Skip to main content
S
ShandRonnie1-Visitor
1-Visitor
November 26, 2012
Question

HP Z600 Workstation slow performance

  • November 26, 2012
  • 14 replies
  • 6110 views

I have a new Z600 workstaion with Windows 7-64that has a quad-core processor with 12 CPUs shown in the Performance windows of the task manager.


The problem is that only one of the 12 CPUs is used at a maximum of 8-10 percent.


I am working with WF5 and tested it while running a Windchill session and a standalone session and both show the slow performance.


There are no search paths, everything is local. While deleting components at a large assembly, the regen at 8-10% can take over 30 minutes.


Because Pro/E is compiled single threaded, it can only use one of the 12 CPUs (in this Z600), but it should use the single CPU at 100%.


What is causing Pro/E to utilize only a fraction of one CPU?


Thanks in advance,


Ronnie Shand


This thread is inactive and closed by the PTC Community Management Team. If you would like to provide a reply and re-open this thread, please notify the moderator and reference the thread. You may also use "Start a topic" button to ask a new question. Please be sure to include what version of the PTC product you are using so another community member knowledgeable about your version may be able to assist.

14 replies

P
ptc-11405601-Visitor
1-Visitor
November 26, 2012
Piggybacking on the same question... Hypothetically if I were to run two sessions with two separate licenses on the same workstation, Would one or two processors be utilized?

Thanks in advance,

Keith

P
ptc-19327451-Visitor
1-Visitor
November 26, 2012
It looks to me as if t you are using one CPU at close to 100%, and are
using all of the CPU's at approx 10% of total. No problems here, one
CPU at 100% would only be 1/12 = 8% of total.



I have the same issue with an 8 core machine. The only way I have to
streamline CREO/WF5 is to use Distributed computing for some tasks
(mostly NC)



I don't think that PTC will be ready for parallel computing in CREO in
my lifetime (too much old code to rewrite). Maybe running several
virtual machines in the background will help, someday maybe.



A real killer would be if a competing software starts using GPU cards to
offload CPU tasks and makes this mainstream (I'm looking at you
Solidworks)

How about concurrent linear static results as you tug on a
model/assembly to change features?





Christopher F. Gosnell



FPD Company

124 Hidden Valley Road

McMurray, PA 15317
B
BrentDrysdale12-Amethyst
12-Amethyst
November 26, 2012
Hi Chris,
I sent the same to Ronnie directly. However you raise an old point.

Can you imagine the game changer if one 3D solid history based modeller
uses multi core and maybe as you suggest optimised for GPU processing too.
Our industry (not just PTC) seems way behind here and for my money this is
a much easier sell than a new graphics interface with some nice to have
new functionality Who would give a rats backside if there were a few more
clicks if you software ran so fast that it actually increased productivity.

Ah well back to the salt mines.


Regards,

*Brent Drysdale*
*Senior Design Engineer*
Tait Communications
J
JonathanHodgson13-Aquamarine
13-Aquamarine
November 27, 2012
If you run two separate sessions, they can fully utilise two of your processor cores (provided that there isn’t another bottleneck).

In fact, on one occasion I’ve had three sessions all simultaneously importing large STEP files (off a RAM disk), making full use of three of my four cores!

However, it does use up three licences (and I’ve grumbled about this on the PTC forums, and to a PTC marketing person).

Jonathan
P
ptc-19327451-Visitor
1-Visitor
November 27, 2012
This is where new software, written with modern API's would have an
advantage over older (UNIX, VMS, IRIX, etc... derived) software due to
much less 'grunt work' required to get the software compiled for
parallel systems. Joel Spolsky (Joel on Software) rightly describes the
perils of rewriting your tested code (It almost killed Autodesk), but if
you could compartmentalize the software into functional libraries and
revise and test them one at a time, you might have a shot. The
bottleneck is of course the modeling kernal, most of which have their
beginnings in the late 70's - early 80's when parallel processing was
still a (very, very expensive) dream.



Since most software I have access to is run by companies that see
parallel computing as a feature to be provided at an extra cost, we need
a paradigm-changer to shake up this market. Even the specialized FEA
software I use that was parallel capable from the start (15-20 years
ago) is not ready yet to run on GPU-type massively parallel systems.
(although I am betting they are testing this as we speak). The speedups
are not one-to-one, you do not get 8x speedup for using 8 processors in
parallel even if the code was written for parallel computing. (I get
about 6x due to increased communication required between parallel
processes)



To answer another question, if you run multiple sessions of CREO, you
will see more cpu's peg at 100% when more copies of the software are
working, but of course you will be paying for extra CREO licenses as
well. Since you can't keep all of those balls in the air at once, most
of the time the cpu's will not be doing anything.



