Hello,

Thanks for replies to all three of you guys :) Let me clarify few things
that might have been unclear:

The box is indeed m38 in BA23A with m76 label on it; m76 died (?) and
m38 was put inside that box.

> The VS3100s came with varying front plates.  Some of them had cut-outs
> for internal 3.5" floppies and internal 5.25" CD-ROMs or tape drives.
> If you don't care so much about aesthetics, and you have the internal
> drives, you can always run the system without its cover.  Other options

I do care about aesthetics :) I tried not to damage the enclosure .. but
i really would like to use a CD-ROM, now i have to find this plastic
front panel thingy for CD-ROMs somewhere. As Antonio says it should be
trivial to replace it (i have to figure that out :-)) but i have to find
one of those first - anybody happens to have a trashed enclosure that
they can rip it out of and sell it for not-some-outrageous sum ? I need
the complete cover because the box is down on the floor with tons of
dust; extra hole wouldn't exactly help in keeping box tidy. Now that i
think about it i have yet to figure out how to place that box (now it;s
sitting in a corner, with back side facing me because that's where the
power switch and all the connectors i care about are - CD-ROM would be
facting the wall in that case. Perhaps i should go for external unit
instead ?

> > The only PCB that i can access inside the box is obviously SCSI
> > controlelr (it has NCR chip). It also has two geatures i'm not sure
> > what function they serve: first is a square chip, about the size of a
> > CPU with number starting with 21... on it; i believe DEC labells their
> > (CPU) chips this way; so what is this chip ?
> 
> Probably a dedicated I/O daughter card controller processor.

This board also has 4 identical chips, quite big ones (40+ pin DIPs),
they look like Z80 CPU and have NEC written on them. Are these some kind
of memory ships or what ? I can't think of any reason why a controller
this old would use extra memory chips.
 
> > Second thing is a two-row, 108 or so pin male connector (pins are
> > close together like those of 68-pin SCSI connector); what is this
> > connector used for ?
> 
> If it's what I think it is, there's a ribbon cable connected to it that
> goes down and under the steel plate; it connects to the system board.
> I'm not sure what the official DEC name for this is, but it's for the
> proprietary I/O bus between the system board and the I/O daughterboard.

Not this one - that one is smaller (shorter). I counted pins on this
thing and i believe it has 54 in 2 rows, making it 108 pins. It is much
closer to the centre of the board than that system bus cable you
mentioned. It looks like larger (longer) 68-pin SCSI connector, it even
has those restraining things at each end.
 
> for use with a flathead screwdriver).  With the I/O board mounted on
> the steel plate, you'll have to disconnect it from the I/O bus and

I must have missed the one screw near the PSU; when i dig the box out
again i'll look for it.
 
> > My HD is RZ26L-W, made by Quantum. I can't find any info about this
> > drive via Google (tried "rz26l w quantum"), just a bunch of resellers.
> > Does anybody have any docs for this drive ? I'm curious as to what 'SP
> > sync' jumper is used for (it's open now); write protection i figured
> > out and 'spn dly' must be spin delay or something (the only one
> > closed). Does that 'SP sync' jumper affect the speed of the drive ?

I found
http://perso.wanadoo.fr/rayp/HD/QUANTUM/general/jmpr_def_scsi.htm which
mentions "SPindle Sync", close enough to "SP Sync" i guess. But that
explanation is ... all gibberish to me - i'm not sure why one would want
to have two drives spindling synchronously anyway ? 

It also seems i have found the explanation for that "clonk" sound i
always hear after powering up the VS but while HD is slowly spinning up:
it must be head locking mechanism releasing the heads after drive spins
to operational rpm.
 
> The VAXstation 3100s have SCSI-1 interfaces.  The controllers and drives
> negotiate the best speed at they can both operate.  The 3100's
> controller is somewhat older than your SCSI-2 drive, so the drive
> t-a-l-k-s sl-ooooow to it so the controller will understand it.

Hmm .. so VS's controller neither speaks SCSI-2 or allows wide transfers
? That's .. bad :( Especially the wide part, it would have helped, right
?
 
