[johnnym]

The upcoming release of OpenBSD 7.3 is still some weeks away - it should appear around end of March IIC - and I also wanted some diversion from OpenBSD/sgi, so I did show my other machines some love in the meantime and tested NetBSD 9.3 on them. I tested a lot of machines and configurations so far for diskless operation and will list them here, starting with my SPARC machines today:

• NetBSD/sparc
  • Sun SPARCclassic w/microSPARC running at 50 MHz and 80 MiB of RAM - actually this one here:
    
    
  • Sun SPARCstation 10 w/2 x SuperSPARC SM61 running at 60 MHz and 512 MiB of RAM
    
  • Sun SPARCstation 10 w/1 x HyperSPARC HM150S-512 running at 150 MHz and 144 MiB of RAM
    
  • Sun SPARCstation 20 w/2 x HyperSPARC HM150S-512 running at 150 MHz and 288 MiB of RAM
    
  • Sun SPARCstation 20 (same as above) w/1 x HyperSPARC HM100W-256 (two processors on one MBus module) running at 100 MHz and 288 MiB of RAM
    
  • Axil 311 w/2x SuperSPARC SM50 running at EDIT: 50 40 MHz and 368 MiB of RAM
    
  
  All of them worked well and stable - even swapping to a NFS swap file didn't bother them - but overall the *SPARC driven machines felt slow, even the SPARCstation 20 when using two HM150S-512 (see The Rough Guide to MBus Modules for details about the nomenclature). The ones w/o L2 cache were exceptionally slow, they even take much longer than the other machines to actually load the kernel over NFS during bootup. I didn't test though if this would improve with a 100 Mbps NIC. But if you imagine that these ran GUIs in the 90ies...
  
  I took the opportunity and also ran some benchmarks on all of them (with 7za and openssl). I plan to gather the results in a usable form in the future (thinking about R scripting and plotting) to avoid spreadsheet software.
  

[johnnym]

I had to correct the above post as I recalled that the SM50s were actually running at 40 MHz only in the Axil 311, as this one is also limited to 40 MHz on the MBus like the SPARCstation 10. So the SM50s are 20 % short of their clock rate in those machines. I also tried them in my SS20, but they didn't work correctly there, most likely due to the ROSS 2.25R firmware as hinted also on the Rough Guide to MBus Modules.

All of those machines have a battery holder and CR2032 battery attached to their NVRAM chips. Years ago when I did that modification I didn't disconnect the original batteries to save time, work and dust but just soldered the battery holder in parallel to their terminals. And I'm happy that this didn't bite me in my ass so far. Most of them still had the correct date since I modified them. Also the NVRAM contents were kept safe (incl. IDPROM) for all of them. As it's so easy to reprogram the IDPROM with fw-control I won't bother to replace those batteries in the future. There's one catch though, I don't know of a way to set time and date from the OBP, so this always has to be done in the OS.