< The real problem was the controller - RQDX series controllers were so
< slow one sometime wondered if the damn thing would ever return a block
< to the os at all...

Retrorevisionism, sigh.  the RQDXn series also represented the state of 
the art and were many times faster (and smarter) than the PC HDCs. Also
the RQDX1/2 were slower than the RQDX3.  But it also saved the OS from 
doing bad block management.

In fairness to the RQDX controllers they are an older design dataing back 
to 1983! They were introduced as part of the MicroPDP-11 systems and for 
1983 were lightyears ahead of the really pokey xybec controllers on pcs
that could barely manage 90k bytes/sec.  the RQDX3 was introduced around
'86-87 and was starting to show it's limits but was a cost/size reduction
while remaining backward compatability.

Also I see people using RD51 and RD3x drives wich are VERY slow compared 
to the RD52, 53 and 54.

< They could have made those controllers much faster, but I suspect not s
< for marketing reasons ;-(...

Why?  Is there any reason to support this?  Keep in mind DEC also sold 
controllers for RAxx drive for Q-bus.  Q-bus devices were slower because 
Qbus is slower than (pick your favorite bus).

Think of where 1986 was... IDE was still an infant.  SCSI/integrated 
controller drives werent quite there and would be expensive as all get
out early on.  EDSI was not yet, SMD and other high end technologies
were either BIG, new or EXPENSIVE.  As an engineer designing for the 
future in 1985 consider whats available, consider the projected price 
point (cost) and manufacturability.

< Just like the RLV12 and RLxx series drives - non intelligent, simple

the RL drive were not very intelligent but the state machines in them were
complex enough to make up for that lack of cpu.

< interface that subjectively blew the pants off RQDX.. controller drive
< combinations, even with the slow access time of the RLxx drives.

Since I run RLV12/RL02 and RQDX3/rd52 in the same 11/73 I can say while 
the RL seems a bit faster for trivial accesses, when it come to really 
stirring the disk it looses every time.  the cpu overhead for RL is higher 
and it moves the head slower.

A similar RL and RQDX3/rd54 combo were on a MVII at one time and even 
under vms it's a noticeable difference in favor of the RQDX.  RLs are 
nice but by the book they are slower.

FYI: by the book RD54 drive is 38ms against 15ms for the RL track to 
track. But the RL can only hold 20480 bytes a cylinder.  RD54 gives you 
130560 bytes before you have to move the head, making it appear about 6x 
faster.  Also the RD54 spins at 3600rpm against the RL02s 2400 meaning
the data you want will come around sooner.  The RD31/2 or RD51 performance
compared to RL however is almost the reverse.  Then again it's never been 
fair to compare drives with voicecoil positioners aganst steppers.

I'd say for over a decade old the hardware works pretty good.  For a 
design that has elements closer to 12-15 years old it's performance over 
time is impressive.  Never forget the MVII was built around mostly 
standard PDP-11 components and was trying to dislodge the PDP-11 for 
some markets with backward hardware compatability and a 32bit upgrade 
path.


Allison