> > says it's a 5400 RPM drive with 0.5 MB of cache (!) and transfer rate
> > of 32.8 MB/s (!!)
> 
> Even with the best SCSI controller in the universe, on the fastest
> Pentium-VII++ 20GHz system money can imagine, I'd say 32.8 MB/s
> throughput is bit optimistic for that drive.

Well yeah, i'm used of marketing lingo like that but that usually boils
down to 1/3 of that speed (which sounds about right, 10 MB/s for
SCSI-2). That is probably internal transfer rate anyway (between drive
and its cache/controller), not transfer rate on the bus (SCSI or
system).
 
> > but it feels slower than my old 130 MB 3600 or so rpm IDE in 386SX/25
> > i used to have.
> 
> Do remember that, in terms of CPU integer performance, the VAXstation
> 3100/m38 probably ranks somewhere in the same area as a 386SX/16 to 25.
> The CVAX+ it uses is clocked at 16.67MHz (60ns cycle time).

Umm, but that was 14 ms average access time ATA drive with 32 KB or so
cache that clogged up the CPU during data transfers in its PIO 0 mode
and this is 10 ms 512 KB chace drive with SCSI controller doing DMA
transfers. 
 
> There are quite a few benchmarks available in the benchmark section of
> pkgsrc.  You can also use `dd' to test sequential access performance
> against the raw disk devices.  Here are some numbers from my MicroVAX
> 3100/20e, using various block sizes for sequential reads:

OK, this is what i got using same block size / block count:
 
>   % dd if=/dev/rsd2c of=/dev/null bs=4k count=4k
>   16777216 bytes transferred in 36.040 secs (465516 bytes/sec)

28.96 seconds (579323 bytes/sec)
 
>   % dd if=/dev/rsd2c of=/dev/null bs=8k count=4k
>   33554432 bytes transferred in 52.510 secs (639010 bytes/sec)

43.38 seconds (773500 bytes/sec)
 
>   % dd if=/dev/rsd2c of=/dev/null bs=16k count=4k
>   67108864 bytes transferred in 85.811 secs (782054 bytes/sec)

72.51 seconds (925511 bytes/sec)
 
>   % dd if=/dev/rsd2c of=/dev/null bs=1024k count=64
>   67108864 bytes transferred in 68.620 secs (977978 bytes/sec)

60.16 seconds (1115135 bytes/sec)

Hmm, read speed isn't that bad afterall (it's not like the box can
process 0.5-1 MB/s for any practical purposes (such as kernel
compilation). Have i really lost the perception of speed ? My PCs aren't
1000x times faster than this VS, one is ~13 times faster BogoMIPS (VPUS
:) wise, one is even slower than that and one is running Windoze XP so
even trivial things like moving files from one location to another on
same HD takes seconds sometimes. Apart from sluggish console/telnet
output and time required to load make it's not bad at all :) 
 
> 1MB block size.  Similar `dd' tests can be done for write speeds, as
> long as you don't mind trashing the contents of the partition you're
> testing against. The tech-perform mailing list will be able to offer you

I do mind trashing them, it took me ages to customize this thing to my
liking :) But since it's write-to-filesystem what i'm more concerned
about i could have it write to a file. 

dd if=/dev/rsd0c of=testfile bs=4k count=4k
74.77 seconds 

now if read took just short of 30 seconds that means write takes ~45
seconds, 50% longer. Now that i think about it, that doesn't sound so
bad either. What is kUmem device all about ? Raw memory access device ?
It would be better for comparisons than hd-to-hd operation.
 
> I've had good luck with old Apple Mac CD-ROM drives and Plextor drives.
> I think nearly at SCSI CD-ROM that supports 512 byte sectors should
> work fine for you.

OK, with HD transfer rates up to 1 MB/sec i guess anything better than 1
Euro (US$1.11) 6x speed drives off eBay are a waste of money ? :)

Another thing that just popped to my mind: VS4k has Turbochannel bus,
right ? Does that mean one can put another network interface into it so
it can function as a router ?

Blaz Antonic