Random change WTT: KA694

From: Chad Fernandez <fernande_at_internet1.net>
Date: Sun Jun 24 16:12:03 2001

I see that explains a lot. I'll have to make a new mail folder for my
"Vax info" :-)

Chad

Chuck McManis wrote:
> Hi Chad,
>
> In the 4000/300 DEC decided to make a "new kind" of Q-bus box, one where
> the CPU was upgradable again.
>
> In the early MicroVAXen the CPU was a single card, memory was a single
> card(s), disk interfaces, and then ethernet. Since the Q-bus is limited to
> about 3Mbytes/sec that set a limit on the I/O performance of the machine.
> With the KA640 DEC put ethernet, disk I/O (DSSI), and memory on the same
> card. This gave them a direct disk channel (4Mbytes/second), ethernet not
> on the bus (saves interrupts and bus latency) and some local memory. I did
> not work as well as they would have hoped apparently. But the ultimate
> expression of this idea was the KA660 which used Q-bus "fingers" had only
> DSSI and Ethernet on the CPU card and pushed the memory bus that ran over a
> 50 pin ribbon cable to its limit.
>
> For the 4000/300 they knew 64bit wide memory was going to help them get
> performance gains and they knew that leaving the processor on the Qbus
> meant they had to connect things with ribbon cables that could have easily
> (and less expensively) been connected on the backplane. So a new bus was
> invented (I don't know if it is named or not) and the "CPU Box" was born.
> This borrowed from the mainframe VAXen of the time that had the
> CPU/memory/bus interconnects in one box, and then either BI, XMI, or even
> UNIBUS boxes hanging off of that. In the 4000/300, 400, 500, 600, 700,
> 700a, 705, and 705a. The backplane consists of about 5 "slots" on the right
> edge that hold the CPU / memory / DSSI interconnects and to the left of the
> CPU is the Q-bus. The KA670,-675,-680,-690,-691,-692,-693,and -694 CPUs use
> this bus. (note they could not go to KA7xx because that would interfere
> with the module numbers of the 700 series processors :-) Everything from
> the KA675 were variations on the NVAX implementation of the VAX
> architecture (as is the 4000/90, 4000/105, etc)
>
> So you're correct in that the back plane connector on the KA694 looks
> almost like a eurocard connector but it actually plugs into a backplane
> that has Q-bus slots to the left of the CPU.
>
> So the KA660 is the fastest "true Q-bus" processor and the KA694 is the
> fastest "integrated Q-bus" processor. The 4000/105 has a desktop enclosure
> and actually brings the Qbus out to a pair of connectors on the back of the
> box, These connect with cables to a B215F or B213F expansion box where the
> first card in that box is the Q-bus extender. I don't consider the 105 to
> be an integrated Q-bus processor like I do the 4000/705a.
>
> --Chuck
Received on Sun Jun 24 2001 - 16:12:03 BST