Apologies if I've confused everyone by posting disk benchmarks from two different computers in two different threads at the same time. Here's the final benchmark from my Performa-6300-in-6214CD-case, a 40GB 7200 RPM Seagate IDE drive:
Conner 1.2 GB 3600 rpm IDE: sustained read: 1789 KB/sec, sustained write: 2179 KB/sec, average access time 24.1 ms
IDE-to-SD with class 10 card: sustained read: 1720 KB/sec, sustained write: 1528 KB/sec, average access time 1.2 ms
IDE-to-SATA with 750 GB SSD: sustained read: 2408 KB/sec, sustained write: 2822 KB/sec, average access time 1.1 ms
Seagate Barracuda 40GB 7200 rpm IDE: sustained read: 2463 KB/sec, sustained write: 2622 KB/sec, average access time 10.5 ms
This 7200 rpm Barracuda drive performed well, roughly matching the throughput of the IDE-to-SATA adapter with SSD, although its access time is slower. I suspect both this drive and the SSD are maxing out the 6300's IDE interface, and roughly 2.5MB/sec to 3.0MB/sec is as fast as it's possible to go.
So which of these four options should be my final choice? It won't be the slow Conner drive, and not the SSD either since I plan to keep that for my G4. I'm leaning towards the IDE-to-SD. Even though the Barracuda has faster throughput, the SD card has faster access time and appears to outperform for small transfer sizes. Using the SD card also makes it easy to transfer new software onto this Mac. I just need to figure out the cable situation so I can cram it into the drive bay!
Commentary: I'm still not sure how to interpret HDT's numbers, or which numbers are most important. Looking at the sustained read and write numbers I listed above, you'd think the IDE-to-SD adapter was even worse than the original Conner drive, but that's definitely not true. The computer starts up nearly twice as fast with the SD adapter. So if startup time is your goal, it's probably more important to focus on access time and transfer speed for small blocks.