I have noticed some things that shoot down the idea of multiple
licenses, though. I find that when I am computing an NC path or
exporting a CL file in one session of CREO, I cannot open files in
another, separate session of CREO.



If I use distributed computing though, I can fire off 4,5,6... tasks at
once and see that many cpu's peg at 100% along with as many copies of
xtop.exe launched and I can run the software without issue. If I run
distributed computing hosted on my own machine, it doesn't seem to have
to grab extra licenses to do this though. It's almost a win-win.





Christopher F. Gosnell



FPD Company

124 Hidden Valley Road

McMurray, PA 15317
A
A.DelNegro1-Visitor
1-Visitor
November 27, 2012
Interesting thought on the multiple licenses thing… I remember once when I had a ‘borrowed’ license that I was able to start multiple sessions on the same machine. So I just did a test. I borrowed a license from the server and then started multiple sessions of Pro. We only have 2 licenses, and I was able to open 3 separate sessions of pro on my machine. I checked the server status and it still shows 1 of the 2 licenses available for use.



Interesting functionality. Not sure if this was intended or not, and I hesitate to broadcast it in case PTC would want to close the loophole. (not that they really listen here, right?)



Anyway, FYI I’m only on WF4 so I can’t vouch for newer stuff, but give it a try if you have the need for multiple sessions…



Tony


D
DamianCastillo1-Visitor
1-Visitor
November 27, 2012

Interesting discussing.


I read a technology created by "Cloud Invent" an they have a new modern Kernel that is designed differently than the old kernels. One of the advantages is that it can do both parametric and direct modeling with the same kernel code. This means you don't need a parametic app and a direct modeling app that are coded to work with each other. This kernel does not distinguish each style of modeling as different. The other thing I noticed is the performance.


Check out their articles and demos here. I am sure they can support parrallel computing with this modern kernel but the chances of a small group like this to launch a new CAD system is very unlikely. What if PTC purchased the technology?


http://www.cloud-invent.com/


"Too many people walk around like Clark Kent, because they don't realize they can Fly like Superman"

B
BenLoosli23-Emerald III
23-Emerald III
November 27, 2012
From a single workstation, PTC is supposed to use 1 license no matter how many Pro/E processes you start. Since you can only be ‘active’ in one session it is done by design. I don’t know if they changed this in Creo, but Wildfire is that way.

Multi-threading of any CAD software across multiple CPUs is hard. Since the majority of them are written as history-based logging of the design build. How do you process feature 33 when features 1-32 have not been built? While Siemens, PTC and Dassault are trying to speed things up at the kernel level, it is hard when features build on top of each other to complete the final model. There are some areas of the code that can be spread out, the biggest we see today is in hidden-line removal of drafting views. Since this is a relative ‘static’ operation, the code can be split while processing.

Chinese proverb – ‘Careful for what you wish for.’


Thank you,

Ben H. Loosli
USEC, INC.
J
JonathanHodgson13-Aquamarine
13-Aquamarine
November 27, 2012
“From a single workstation, PTC is supposed to use 1 license no matter how many Pro/E processes you start.”

Do you have a source for that? We definitely claim one licence per session – if I have two xtop processes, I’m using two licences.

It’s a shame because there are times when it’s really useful to have more than one session running, but for obvious reasons we’re strongly discouraged from doing so.


On multi-threading, that’s a fair point regarding model regeneration. It’s not an easy question, but I’d like to think that there are ways to get round it – perhaps look at the parent-child tree and see whether there are separate branches that can be regenerated in parallel. Failing that, let me at least open the next command, even if I then have to wait for the ‘ready’ flag before I can use it… calculate mass props in the background… export and import STEP in a separate thread…

CPUs aren’t going to keep getting faster at the same rate as they used to, so programmers are going to have to find imaginative ways to multi-thread if they want their software to continue to handle bigger, more complex models.

Jonathan

D
dgschaefer21-Topaz II
21-Topaz II
November 27, 2012
I was just going to post the same thing. On the rare occasions that we run out of seats, it’s because someone accidentally (or not) fired up two sessions, grabbing two licenses.



I’d love to have one license per workstation.



--
Show more replies
Terms & ConditionsAccessibility statement