3.5″ Floppy disk drive benchmarking (WHL #57)

Yeah, you read that right. Time for some serious benchmarking of good old floppy drives. The type of benchmark that the interwebs have been waiting for.

In case you’re questioning my mental health right now: Nope, not affected by COVID-19 or the “quarantine” (that only keeps me from delivering a 2U server and causes people to act strange in the supermarkets). I’ve gone mad quite a long time ago. Probably a glitch during assembly.

Now, I got four bad boys here, revving their, umm, stepper drives?
Alps Electric DF354H(121G), 33wk 2007 – full size 3.5″ drive (5V 1000mA)
IBM FRU 06P5223 ASM 08K9835, Feb 2004, USB drive (5V 500mA)
Sony MPF820, Dec 2005, Slim drive (5V 750mA)
Teac FD-05HF, May (?) 2008, Slim drive with unknown power consumption

The Alps was recently discarded by a colleague and I was out of drives, so he gave it to me. Who needs floppy drives in 2020 anyway. Alps was/is a major player with disk drives (like, the motors) and does a lot of potentiometer stuff nowadays.

The IBM was bought off eBay quite a while ago when I captured the family VHS cassettes (and 3.5″ disks and CDs). IBM forced the floppy for its personal computer, starting from the invention of 8″ drives to 5.25″ and 3.5″.

The Sony is an IBM branded device from a recently salvaged server (x346 I think, probably too old for a 3550 M1 or 3650 M1 X-series that was also tossed). Sony made the 3.5″ standard with mass-producing it for the original Apple Macintosh.

And the Teac is from the old dumpster-dived cousin chassis of my current ZFS server. Older style Supermicro case with floppy drive and RS232 in the front. I did modify that like two weeks ago while I was also stripping the other IBMs, HPs and Supermicros that I was recently donated. Teac was a major player in the upscale floppy and CD drive market and still makes professional audio gear.

When I was donating quite a lot of parts from those servers to random guys in German hardware forums, I also asked if anyone wanted one of my old server boards, Supermicro X7DBE and X8DTL-iF (as recently diagnosed with high battery drain!). I was going to keep one, and random.org decided it was the much less asked for X7DBE. Both were bundles with board, CPU, RAM, heat sinks and I/O shields.
So I shipped the X8DTL and rebuilt the X7DBE onto the motherboard tray from the scrapped Supermicro case. Nice board with 8x4GB FB-DDR2, dual-E5420 Harpertowns (Quad 2.5 GHz, Penryn 45nm, SSE4.1!), IPMI via card in a dedicated slot, three PCIe slots and three PCI-X slots. It also carries SATA and PATA ports (unfortunately no U320 SCSI which is a real bummer – that would have been the X7DB8 variant). About 200W of idle power draw, probably half of that by the DRAM. And this unit has a floppy drive connector! (fans moved to the back just for the photo, laying on top of RAM and heat sinks during tests)

So naturally I tested the newly salvaged two slim drives and the previously just stocked full-sized drive. I can’t stand having untested hardware on the shelves.

Long story short: They internal drives either all work basically at the same speed, or they are limited by the controller. It’s a ridiculous 38KB/s max speed with some reproducible dip at medium block sizes. The external drive however has other limitations and is slower in general, and especially slow with low block sizes (<5 IOPS). All drives offer the same speed at reading and writing, which I never knew until today. Read benchmark:

Write benchmark:

Unfortunately ATTO doesn’t really scale the results graph, the lowest setting has 100MB/s full scale. One can convert to IOPS which does show some bars, but that’s a tad boring as speed seems limited by other factors. It’s not like an SSD benchmark that really shows controller/flash limits – it’s always around 30KB/s.

Example screen:

AS-SSD doesn’t allow for floppy disk drives to be tested. Others have troubles as well, likely caused by the lack of free disk space for some 1GB++ of test data that is usually required to make SSDs cry.

Averaging reads and writes over the entire block size range, the results are as follows:
1. Sony: 32.49 KB/s
2. Alps: 32.43 KB/s (99.8%)
3. Teac: 32.21 KB/s (99.1%)
4. IBM 20.74 KB/s (63.8%)

Conclusion: All floppy disk drives are the same – except for the USB ones, those are terrible. But unless your desktop computer is from the stone age (let alone your laptop or even tablet), it’s also the only way to read 40 year old disks that might still be in use in your 80s/90s audio gear, CNC machines, or the local nuclear bomb guidance system.


Leave a Reply

Your email address will not be published. Required fields are marked *

:mrgreen: 
:neutral: 
:twisted: 
:arrow: 
:shock: 
:smile: 
:???: 
:cool: 
:evil: 
:grin: 
:idea: 
:oops: 
:razz: 
:roll: 
;-) 
:cry: 
:eek: 
:lol: 
:mad: 
:sad: 
:suspect: 
:!: 
:?: 
:bye: 
:good: 
:negative: 
:scratch: 
:wacko: 
:yahoo: 
:heart: 
B-) 
:rose: 
:whistle: 
:yes: 
:cry2: 
:mail: 
:-(( 
:unsure: 
:wink: