First post, by douglar
- Rank
- Oldbie
I'm trying to put together some DOS storage benchmarks and I wanted to throw some thoughts into a post so that I can get feedback from other Vogons members.
First, it seems important to have an easy to use benchmark took that gives consistent results.
Looking at available software, I see three contenders:
Speedsys (https://www.philscomputerlab.com/dos-benchmark-pack.html)
- Directly sends commands INT13h, bypassing the file system {edited}
- Often identifies the difference between SSD and Disk storage and does different seek tests
- Some BIOS extensions seem to confuse it and it defaults to the Disk tests ( XTIDE UBE ) {edit} when the rom is not shadowed
- Some devices report unexpectedly high values on the linear verify test, suggesting that they are not actually verifying the sectors and just immediately returning "success'
- Doesn't work with my Atronics caching controller. Always reports 0MB/s.
- Samples the entire storage device
- The write test is destructive
- Requires a 386
- Has some command line scripting functionality
DiskMBPS & DiskRand (http://www.roylongbottom.org.uk/oldones.htm#anchorDisk)
- Uses the filesystem, so it is affected by BIOS speed, DOS buffers, and anything that hooks INT21h like smartdrv.exe
- Samples inside a single file
- Requires a 386
- Very scriptable and configurable
DOS Disk Tester (https://www.lo-tech.co.uk/wiki/DOS_Disk_Tester)
- Uses the filesystem, so it is affected by BIOS speed, DOS buffers, and anything that hooks INT21h like smartdrv.exe
- Simple and works on 8088, DOS >= 2.0
- Scriptable
I'm leaning toward DOS Disk Tester because it is pretty straight forward. Hopefully the results are pretty consistent over time.
Here's the items that I see affecting the storage stack:
INT21h & INT13h
- Partition size, FAT type, & DOS buffers
- BIOS & BIOS Extension performance (Optimized? RAM shadowed?)
- Drive Overly Performance
- DOS Driver - Does your storage device need a device driver to reach max performance
- Disk caches like smartdrv.exe
System Speed
- CPU - In the days before bus mastering, CPU speed was important for I/O throughput
- Bus Speed - a 12Mhz ISA bus is faster than 8Mhz. A 50Mhz VLB should be faster than a 33Mhz.
- PCI Bridge - systems that used a bridge chip were often much slower
- Hard drive controller - Once we get into >= VLB, there is a controller chip that affects performance
ATA Protocol
- Is the storage device using the best ATA protocol?
- Is there a slow device on the same cable slowing you down?
Storage device
- Storage devices don't always perform the same over time
- https://pcpartpicker.com/benchmarks/methodolo … we-test-and-why
