SGI WorkStation
dave_h...@my-deja.com
and I was hoping to get some additional information together for them
so that we get the right machine.
I will be using the machine to make architectural 3d renderings, and
the office wants them as realistic as possible. So that means nice
landscaping (Brics), excellent material texturing, and great glass
effects, and so on. Oh yeah, I forgot to mention, they want all this
in one day, ha.
Anyway, I've been using Triforma J with BricsView to create my
renderings in the past, and I've been using a Workstation with 512 mb,
PIII 500 and an Oxygen GVX1 video card. So I've been getting by with
this, however my biggest hold up with the renderings is the process of
actually producing the image on the screen to check it out. Because of
the deadlines, I usually get one crack at seeing the image on the
screen before I need to save it away. So fixing up errors isn't always
possible. If I had a faster machine, I could make better corrections
to the image before saving the file to the real thing.
So, that's the deal. We seem to think that the SGI machines are the
answer. I was hoping that someone out there can make a few suggestions
on what to look for and what to avoid, etc....
Also, what's the opinion on the software that I chose to do my
renderings? We have Triforma J, so that's what I use. But I am open
to learning other softwares such as 3d Studio Viz or Max, or any other
that will help me create realistic architectural renderings in an
efficient manner.
Also, how come nobody has any decent cars for microstation? None of my
renderings have cars (real cars, but not memory busters).
Thanks for any input,
Dave Horvath
Uniland Development Company
Amherst, NY
Sent via Deja.com
http://www.deja.com/
Bear
You can get some quite good cars from ModelVision. They have textures with
each and also have trees and so on.
I've also got a few vehicles on my site if you want to have a look.
If you haven't already, I'd consider joining the Viz mail group. There are
some very
good people there that have lots of good ideas.
As far as rendering to the screen goes, I tend only to do this for small
models. With
anything of size it's easier to render a small image as a test. JMO.
When you get to do you final image, try using the banded rendering method.
This
allows you to use several machines across a network to render one image. It
does
make a big difference.
If you are rendering a few images from one setting in ray trace, remember to
have
the 'keep database in memory' checked. This means that once you have done
the
pre-processing you don't need to do it again. Be careful with this though,
if you change
any rendering settings you must uncheck this and recheck it. Sometimes it
does not
reset and the changes will not show up.
The other setting to remember is 'render all objects', make sure this is
also checked.
This is at least a start.
HTH.
--
Bear
http://dropby.at/forwardthinking
<dave_h...@my-deja.com> wrote in message
news:93utj0$ao3$1...@nnrp1.deja.com...
dave_h...@my-deja.com
What you've mentioned, I currently practice, except for the Banded
Rendering. From what I've read that isn't reccommended for my setup
here. Maybe you could explain that method more to me.
We have 5 workstations; mine has 512 ram, the others each have 256 and
we are linked to a network. So maybe with that info you could let me
know what the next step would be.
Thanks,
Dave
Any other SGI related comments out there????
CADDCOP
should render in 1/4 of the time. You might even find you an live with
1/4 to 1 /16 size images for testing and then switch to banded
rendering wher ready for final results. I used to do all development
and testing during the day and final renderings overnight. And that was
on a 486!
Sorry, No SGI experience however.
--
Chuck Rheault (c_rheault[at]gpinet[dot]com)
Greenman-Pedersen, Inc.
Laurel, Maryland Branch
Bear
use banded rendering.
What this does is when you start the save,a *.bnd file is also
created. Other machined can then access this file from the
image save option as help to render the image.
One thing to be careful of is to make sure each machine has
the same settings for the file. For instance, if one machine has a white
background and one a black background, the your final image
is a Zebra.
HTH.
--
Bear
http://dropby.at/forwardthinking
<dave_h...@my-deja.com> wrote in message
news:93vap9$mgp$1...@nnrp1.deja.com...
dm...@my-deja.com
range. It was very poorly setup and I had repeated problems with it
resulting in (eventually) a new hard disk.
SGI support were loath to send out an engineer and made me update the
bios (On a brand new machine!!) and loads of other stuff before they
finally agreed to send an engineer.
The 230 came with Windows 2000 and although they supplied speakers they
did not configure the soundcard to work. (I know that relatively minor
but it shows a lack of care). Also you have to be very careful using
the windows update feature as some of the stuff it recommends you
install will cause you no end of grief (ie constant crashes).
This may be a rogue machine but It's put me off buying any more.
Regards
Duncan Macdonald
Keith Laurie
while back, I would stick to standard PC framework, but max out the critical
areas e.g.
- as grunty a CPU as is cost effective (one down from the latest release?)
- as much ram as you can fit running at a fast bus speed
- as good a Graphics Card as you can afford (ask Sam. He seems to know his
stuff on this)
- as fast a Hard drive as you can find (wide and fast rpm. I am not up with
the latest guff, but last I heard Scsi was not all that much faster than
others, and was more expensive, so shop around)
- a "smart" motherboard that can maximise the throughput between components
with least amount of bottle necks.
You will probably find for the premium you pay for a pretty looking SGI box,
you can get a better performance run of the mill box.
Regards
Keith Laurie
<dm...@my-deja.com> wrote in message news:941g4b$gnu$1...@nnrp1.deja.com...
Chris Zakrewsky
> the latest guff, but last I heard Scsi was not all that much faster than
> others, and was more expensive, so shop around)
To make a long story short:
SCSI allows asynch read/write request whether EIDE/ATA not.
SCSI drives rotate faster than EIDE/ATA devices.
SCSI creates less CPU overhead than EIDE/ATA.
