Community Tip - Your Friends List is a way to easily have access to the community members that you interact with the most! X
While I am wating for Creo 2.0 to finish updating drawing views on drawing, I was thinking what would be the best/cheapest PC configuration so I do not have to wait 10-15 minutes for drawings views to update. I am using a 3D SURFACE model of cylinder head and it is a very hard to work with it in drawing mode. For smallest change in 3D model, I have to wait 10-15, sometimes 20 minutes, for drawing to update. Aside the fact that if you want to show cross section on surfaced part, view will also show everything behind the cross section therefore it is useless, and I have to jump through hoops to get what I want. Anyway, I wonder if anyone else has any experience with dual CPU (not core) PC configuration and Creo 2.0?
I was thinking making my own dual CPU configuration: 2xIntel Xeon E5-2630, ASUS Z9PE-D8 WS and 32Gb RAM with either W7000 or K4000 GPU.
Is everything loaded into the system when you make these changes, or is it having to access the server? If so, it could also be traffic on the network and not just the computer itself.
Everything is loaded locally. I was even using "ImDisk Virtual Disk Driver" to make a fake HDD in RAM memory but no change in speed. My guess is that it takes a lots of time to check which line/surface is visible when surface model is present. I think if I speed up the system with dual CPU's it should be faster. Or not?
Currently, I am using DELl XPS with I7-930, 12Gb RAM and V7900 GPU. Before this one I had some Dell workstation that needed 30-45 minutes to do the same thing (with Pro/E 4.0).
I think a dual-CPU system will be wasted, because Pro/E is almost entirely single-threaded (although maybe Creo is a little better).
I think you want a processor with the highest single-thread speed possible:
http://www.cpubenchmark.net/singleThread.html
and then ideally over-clocked, if this is possible / allowed where you work.
Try keeping Windows Task Manager visible, on the Performance tab, set to View -> CPU History -> One Graph, All CPUs. I suspect that when you're regenerating or updating, you'll see it lock to 25% (if you've got a 4-core CPU; or 12/13% for an 8-core, etc) - if it does this then it's almost certainly limited to a single CPU thread and not bottle-necked by anything else.
According to this (recent) thread:
http://hardforum.com/archive/index.php/t-1756823.html
you should disable Hyperthreading (if enabled) for better single-thread performance. Apparently.
I get exactly what you have said. CPU usage tops at 24% and 103% clock speed. I will try to find out how I can turn off hyperthreading.
I suspect the improvement from diabling hyperthreading (if it even exists on your i7-930) will be a small percentage. Really, it seems that you want a big improvement which means a new CPU - perhaps something like an i5-3570k which will give you nearly twice the performance out of the box, and should have decent overclocking potential (it sounds like an extra 30% or even more may be possible).
Remember though that overclocking:
a) will void the warranty (and can cause damage, although the risk seems low nowadays)
b) requires better-than-standard CPU cooling (ideally with careful attention to airflow through the case and coolers)
c) can be a frustrating and time-consuming trial-and-error process
d) is probably not possible on a Dell (they often do not have access to some of the settings required)
e) probably does not constitute a 'PTC supported configuration'!
I am pretty comfortable with overclockig. Never had any issue with it. I will have to evaluate getting i5-3570k CPU against i7-930. Speaking of "PTC supported configuration" - no such a thing exist, per my experience. I disabled hyperthreading and will test to see if any gain is present.
Use K10 stat to overclock it when needed,don't burn it (please do it at your own risk) .i sometimes use it to get upto 4 Ghz
'Use K10 stat to overclock it" - how to do this?
Download that application,under P-stat tab,there is FID and DID,FID is core multiplier and DID is core divider.
select the FID and see what frequency you are comfortable with,don't increase way too much.
push in the frequency buttons and then apply and from tray icon enable clock control.
This works only with AMD CPU's not Intel ones.
Ok. I turned the hyperthreading OFF and I saw no change in Creo's performance. CPU load is at 25% for Creo per core which gives me 100% per CPU. CPU load shows evenly on all four cores (at 25%). Hyperthreading cores were "parked" before HT was turned off. So, no noticable change detected.
25% cpu usage is fine for a single threaded process using a 4 core processor. It simply means the the process has fully loaded 1 core (which is all a single threaded process can do). That's why processor speed becomes more important than core count for single threaded applications.
Hyperthreading is a processor feature that makes it appear like you have more cores/processors than you really do. For certain types of work loads the processor is able to make better use of the available resources with hyperthreading on. For processor intensive, single-threaded work loads it will not improve performance. It CAN degrade performance by allowing other threads processing time on the same core that your heavy work load is running on. This is less likely to happen though if the rest of the cores aren't fully loaded. (The processor scheduler will spread the work to non-busy cores.)
I would agree with you a 100% if it wasn't for one thing. Resource monitor shows equal load on all four cores. Which is 25%. I would expect to see 1 core loaded 100% and others less. This is because Creo would take only 1 core (single thread) and other applications would use other three cores. I do not understand why resource monitor shows that each core is equally loaded. Is there something that I am missing?
A single threaded process has a series of sequential steps needed to complete (multiple loops through the processor). The CPU scheduler will constantly "move" these sequential process steps around to different cores (load balancing basically), but the next step won't be started until the previous one finishes, regardless of what core it runs on. The long string on commands or processes completes sequentially, but only on one core at a time, and not necessarily on the same core.
This makes more sense.
By the way, you can force any process to only use a specific core. Using Task Manager (on Windows), you can right click on any process in the processes list and "set affinity". This means you are choosing exactly which processor/core that specific process will be allowed to run on. While I don't recommend doing it, you could theoretically set xtop.exe to only run on one core and restrict every other process from using that same core. Practically I don't think you will see any improvement. Windows is good enough at processor scheduling that it will make sure xtop.exe is getting all the processing time it needs.
For more reading, see this article:
http://scalibq.wordpress.com/2012/06/01/multi-core-and-multi-threading/
Thank you. What I was seeing in resources monitor, that is exactly what you have said. I guess, I will have to live with it.
I hate when I have to work with quilts in Creo. They are so hard to work with in drawing mode. It takes enourmos time of waiting. Sometimes, I even forget what I was supposed to do next. For same model in SolidWorks, it takes no more than 1 minute to laid out views. In Creo, it takes forever + 1 hour.