I'm planning to buy a new computer, and its main task will be for
antenna simulations using NEC. The issues appear to be: Intel verses
AMD, single verses dual core, front side bus and RAM speed, and of
course processor speed. I would appreciate any guidance.

It would be interesting to compare computation speeds of various
computers for the same antenna. One possible antenna to use for
comparison is The Grid Yagi from my website:
http://home.comcast.net/~ross_anderson/GridYagi.htm
http://home.comcast.net/~ross_anderson/GridYagiD1.htm
My three year old Dell with a Pentium 4 2.8GHz with HT and 3GB DDR of
RAM at 400MHz using the Nec2dXS*.exe calculating machine gives:
- - - MATRIX TIMING - - -FILL= 12.469 SEC., FACTOR= 11.422 SEC.
I'm interested to see how this compares to more modern computers.

Ross W1HBQ

On 16 Oct 2006 15:43:35 -0700, "W1HBQ" wrote:

The Grid Yagi from my website:
http://home.comcast.net/~ross_anderson/GridYagi.htm
http://home.comcast.net/~ross_anderson/GridYagiD1.htm

Hi Ross,

Can't say much for computer speed, but the antenna is interesting. It
is certainly resource hungry to model and I can see why you want
speed.

73's
Richard Clark, KB7QHC

I haqve a Toshiba laptop with 3.2 gig processor which I use for antenna
work. I really can't see why I shoulde look for a speedier processor at
this time and really it is not that much faster than what you have, if
they had a processor that doubled the speed then I would be interested.
I use AO-pro which I suppose is a larger program for antennas than NEC4
and sometimes it takes two or three hours to complete a 15 element
array and with the processor going at full speed things get a bit hot
but then again the processor is not going all the time as one has to
print and set things up , so for the total project I see no gain in
getting a faster machine. Had a 5150 dell last year of the same speed
but oh so many problems with Dell and them refusing to give out repair
information that thousands of them were thrown away in anger. They have
been sued for a few million but if a failed one had been opened by
somebody other than them it can be put down to user abuse where they
should not have touched it and waited to see if they were going to be
sued. The lawyers by the way get 3 million to sell out all users. Dell
suffered a tremendous down turn on all sales last year because of bad
service in the face of widespread failures across the board on all
their products where they only allowed repaires by their own repair
places at an inflated price
with no details of repair for DIY.. Their products are produced by
multi suppliers in competition so failures are all over the spectrum
and difficult to isolate.To sum up don't buy a Dell!
Cheers
Art


W1HBQ wrote:
I'm planning to buy a new computer, and its main task will be for
antenna simulations using NEC. The issues appear to be: Intel verses
AMD, single verses dual core, front side bus and RAM speed, and of
course processor speed. I would appreciate any guidance.

It would be interesting to compare computation speeds of various
computers for the same antenna. One possible antenna to use for
comparison is The Grid Yagi from my website:
http://home.comcast.net/~ross_anderson/GridYagi.htm
http://home.comcast.net/~ross_anderson/GridYagiD1.htm
My three year old Dell with a Pentium 4 2.8GHz with HT and 3GB DDR of
RAM at 400MHz using the Nec2dXS*.exe calculating machine gives:
- - - MATRIX TIMING - - -FILL= 12.469 SEC., FACTOR= 11.422 SEC.
I'm interested to see how this compares to more modern computers.

Ross W1HBQ

It would be interesting to compare computation speeds of various
computers for the same antenna. One possible antenna to use for
comparison is The Grid Yagi from my website:
http://home.comcast.net/~ross_anderson/GridYagi.htm
http://home.comcast.net/~ross_anderson/GridYagiD1.htm
My three year old Dell with a Pentium 4 2.8GHz with HT and 3GB DDR of
RAM at 400MHz using the Nec2dXS*.exe calculating machine gives:
- - - MATRIX TIMING - - -FILL= 12.469 SEC., FACTOR= 11.422 SEC.
I'm interested to see how this compares to more modern computers.

Ross W1HBQ

FILL= 14.641 SEC., FACTOR= 16.734 SEC., RUN TIME = 31.641

On a 2.4 Ghz Pentium 4, single processor 512 Mb, Windows 2000, and Yes,
a Dell.

Arie.

FILL= 15.922 SEC., FACTOR= 19.156 SEC., RUN TIME = 35.406

Pentium 4, 2.8Ghz, 1 Gb (W2000, cpu running at only 50% during calc's,
Dell optiplex GX620)

It seems to be some kind of dual core processor system, because on
windows task manager
I see two cpu usage history graphs. So 2.4 Ghz single core seems to be
faster than 2.8 Ghz double core...

Arie.

Dear W1HBQ:

To minimize compatibility issues, I use a stand-alone computer that uses an
Intel mother-board, Intel chip set, and Windows 2000. The board has
temperature sensors that are accessible through a free Intel computer
program. On long runs, one sees the temperatures go way up with 100% CPU
use. Obviously, the computer is not a Dell. Dell makes a good computer,
but uses non-standard boards. The work horse computer is an old Dell.

The NEC computer has no connections to the outside world and very little
software other than NEC4 and the pro-version of EZNEC4. As I recall, the
clock is about 2.8 or 2.9 G, 400 MHz FSB, and only 1 GB of ram. I
contemplate adding a second 1 GB of ram as some large models do push memory.

If I can find the time, I will benchmark your code.
If you are interested, I can find more details.

73 Mac N8TT
--
J. Mc Laughlin; Michigan U.S.A.
Home:

I'll admit to not being an expert on this... However here are some
points to ponder.

