Re: DMA or not DMA...

[nathan@visi.com (Nathan Mates)]

In article <10JUN199718075585@vax2.concordia.ca>,
Mitchell Spector <spec@vax2.concordia.ca> wrote:
>    Hmmm, I just tried it now (going into the RamFAST's 'options'
>menu and toggling DMA to 'no') and I don't see any difference in
>performance from the numbers I originally posted. Does that option
>really disable DMA?

   There was a 25% speed hit on my system (Zip 9/32, 256K Rev C
Ramfast ROM 3.00j or 3.00l) when I tested. I always made sure to power
down my GS after changing this setting and saving it, as this is quite
possibly something that's read in on Ramfast powerup and not designed
to be changed on the fly. Try that to see if the powerdown affects
your numbers. [Rule 1 of a scientist: isolate off as many variables
as possible.]

>    The DMA issue aside for a moment (but for the record, I do _not_
>think DMA is an overrated feature)  I think an accelerator makes a
>world of difference to hardrive performance. 

   So if a 9/32 (probably about average for accelerated GSs; there are
lots of people who haven't upgraded) gets maybe 25% slowdown w/o DMA,
and 12+ a 5% (number unverified), what about all the thousands of GS
users without accelerators? I'd expect that at most 25% of GSs have
accelerators (and I'm being generous there), so for the majority of
the populace, DMA matters a GREAT deal.

   Hmmm. Anyone have any hard stats on what GSs are actually running?
There's far too many old ROM 00/01s with 256 or 512K floating around
from schools simply because Apple couldn't bother selling a computer
beefy enough to do lots of stuff.

Nathan Mates

--
<*> Nathan Mates http://www.visi.com/~nathan/      <*>
# What are the facts? Again and again and again-- what are the _facts_?
# Shun wishful thinking, avoid opinion, care not what the neighbors
# think-- what are the facts, and to how many decimal places?  -R.A. Heinlein