Go SCSI or go home 🤣.

I used to use a CMD IDE controller on my older computers, but I was using Windows XP most of the time. It worked fine I guess.

For the moment, I tend to stick to the onboard controllers (as they have better bandwidth to the chipset) and would use SATA to IDE converters where needed.

I am getting a second IDE to SATA converter soon, will see how good it is.

An SSD or flash card can already take full advantage of the bandwidth of an IDE controller. The interface is also ubiquitous. Why go through the trouble of using SCSI instead?

Kahenraz wrote on 2021-09-29, 02:35:

An SSD or flash card can already take full advantage of the bandwidth of an IDE controller. The interface is also ubiquitous. Why go through the trouble of using SCSI instead?

Why go to the trouble of using a PCI IDE controller when the motherboard already has one built in?

A controller can be faster than what is supported by the onboard IDE and the ROM can work around 8GB and 32GB limitations on disk or partition size.

An older controller may even be desirable on a newer motherboard when running Windows 9x or DOS if no UDMA drivers are available for these older operating systems.

Or you can have a lovely M919 or M571 or M537 that barfs the onboard IDE when you overclock it crazy high... (What's that, get an Asus, because they'll prevent you from doing that?) 🤣

I've used a SIL0680 IDE card before. They should work just fine. It's advised to use the non-RAID one if you don't need RAID and also need to connect optical drives with it.

However, I'm not really a fan of adding such cards in systems that I intend to run DOS on, as their option ROM can take a good amount of UMB.

A problem would be that PCI bandwidth is limited and shared among all devices on that bus so you might not be able to reach optimal performance in some scenarios.

I haven't tested this card yet but have instead been doing research to find the latest BIOS and software that was available. Silicon Image used to provide links for these things on their website but it no longer exists. And Lattice Semiconductor does not provide these files in its legacy archive.

I found the BIOS page on the Wayback Machine for this controller on Silicon Image's website but the latest BIOS version 3.4.0 was not archived.

https://web.archive.org/web/20100328145843/ht … x?pid=31&cat=15

It took some digging but I managed to find it and have attached it here for anyone who may need it.

This is the latest non-RAID BIOS. For the latest RAID BIOS download version 3.2.20 as linked by the Wayback Machine.

canthearu wrote on 2021-09-29, 02:19:

For the moment, I tend to stick to the onboard controllers (as they have better bandwidth to the chipset) and would use SATA to IDE converters where needed.

If the SATA/IDE bridge isn't Marvell or one of the "blessed" Silicon Image species, then a proper PCI card would outperform them easily. And that's not counting Jmicron and Sunplus trash can be a source of compatibility and reliability headache.

I wince whenever I see the CMD logo on an IDE controller chip. CMD0640B was one of the most notorious IDE corruptors of its day. Those, and PC Tech RZ1000.

The Silicon Image chips are allegedly descended from the CMD chips, but supposedly they don't have those bugs?

I'm getting excellent speed for this controller in Windows ME with it managing over 160MB/s reads from my fastest SanDisk Extreme Pro CompactFlash card.

The only issue I'm having so far is that the system hangs during shutdown and must be forcibly turned off. This triggers ScanDisk on reboot. What a waste.

Edit:

I was using PCMark2000 for disk benchmarking but I'm getting different results with ATTO Disk Benchmark which shows a maximum of about 20MB/s read/write.

I also received a blue screen indicating a Windows Protection Error on shutdown after some benchmarking. This is not reassuring.