I've used CF cards on and off over the years and they hold up pretty well depending how they are used as well what is running on them. Ran a 32gb 400x in a iBook G3 some years back and used it pretty heavily which took months before slowing down a tad. Used a second gen 30gb kingston ssd for swap and it took a year and a half before doing the same. For retro use they are ideal considering how most rigs get used these days. Good as gold for retro laptops.

Anyone have any experience with a PCI -> CF card?

http://www.ebay.com.au/itm/PCI-To-4-Port-CF-R … =item27c24186fb

konc wrote:

Just an addition to the thread, I've found only by experience (and contrary to what theory suggests) that DOM's work much smoother in windows environments than a CF card+adapter. For example, this is the SAME pc with a 600x CF on an adapter running windows XP vs a rather cheap DOM from ebay.

You know your hardware knowledge is no longer quite up to date when you think to yourself "what the hell is a DOM" and proceed to Google it. 🙁

Hahaha. Guilty here as well!

adslh8yp wrote:

Anyone have any experience with a PCI -> CF card?

http://www.ebay.com.au/itm/PCI-To-4-Port-CF-R … =item27c24186fb

Interesting. Has a Silicon Image chip and a RAID BIOS. Seems to be a combination of PCI SATA controller, SATA to IDE bridge and IDE to CF adapter. Personally you might as well go with SATA hard drives.

Curious as to why Silicon image sata and IDE ATA\100 controllers are recommended so often for 9x.
As I've not ever used one.

Does 98se already have a default driver for them? Meaning no chance for the boot delay bug to appear later on? Which always seems to happen with Promise and via pci controllers in 9x.
Are they immune to this bug?

Boot delay bug?

SI controllers are cheap and easy to obtain. It might have helped that I keep bringing them up and showing the solution in my videos 😀

So far I tried SI and Promise (SATA 300 TX2 I believe). Both will work fine without drivers but a PCI device is left undetected in device manager. Then you just load a driver and it will show up under SCSI devices in device manager.

For DOS it just works.

Personally I now prefer using SATA to IDE bridges. The 5.25" ones that you put on the back of a SATA drive. It just makes for an easier configuration.

What I couldn't get to work are VIA based controllers despite coming with RAID BIOS...

Yes after about 30-40 reboots, if your using a Promise or Via card as a boot drive in a 9x machine, you will get a long 30secondish pause during boot.
Some reason the cards get moved to the end of 9x initializing them or something and a delay gets added.
Its annoying they do it on any machine no matter the model from ATA66 to Sata1, in any machine I've tried using them in.
Just wondering if SI does this too.
This problem never occurs with NT4\2k\XP\
Just 9x.

Fascinating. Got script that reboots 50 times 🤣

I got a promise currently in a GB 586ATX with 430TX chipset. I guess I never noticed this bug because my projects are usually finished within that many reboots.

I just deal with it on my Tualatin with the promise controller.
This problem doesn't happen if it, is not being used to boot the system.
Theres actually old posts you can google on this same problem with Mobos that have promise ATA\100 raid controllers from the Athlon XP era that had these promise controllers integrated into them.

Funny how adding something bring a whole lot of new problems with it 😀

On the 586ATX, performance of the Promise controller is quite weak. A CF card in PIO mode beats it. I will try DMA mode next and see if it approaches the theoretical 33 MB/s. I'm not sure why the board doesn't like the SATA to IDE converters but I still got another model to try.

I use the SI controller really for MS-DOS in my Super Socket 7 time machine. But it also has a bay for CF cards. So it's a versatile solution. But under DOS I haven't had any issues.

CF cards... they really do depend on the controller accepting them.
A topgun mobo that I have which is just an ali rebranded socket 7 AT board.
The onboard controller will refuse to enable DMA mode in 9x and NT even with my transcend CF cards.
But you can use the dos UDMA utility for this chipset to enable it under dos and it works fine... 😐 But of course 9x will think its working in MS-DOS mode if you boot into windows (from dos), slowing the system down.

I guess sticking with Intel will also avoid issues. At least that's what I do, sticking with Intel, Nvidia, 3dfx, Matrox and you got less issues to begin with.

I've never measured read\write performance on that machine, but like you found above PIO mode still feels pretty fast on a CF card. Even if its eating cpu cycles 🤣 .
Prefer them over Hard-drives on old machines no-doubt.

ATTO disk benchmark is great for this. Works under 98 and I use it all the time.

The onboard controller will refuse to enable DMA mode in 9x and NT even with my transcend CF cards.

Same issue on my 586ATX. I got some CF microdrives, Seagate and Hitachi, here it enables it fine 😀

But these microdrives are so slow to begin with...

I got some more IDE to SATA bridges coming from eBay. Just wanting to get an overview of all the different chips. So far performance has been really identical. They all max out the IDE controller. To enable DMA a small workaround is required. Just got to go to device manager and change the driver for the the "generic IDE drive" to "IDE drive". No idea why it works, but without doing this it won't boot again (J Micro) or boot in slow motion (other controllers). I really got to document this all one day on my website 😵

Here some benchmarks on a 586ATX with Intel 430TX and Windows 98:

32 GB limited modern SATA notebook drive on Promise controller:

32 GB CF card through IDE to CF adapter:

4 GB Seagate microdrive quickly reaching it's limit (this is with DMA box ticked):

It looks like the CF card performs better with large files than the SSD, while the SSD trumps it when it comes to smaller files. I wonder, which do you think "feels" faster from a subjective standpoint, just running games and such?

They all feel fine, I mean it's a Pentium 200, so the CPU is often holding things back. The micro drive is very slow when seeking. So I won't be using that. I find the CF card the easiest to work with, but I would still prefer using a SATA to IDE adapter if it would work in this board. I'' try another board with a different chipset...

Ok the SATA to IDE adapter is working on an ASUS SS7 board with SIS chipset. I will see if I can enable DMA with the stock Windows 98 SE drivers or if I need to get some chipset drivers. It's a SIS 5598 chipset.

Here the results with the SIS chipset board and the same Pentium 200 and S3 graphics.

Also! When another board wouldn't detect the HDD through the SATA to IDE adapter I investigated some more. Turned out the IDE cable was faulty 😀 So stay tuned for GB 586ATX tests with DMA enabled and a modern 32 GB capacity limited SATA drive through a IDE to SATA adapter.

IDE to SATA adapter with modern 32GB capacity limited SATA drive, stock Windows IDE drivers, no DMA:

With DMA ticked (Yes it's slower!):

Loading the SIS IDE driver:

Poor performance 😒

Well the DMA setting won't stick, guess this board just won't play ball.

Still it now works with the SATA to IDE adapter. Here are the results for the 586ATX:

Got a 20 GB IDE Western Digital drive, but the 430TX chipset board still won't enable DMA 🙁

So I tried a GA-5AX with ALI chipset. Look at these results! This is with a MMX 233 though.

Stock Windows drivers, just enabled DMA in device manager: