Great article! I looked at the IDE SSDs and when I saw the price I decided to go down the CF route as the adapters are cheap (they dont do a lot reallt as CF are in effect IDE) and got a load of random CF cards from a second hand shop, it was pot luck what brand you got but most were Kingston and the odd one was a random brand like Jessops. So far the couple of laptops I have put them in are going strong! 😀

My experience with measuring max though put on solid state ATA devices (CF, DOM, and SSD) in old hardware is that the max through put is often (usually?) limited by the ATA protocol negotiated between the controller and the storage device. HWiNFO for dos is a good quick way to determine what ATA / UDMA level the device is operating at. I found that contemporary storage devices did not always negotiate the best speed with legacy IDE controllers and that it is not unusual to see a device negotiate slower speeds with different controllers even if everything is rated at a faster ATA speed. Using an add in PCI controller or a different motherboard can significantly change the max throughput of the storage device if it allows for a faster protocol. Some controllers don’t recognize DOM devices as 80 pin cable connected and negotiate a slower speed.

Predicting max through put for SD and USB storage devices on an IDE bus is different because they are not native ATA devices and have a bridge chip. The sintech bridge was frequently the weak link when measuring max SD throughput on socket 7 boards. It looks like the bridge is limited to 25mbs regardless of what SD card I used. Protocol negotiation between the bridge chip and the legacy controller was also still an issue in some cases but it was more a compatibility problem than a performance problem. I frequently had trouble getting the sintech to negotiate back to ATA-0 on old 486 systems. The SD adapter would not work in those cases.

Sata/pata bridges were almost always limited by the negotiated ATA protocol.

The small random transfers are a different story and probably have a bigger impact on Windows performance. Device buffer & speed, writing algorithms and flash speed probably factor into this along with protocol negotiation.

I think the devices that have slower writes than reads would be the result of the type of flash used, yes?

Compact flash card is solid-state device that in just a smaller factor than SSD.

It has the same architecture with PATA interface with different connector, a controller and one or more flash chips:

The performance only depends on the controller used and number or speed of chips. So SSD can be faster or CF, depending on what's inside.

Given the PCB of PATA SSD can be bigger, sometimes SSD have more chips which somewhat improve durability and performance. But often especially when comparing smaller SSD drives there is no practical difference and they often share the same parts inside and even the PCB size isn't that much different.