Re: Dedicated CAD computer?

On Sun, 16 Jul 2006 04:30:31 GMT, "Brad Velander"
<bveland@xxxxxxxxxxxx> wrote:

Howdy Mike,
So you are liking AD6.

One suggestion regarding your faults/GPF faults. Years ago any number of
P99SE users were getting numbers of crashes and faults. After much
discussion, trials and testing, the problems were generally eliminated by
one of two things. Rigorous memory testing (long periods of in-depth pattern
testing) showed intermittent faults. Also video cards were the cause of a
lot of the faults. The guys either updated video drivers or bought other
video cards. Those two solutions have always seemed to relieve the problems
for most Altium/Protel users.

To assist in your quest you could check the Altium user forum, I know
that not that long back (maybe 3 or 4 weeks max) there was a similar
discussion when someone asked the same question regarding what was a good
machine for running AD6. I believe that a significant number of the
responses were running Nvidia Quattro 4400 video cards (don't quote me on
that though).

As Stephan mentioned, don't bother with a dual core processor because
AD6 definitely doesn't use it and probably never will. Some users were even
turning off their dual core capabilities through their Bios and claiming
improvements in speed. I don't know if that makes sense or not, but that is
what they claimed.

It actually does make sense. When Dual Core is enabled there is
additional overhead required to distribute the workload across the two
cores, whichever way you slice it and dice it, this requires
additional CPU time.

Now if you have multiple tasks running in parallel then this overhead
is irrelevant because the benefits by far outweigh it.

But in case of a single threaded app the other core sits there idle
while you still have the overhead to check and see if there is
something that the other core could do (and the answer is always no).
So this time ends up wasted slowing the app down.

Sadly from a programming perspective, multithreading is "difficult" to
implement for someone that's been writing single threaded all their
life. There are many things that need to be taken into consideration
that you just don't need to worry about when writing single threaded
apps (deadlocks for example) and it's easy to make a mistake.

In addition to that algorithms also need to be written very
differently to be able to even take advantage of multi threading.

My EDA app actually is multi threaded where it makes sense to do so
such as my rendering loop for PCB Boards. The rendering loop is setup
to distribute the workload across all available CPUs (in case of
hyperthreading, dual core, or multiple CPUs) and both threads load the
GPU with video data in parallel as well. Allows me to render a board
with almost 20,000 objects (lines, vias, pads, text, etc.) including
all necessary triangulation resulting in almost 100k triangles in less
than 90ms. Layer sorting included (done via hardware z-buffering).

I only worry about this though to that extend for the PCB rendering
pipeline. Pattern editor, symbol editor, and schematic capture all run
single threaded as I've yet to come across a data set that requires me
to get that fancy =)

--
Stephan
2003 Yamaha R6

kimi no koto omoidasu hi
nante nai no wa
kimi no koto wasureta toki ga nai kara
.