EIDE/ATA cost half the price of SCSI (in fact even less, given one need
SCSI controller as well, when EIDE/ATA usually is 'for free')
HTH
--
Remaining With Our Very Best Regards,
/Chris Zakrewsky
+------------------------------------------------------+
| ch...@cadperf.se http://www.ustation.se |
+------------------------------------------------------+
| Team uStation AB (CAD Perfect) |
| Älta Strandväg 37B, S-138 33 ÄLTA, SWEDEN |
| tel +46 8 773 1003 fax +46 8 773 3004 |
+------------------------------------------------------+
| Bentley MVAR, Independent Software Developer |
+------------------------------------------------------+
"Keith Laurie" <cadman...@xtra.co.nz> wrote in message news:yI596.206$fyl.13...@news.xtra.co.nz...
dm...@my-deja.com
comes with EIDE.
Duncan
In article <eWb96.1038$TI3....@nntpserver.swip.net>,
Jim Weisgram
<ch...@cadperf.se> wrote:
>Hi,
>
>> the latest guff, but last I heard Scsi was not all that much faster than
>> others, and was more expensive, so shop around)
>
>To make a long story short:
>
>SCSI allows asynch read/write request whether EIDE/ATA not.
>SCSI drives rotate faster than EIDE/ATA devices.
>SCSI creates less CPU overhead than EIDE/ATA.
>EIDE/ATA cost half the price of SCSI (in fact even less, given one need
>SCSI controller as well, when EIDE/ATA usually is 'for free')
>
>HTH
Help me decipher SCSI vs. IDE tests. 2 identical Dell systems, except
one has a SCSI interface and drive. Win NT 4.0. Comparable drives,
presumably latest super duper wide/fast SCSI etc. Load a 280 Mb raster
reference file, rotate. IDE completes in 11 min, SCSI 28 min.
Under most conditions, both run typical applications in identical
times.
---
Jim Weisgram, Oregon Department of Transportation
Disclaimer: all opinions posted are mine, not my employer's.
Although, you never know. Sometimes they might agree with me.
Chris Zakrewsky
I miss some details here, can you confirm?
1. File systems? Are they identical? Both FAT, both NTFS, or...?
2. Both discs equally defragmented? The same defrag tool? or..?
3. Identical (cloned) MicroStation cofigurations, right?
4. No external databases (Oracle, ODBC, or whatever)?
5. By 'rotate' you mean as Fence Rotate, right?
(p1 thru p4 above have MUCH larger impact on performance than
any of IDE/SCSI architecture differences)
--
Best Regards,
/Chris Zakrewsky
"Jim Weisgram" <jwei...@hotmail.com> wrote in message news:sm0e6t8gccskgaofn...@4ax.com...
fourt...@teacher.com
Microstation J Benchmark program. Here are the results:
SGI Workstation - 1 GB Ram - SCSI Hard Drive - Wildcat Vid Card
600 MHz Chip (I think)
MSMBench Benchmark Report
MicroStation/J Standard Benchmark Test
Test #1:Video Update1
Test Time: 0.57 secs
Test #2:Video Update2
Test Time: 1.12 secs
Test #3: Dynamic Panning Simulation
Test Time: 7.83 secs
Test #4: Typical heavy design session1
Test Time: 0.58 secs
Test #5: Fence copy multiple times
Test Time: 0.74 secs
Test #6: Process render options
Test Time: 39.28 secs
Test #7: Process render options
Test Time: 16.32 secs
Total Test Time: 66.44 secs
DELL Precision 410 - PII450
256 MB Ram - Elsa Gloria Card - Dual Striped SCSI Hard Drives
MSMBench Benchmark Report
MicroStation/J Standard Benchmark Test
Test #1:Video Update1
Test Time: 0.375 secs
Test #2:Video Update2
Test Time: 0.906 secs
Test #3: Dynamic Panning Simulation
Test Time: 6.5 secs
Test #4: Typical heavy design session1
Test Time: 14.125 secs
Test #5: Fence copy multiple times
Test Time: 7.328 secs
Test #6: Process render options
Test Time: 17.625 secs
Test #7: Process render options
Test Time: 10.781 secs
Total Test Time: 57.64 secs
Intergraph TDZ2000 W/ 256MB RAM - 600 MHz Chip (I think)
Wildcat Card - SCSI Hard Drive
MSMBench Benchmark Report
MicroStation/J Standard Benchmark Test
Test #1:Video Update1
Test Time: 0.39 secs
Test #2:Video Update2
Test Time: 0.672 secs
Test #3: Dynamic Panning Simulation
Test Time: 4.437 secs
Test #4: Typical heavy design session1
Test Time: 11.719 secs
Test #5: Fence copy multiple times
Test Time: 5.953 secs
Test #6: Process render options
Test Time: 16.734 secs
Test #7: Process render options
Test Time: 9.625 secs
Total Test Time: 49.53 secs
Jim Weisgram
<ch...@cadperf.se> wrote:
>Jim,
>
>I miss some details here, can you confirm?
>
>1. File systems? Are they identical? Both FAT, both NTFS, or...?
Both NTFS. But I did learn since the first post that the IDE system is
partitioned as a 2 Gb partition on a 20 Gb drive, something to do with
unfamiliarity with the Ghost utility by the installer. The SCSI is
partitioned as a 20 Gb partition.
So they aren't as similar as I thought. However, I don't know that I
like the idea that such a partition size should make such a large
difference on NTFS.
>2. Both discs equally defragmented? The same defrag tool? or..?
Yes, both brand new systems, both the same defrag tool.
>3. Identical (cloned) MicroStation cofigurations, right?
Yes, both installed from the same installation image.
>4. No external databases (Oracle, ODBC, or whatever)?
None
>5. By 'rotate' you mean as Fence Rotate, right?
No, I mean the rotate raster reference file command that rotates by 90
degrees.
>
>(p1 thru p4 above have MUCH larger impact on performance than
>any of IDE/SCSI architecture differences)
---