limitation.
They did show all kinds of partitions on removable media.
Windows 11, of course, does show them, too.
So users of Windows 1x may think that something must be broken with the flash medium if a non-Win 1x user can't see the extra partitions.

konc wrote on 2022-05-06, 09:57:

chiveicrook wrote on 2022-05-06, 09:21:

It should be possible to access all partitions regardless.

There is an issue with partitions and removable devices (=CF card on usb card reader). Windows, at least until some version of windows 10, only display the first.

Argh, too much linux usage for me I guess 😀

To make matters more interesting I decided to experiment with this:
1. Grabbed an old 1GB card, plugged into internal usb cf card reader, removed partition and created 2 of them in Windows 10 with aomei partition manager
Result: Only 1st partition visible in Win10
2. Took the same card, put it in my 440CDT running Debian Jessie via passive PCMCIA adapter, used cfdisk to wipe partitions and create 2 new ones.
Result: Both partitions are visible in Win10

Unfortunately I don't have a spare reader to test in 9X or XP and I've nuked my msdos on 440CDT when realigning partitions after HDD->SSD migration.
Once I restore it I'll check what cardwiz software under dos sees but I vaguely remember there is a switch option to set number of partitions in atadrv and/or mtddrv.

EDIT:
Tried also 3. Grabbed an old 1GB card, plugged into internal usb cf card reader, removed partition and created 2 of them in Windows 10 with builtin disk-management
Result: Both partitions are visible in Win10

that is good info to know about win10/11 dropping the asinine partition restriction on removable media introduced with winnt. i can confirm win95 rtm nor win98 se have this restriction when reading from a pcmcia adapter; however, my usb cf card reader is limited to only the first partition (even in win98 se). the reader is an old 6in1 usb deal so that's probably it. i confirmed it's using the ehci (not ohci) bus, but it's showing its age since the sd reader can't read sdhc.

the laptop is all about 16bit as i can confirm there is no pci bus whatsoever (to my consternation as i can't use umbpci). the pc chips 65545 vga controller *might* be 32bit, but there are isa, vlb even pci revisions of this chip. i actually owned this physical laptop back in the day so there's nostalgia, but it really perseveres since it has non-pnp ess688 sound with real opl3.

i have the laptop manufacturer's specific cardsoft drivers and they work! the cf adapter is properly identified (beeeep) and drive letters are assigned, but even accessing the drive letter instantly crashes the system.

i didn't think about pio or cpu bottleneck. i suppose i'm getting it from both sides when transferring from pcmcia ata to ide drives. it's a p54c 75mhz on a 25mhz fsb.

i guess i'll be looking into getting a usb 3.0 cf reader (probably an sd reader too).

I've just finished experimenting with msdos. It seems that my memory was flaky and it doesn't support multiple partitions on cf cards. Best I can do is to get a card recognized with two drive letters but access to them fails.

stanwebber wrote on 2022-05-06, 15:04:

i have the laptop manufacturer's specific cardsoft drivers and they work! the cf adapter is properly identified (beeeep) and drive letters are assigned, but even accessing the drive letter instantly crashes the system.

Best bet for compatibility is to use bundled atainit.exe to "partition" and then standard dos "format" to format CF card under dos. It doesn't seem to support more than 1 partition however.
Also I don't have >1GB CF cards to test compatibility with larger capacities.

stanwebber wrote on 2022-05-06, 15:04:

i didn't think about pio or cpu bottleneck. i suppose i'm getting it from both sides when transferring from pcmcia ata to ide drives. it's a p54c 75mhz on a 25mhz fsb.

I might have been slightly wrong about cpu bottleneck (after retesting P133mmx can do 4MB/s on internal drive and 2MB/s on pcmcia cf) but both CPU speed and ISA bus (8MB/s max or so) severely limit achievable speeds.

There's a high chance that your CF card is being configured as Removable and it will appear to Windows as a Removable disk even when you connect it directly to IDE or using PCMCIA adapter.

A few years ago I tried experimenting with them as substitutes to mechanical HDDs for some use cases, as I thought it was the same as IDE, but in the end I was wrong, thanks to this "Removable bit".

Non-removable (aka TrueIDE) cards are very rare and sellers usually do not announce or document such so it's very difficult to know beforehand.

Until some recent builds of Win10 only the first partition of any removable device would be visible. For Windows NT family (since 2000), this also applies to "Removable" CF cards. Win9x/ME does not consider these CF cards as Removable, though.

Later versions disabled that partition limit for removable disks, probably because some USB live media creators (like Rufus) started creating a second ESP partition containing a NTFS-compatible loader, so you can use NTFS as opposed to FAT32 when creating Win10/11 install media for UEFI scenarios (need to disable Secure Boot, though).

Certain functionalities may not be usable on a disk considered "removable". My experience is that Windows NT family (since 2000) will not behave well if installed and booted from a "Removable" CF card as certain specialized syscalls for listing disks and partitions would not report the removable ones causing Windows to try accessing floppy drives then error out with obscure error codes under certain circumstances (namely with installers).

There's a driver called cfadisk which can make a "Removable" CF card appear as a fixed one, so Windows will treat the card as a normal SSD. However, on later Windows, you need to enable Test Mode and test sign the driver to avoid having to take a detour to system recovery (to disable signature enforcement) every time you boot.

that tip about cfadisk is brilliant!! i actually was mistaken about my usb cf card reader only presenting the first partition. the reader presents multiple devices to the os from an internal (to device) hub and i got things mixed up. i replaced the correct device driver with cfadisk and all partitions are now displayed in winxp (and win98 se). speed still sucks though. it's taking 160min to copy 7gb from the cf card to a local drive.

in any case i now have a viable setup with the cf card partitioned into 2 fat16 and 1 fat32 drives. all the partitions are accessible in dos 7.0/1, win95 rtm, win98 se and winxp. to access the fat32 partition on the laptop i use a win98 se boot floppy with ataenab and doslfn tsr along with a long file name compatible dos file manager (dn osp).

i don't think i can do anything to improve performance on the laptop side (pcmcia 802.11b would be similarly bus constrained), but a usb 3.0 reader will hopefully alleviate transfer bottlenecks on the other side.

stanwebber wrote on 2022-05-06, 15:04:

p54c 75mhz on a 25mhz fsb.

Just noticed this. I thought that all Pentium 75 models ran on 50mhz fsb with 1.5x multiplier?
What laptop is this exactly? I'm curious about its underlying architecture.

chiveicrook wrote on 2022-05-06, 19:10:

Just noticed this. I thought that all Pentium 75 models ran on 50mhz fsb with 1.5x multiplier?
What laptop is this exactly? I'm curious about its underlying architecture.

you had me second guessing myself for a second, but yeah, the pentium is crippled on essentially a 486 mainboard. it's an nec versa p/75. i got the below excerpts from the service manual.

The CPU board is a 75-MHz (model G8SBJ) board. It connects to the system board via
connectors P1 and P2. The system integrates the Intel’s Pentium P54C chip (75 Mhz inter-
nal, 25 Mhz external).

CPU P54C-75
Clock Speed 75 MHz
System Bus Speed 25 MHz

This is not compatible due to different controller in this Trendscend C170. Get a plain Trendscend industrial without anything else, just speed rating.

One that works as well is Western Digital industrial in 512MB and 1GB. I already tested.

Cheers,