- The NEC2 etc executable more or less runs in a DOS window. I would
guess that it runs primarily in realmode and swaps its data array in and
out of high memory. Kind of like the EMM of old. Does this mean it is
only a 16 bit app and therefore a waste on a 32 bit processor? This is a
similar logic one applies when using a RISC workstation for CAD work.

- I had thought that to a degree software has to be aware of the dual
processor/core capability before it can use it. I would think that a
"dos box" probably isn't aware of same. This would explain 50% use.

- If you are doing a lot of modeling it may be worthwhile setting up a
networked pair of PC's. One that has the user interface and the other
doing the calcs. With some modification to the bat/cmd file that calls
the NEC exe you could hand the task off to the other machine quite
easily. Setting up a queuing function will be the way to go. This would
then also free you up to use a purpose built processing engine, possibly
even use Linux or even a later DOS version.

Am prepared to be shot down on any of this!

Cheers Bob

Arie wrote:

FILL= 15.922 SEC., FACTOR= 19.156 SEC., RUN TIME = 35.406

Pentium 4, 2.8Ghz, 1 Gb (W2000, cpu running at only 50% during calc's,
Dell optiplex GX620)

It seems to be some kind of dual core processor system, because on
windows task manager
I see two cpu usage history graphs. So 2.4 Ghz single core seems to be
faster than 2.8 Ghz double core...

Arie.

Bob Bob wrote:

- The NEC2 etc executable more or less runs in a DOS window. I would
guess that it runs primarily in realmode and swaps its data array in and
out of high memory. Kind of like the EMM of old. Does this mean it is
only a 16 bit app and therefore a waste on a 32 bit processor? This is a
similar logic one applies when using a RISC workstation for CAD work.

I don't know. When it comes to adressing range, normal windows
application can only adress a max of 2Gb (appr 11000 segments) out of a
maximum total of 4Gb, if available. I know from experiments at work
using VMware system virtualization, where max memory is also an issue,
that for some hardware special work-arounds are available to increase
the 2Gb limit to a max of 3Gb.

- I had thought that to a degree software has to be aware of the dual
processor/core capability before it can use it. I would think that a
"dos box" probably isn't aware of same. This would explain 50% use.

Yes, one would normally say so, but when looking at both processor
utilizations one showed a load of appr 40%, the other of appr 60%.

Bob Bob wrote:
The NEC2 etc executable more or less runs in a DOS window.
[edited for brevity]
Does this mean it is only a 16 bit app and therefore a waste
on a 32 bit processor?

I have no experience with NEC2 at all so my comments are not
specifically about that application.

I just want to point out that running as a "console app" does not
necessarily mean the app is 16-bit. Using a modern compiler one
commonly builds 32-bit applications, both console and graphical user
interface. One would have to use an 'ancient' compiler to build a
16-bit application. If NEC2 was built a long time ago that may
certainly be the case.

[more editing]
- I had thought that to a degree software has to be aware of the dual
processor/core capability before it can use it. I would think that a
"dos box" probably isn't aware of same. This would explain 50% use.

Absolutely. If the application in question is a single threaded
application then it will NOT be able to directly take advantage multiple
processing units. (It's single thread may well be switched around
between multiple processing units, at the whim of the OS, but only the
single thread will run.) You are correct that a single threaded
application running a CPU intensive task would likely show 50% CPU
utilization on a dual core machine.

Any application written for windows must be designed and written
specifically to use multiple threads of execution in order to directly
take advantage of multiple CPUs or multiple cores. And, yes, console
applications *can* be designed and developed to be multi-thread capable.

Almost conversely, just because an application is multi-threaded does
NOT necessarily mean that it's threads will be run on separate
processors. It is (mostly) up to the OS to decide which thread runs on
which processing unit, at what time and for how long. Think of an
application with 10 threads running on a dual core machine.

Threads are still pre-emptively interrupted so that other threads can
gain access to compute resources. The machine you are reading this
message on is probably switching between 30 to 100 active threads right now.

One big advantage of a multi-CPU or multi-core machine is that when
running a highly CPU intensive task (as I gather these modeling
applications must be) the other CPU or core is still free to serve up
CPU cycles to the OS and any other applications that are running. This
would make the machine feel more responsive to the user.

Sorry to ramble on..
--
73,
HZ

W1HBQ wrote:

I'm planning to buy a new computer, and its main task will be for
antenna simulations using NEC. The issues appear to be: Intel verses
AMD, single verses dual core, front side bus and RAM speed, and of
course processor speed. I would appreciate any guidance.

It would be interesting to compare computation speeds of various
computers for the same antenna. One possible antenna to use for
comparison is The Grid Yagi from my website:
http://home.comcast.net/~ross_anderson/GridYagi.htm
http://home.comcast.net/~ross_anderson/GridYagiD1.htm
My three year old Dell with a Pentium 4 2.8GHz with HT and 3GB DDR of
RAM at 400MHz using the Nec2dXS*.exe calculating machine gives:
- - - MATRIX TIMING - - -FILL= 12.469 SEC., FACTOR= 11.422 SEC.
I'm interested to see how this compares to more modern computers.

Ross W1HBQ


Most of the antenna work I have done was with 20 - 40 MHz 386 class
processors using YO, AO, AOPro. The speed is not really needed so much
as an understanding of what is happening and a guiding hand that knows
how to tweak the optinization if that's what you are doing.

My longest run was about a month with a 20MHz 80386 plus coprocesser on
a wasted effort to make a 2/220/432 interlaced beam with regular
(parallel) elements. Not really possible to do well that way, but can
be done another way after you know how. Which I did not learn, of
course, for another 15 years. hihi

The faster processors, when optimizing, are actually a pain, as the
optimization often runs away before you can pause and tweak it.

tom
K0TAR