[comp.sys.dec] Iverson 11/750 accelerator

roy@phri.UUCP (04/07/87)

	The latest issue of Hardcopy (A DEC-oriented trade rag) has an
interesting item.  A company called Iverson, Inc in Fremont, CA is selling
a card that plugs into the back of a VAX-11/750 backplane and supposedly
speeds up the machine by 15-25%.  They say "System speed is improved by a
new clock that pulses variably, according to the exact time required for
each instruction, rather than a fixed pulse that is standard on the VAX
750s".  The card is supposed to be completely transparant.

	Has anybody used one of these?  Do they actually work?  I don't
have a price, but if it's just a few $k, it might be worth it if it really
does get 25% (or even 15%) more performance out of the machine.  On the
other hand, the idea of plugging in some random board that purports to play
with (I'm guessing now) the microsequencer clock sounds kind of scary.

	Next question: why would somebody invest good engineering time to
make a product which is specific to an 11/750, a machine which isn't made
any more and will probably be obsolete in a couple more years?
-- 
Roy Smith, {allegra,cmcl2,philabs}!phri!roy
System Administrator, Public Health Research Institute
455 First Avenue, New York, NY 10016

"you can't spell deoxyribonucleic without unix!"

kellym@tektools.TEK.COM (Kelly McConnell) (04/16/87)

In article <2637@phri.UUCP> roy@phri.UUCP (Roy Smith) writes:
>
>a card that plugs into the back of a VAX-11/750 backplane and supposedly
>speeds up the machine by 15-25%.  They say "System speed is improved by a
>new clock that pulses variably, according to the exact time required for
>each instruction, rather than a fixed pulse that is standard on the VAX
>750s".  The card is supposed to be completely transparant.
>
>	Has anybody used one of these?  Do they actually work?  I don't
>have a price, but if it's just a few $k, it might be worth it if it really
>does get 25% (or even 15%) more performance out of the machine.  On the
>other hand, the idea of plugging in some random board that purports to play
>with (I'm guessing now) the microsequencer clock sounds kind of scary.
>

We had a loaner board from Iverson for evaluation and installed it on 2
different 750's running VMS and 2 more 750's running UNIX.  The UNIX 
support group here ran some benchmarks and found that the board really did
deliver 15-25% increase in speed (depending on the test).  The board is
indeed transparent. If you suspect the board is causing a problem you just
flip a switch on the board which effectively removes it from the system.
What happens when the board is turned on is that the clock is sped up for 
selected micro words.  We saw no indication (other than an increase in speed)
that the board was installed on any of the 4 machines we tested.  Installation
was very easy and very well documented.  I think the price was around $9K.
That seems a little steep to me but if you can't afford a bigger machine it
may well be worth it.