🚀 go-pugleaf

In article <830802086snz@cdsl.demon.co.uk>, Peter Hayward
<peter@cdsl.demon.co.uk> writes
>   I'm sorry if this is a duplicate but the original, posted on Sunday,
>   seems to be "one of the small number of news articles lost by demon".

I saw the original.


>   I own an ISA caching disk controller fitted with 4MB of 30 pin simms
>   (well what use are 30 pin simms these days?) but I'm thinking of
>   getting a new motherboard and CPU that is likely to have it's own on
>   board (E)IDE controller.  The question is:-
>
>   Would I better off keeping my caching contoller or moving to the on
>   board (E)IDE?
>
>   My gut feeling is that keeping the cache controller is the better
>   option.  I appreciate that the ISA card will have a much lower data
>   transfer rate, say 5 times slower than the on board controller.  But
>   the data transfer portion of a block read is modest compared with
>   rotational latencies and even more so with seek times.  It's in the
>   last two area's that the cache will show gains, but will that balance
>   out the lower transfer rate?
>
>   Any comments?

I didn't comment originally because I didn't have a simple answer.

My gut feeling is the EIDE controler on my P/133 board Ps all over the
DC200 because it can sustain higher throughput. I wonder too if there is
some latency caused by the extra administrative layers of cache control
in the DC200 ?

My experience of the DC200 has not on the whole been good. It shat up
mightily on one vl bus machine, it would corrupt drives on that system
after any soft reset too.

In another machine it exhibited poor comms performance as its drivers
leave IRQs disable for long periods.

In addition, its RAM test is pretty useless at detecting RAM
incompatabilities. When it boots with incompatible RAM, it craps on your
data silently.

I guess the real performance answer will be down to mode of use. Do you
normally write and reread small files that fit the cache or do you write
vast objects to an unfragmented file system ?  Are the alternative
controlers bus mastering ?  If so, can your os run other processes while
waiting for IO to complete ?


Either way, my spare DC200 is now on its way to you with 4 256k simms in
it which ought to work but I've not tested them !  Beware.


Cheers, J/.
-- 
John Beardmore