Re: MicroVAX 3100/20 is apocalyptically slow even with 32MB of RAM

[David Brownlee <abs%netbsd.org@localhost>]

On Mon, 20 Apr 2026 at 16:28, Daniel Seagraves <dseagrav%lunar-tokyo.net@localhost> wrote:
>
> > On Apr 17, 2026, at 4:03 PM, John Klos <john%klos.com@localhost> wrote:
> >
> >> I have a MicroVAX 3100/20 with 32MB of RAM that once ran NetBSD many years ago. I recently had occasion to replace its disk and install 10.1, and it is superlatively slower than I remember it being. It?s so slow that it?s been trying to just bootstrap pkgsrc for about a week now, with pauses to allow the daily cron script to complete otherwise it won?t complete in less than 24 hours. The system time loses about 10-15 minutes a day, presumably because it can?t keep up with the clock interrupts. This is after I copied the GENERIC kernel config, stripped out all of the VAX models and devices that weren?t needed, and stripped out all of the RAM-based filesystems.
> >
> > I have a VAXstation 4000/30 that has a defective CPU cache. Because of this, it benchmarks quite similarly to a 5 MHz 11/780, even though it runs at 25 MHz. I wonder if your cache isn't getting turned on correctly, or if your cache is failing...
>
> I’ll have to look at the test results next time it powers up. I don’t think it flagged any hardware faults, and I’ve even replaced the nvram battery with a modern substitute.
>
> >> I don?t think RAM is the issue because I haven?t seen it go above 12-14MB active in top. It uses more swap than it does RAM.
> >>
> >> Is there something configuration-wise I should try, or is this model just too slow to be supported anymore?
> >
> > Your machine should be at least half as fast as my fully working VAXstation VLC. I'm running NetBSD 11, and it really isn't any slower than NetBSD 9 or 10. Do you know what version of NetBSD you had on it many years ago?
>
> I want to say NetBSD 4 or 5? Maybe earlier? It was around the time that SCSI DMA was fixed on the 3100s and we didn’t have to use PIO mode anymore.
>

One aspect which has changed massively has been gcc, and in particular
the memory typically used to build anything. Something which would
have (relatively) zipped through on earlier versions of gcc on vax now
page a system to death.

If you run "vmstat 5" during the build does it show any significant
numbers in the "po" column?

It would be interesting to see benchmarks to see how the performance
has changed (if you have a spare disk onto which you could install
NetBSD-4 or 5), though you would probably also want to benchmark
something which does not use gcc - maybe a complex awk script
processing a reasonably sized file?

David