The computer I am running SolidWorks on has a 12 core processor, with 32GB of RAM. Solidworks seems to level off at 10% CPU capacity and 1.5GB of Ram. How can I get it to use more than that? It is a Student edition, does that make a difference?
The computer I am running SolidWorks on has a 12 core processor, with 32GB of RAM. Solidworks seems to level off at 10% CPU capacity and 1.5GB of Ram. How can I get it to use more than that? It is a Student edition, does that make a difference?
During drafting that is fine, but I am trying to mesh a component and it still will not use more than one core. With meshing times getting into the couple of hours.
David,
I've played with PhotoView a bit recently and know it will use all the cores I've got (just 8, and I can't to do much else with the machine 'til it's done) As far as meshing goes I haven't had an opportunity to mesh anything for a long time and that was with a different program. From what I've read here FEA/simulation type stuff should use everything available. As for the meshing, I recall that was sort of a separate step, and I don't know how SW deals with it. There are folks here who I'm sure do, but you may need to move this question over to simulation or some other category to get their attention.
I'm assuming, of course, that you have checked that your machine is properly set up (correct type of RAM, installed correctly etc.) and everything is playing nice. Other programs manage to utilize more CPU & RAM?
Erik,
There must be something in the program that does not allow the mesher to use more of the CPU (or the process itself). I left the mesher running over night, and it looks like it finished ( it is giving me other issues now) after about 12 hours, When I tried to run the simulator it did use around 60% of the CPU. I never had these issues with 2013, but 2014 has been giving me grief.
What processor do you have that has 12 cores? How are you coming up with 12 cores? check your bios settings as you may have hyperthreading enabled in your bios. this splits your physical cores in two which doubles what you really have. Its great for large calculation programs that use more than one core. Otherwise in your case it may be keeping you from using a full core.
Although Intel markets their multi-core processors as the way of the future, many programs are not written to use multiple cores effectively, and nothing can be done about this by the end-user. The programmers of the software need to completely rewrite their code for multi-core support, which is not trivial due to needing to keep separate but parallel tasks synchronized across the multiple cores (race conditions, file locking, atomic operations).
After the single core Pentium D series back in about 2007, Intel went with multiple cores in order to reduce problems with cooling. It is much easier for Intel to cool many separate slow cores spread across a large area than a single very fast core with all the heat focused in one tiny area.
If or when you upgrade in the future, you should look for Intel processors with the fewest number of cores but also the highest TDP (thermal design power). This will give you the fastest possible speed per individual core, and the most effective processor usage by single-threaded software (although still, typically only using 1 core out of several).
Of Intel's processors intended for workstations, this is likely the current top performer for single-threaded programs:
Xeon E5-1630 v3: 4 cores, 3.7 ghz base speed / 3.8ghz Turbo, 140W TDP
ARK | Intel® Xeon® Processor E5-1630 v3 (10M Cache, 3.70 GHz)
Here is one interesting study I've found:
https://www.pugetsystems.com/labs/articles/Solidworks-2016-Multi-Core-Performance-741/
More up-to-date article focusing on comparison of a new processor (Ryzen) to others-- but in the Conclusion you can see the impact of fewer cores (4) with higher frequency (4.2 GHz) (the Intel 7700K) on everything but rendering (I don't need rendering): https://www.pugetsystems.com/labs/articles/SOLIDWORKS-2017-AMD-Ryzen-7-1700X-1800X-Performance-908/
That article appears to be reccomending two 6 core xeons, when a single 6 or quad core would be more than sufficient for general solidworks use. Anything else is a waste of money. High clock speeds > High number of threads
Jack Brooker wrote:
That article appears to be reccomending two 6 core xeons, when a single 6 or quad core would be more than sufficient for general solidworks use. Anything else is a waste of money. High clock speeds > High number of threads
Not really....see this on page 13:
"CATI’s tests showed that increasing the core count from 2 to 4 saved 35 minutes when running the macro script, which is a 12% performance increase. CATI tested performance at the 6-, 8- and 12-core levels, which showed some, but not a substantial gain. That wasn’t a surprise."
Followed by this on the next page:
"...raw processor speed can also make a big difference. Because SolidWorks is largely singlethreaded, processor clock-speed offers a straightforward return on investment. Simply put, the faster the processor speed is, the more instructions it can complete in a set amount of time"
Sorry Dan but that's what I was saying, a Quad core processor allows Solidworks to utilize threads at a higher efficiency than using a 6, 8 or 12 core processor. Sure, that is an increase in performance, but if you're looking to save money on a processor, the minor increase in performance between a 3.8GHz Quad core and a 3.8GHz 8 Core is not worth the exponential monetary increase. Increasing clock speed is the most sure-fire way to get the most performance out of your money, which is why I stated "High clock speeds > High number of threads".
I did not see that paragraph on the following page though so I'll apologize for that. But I stand firm by my statement that, as far as monetary value is concerned, as long as the processor is at least a quad core, clock speed should be a priority over core count
Jack Brooker wrote:
Sorry Dan but that's what I was saying, a Quad core processor allows Solidworks to utilize threads at a higher efficiency than using a 6, 8 or 12 core processor. Sure, that is an increase in performance, but if you're looking to save money on a processor, the minor increase in performance between a 3.8GHz Quad core and a 3.8GHz 8 Core is not worth the exponential monetary increase. Increasing clock speed is the most sure-fire way to get the most performance out of your money, which is why I stated "High clock speeds > High number of threads".
I did not see that paragraph on the following page though so I'll apologize for that. But I stand firm by my statement that, as far as monetary value is concerned, as long as the processor is at least a quad core, clock speed should be a priority over core count
It specifically said that the biggest increase in performance was going from 2 to 4 cores.....after that the performance increase dropped off. So I am trying to say that it agrees with your statement.
"CATI tested performance at the 6-, 8- and 12-core levels, which showed some, but not a substantial gain. That wasn’t a surprise"
Dan Pihlaja wrote:
Here is a good article to read...
Thanks for the Article. It looks like something I'm very interested in.
Do you know what was the timeline of the testing?
When was this article published?
Rick Becker wrote:
Thanks for the Article. It looks like something I'm very interested in.
Do you know what was the timeline of the testing?
When was this article published?
Here is the actual link:
http://files.solidworks.com/partners/pdfs/WP_PARTNER_BOXX_Maximizing_SolidWorks_Performance_ENG.pdf
I believe that this was written in July of 2013, so it may be outdated. But I still think that there is a lot of good information.
Hi David,
Sorry for the delay, but I have some useful information that can help. I wrote a blog a little while back explaining exactly what you need to know regarding hardware utilisation and SOLIDWORKS Simulation. There is a lot of scatterred information out there so I thought I would compile my own research and findings and after confirming this with SOLIDWORKS I published it.
Basically to summarise the document, only certain areas within Simulation use multi-core functionality i.e. curvature mesh not standard mesh, and particular solvers, etc. Hyper-threading is not supported whatsoever so if it is enabled, SW will only see 6 cores (if 12 with hyper-threading enabled). Always choose faster clock speed over multiple cores. (Unless you are rendering) You can always manually assign cores to carry out particular cores as well, have a read and if you have any questions just ask.
Phil.
David,
SolidWorks is only going to use one core. Hence your 10% usage. Rendering with PhotoView, or FEA stuff can use more cores.
For RAM usage, just make big complicated assemblies with lots of complicated components, you'll get there. But unlikely to all 32GB. The best I've done is about 15 and that was only by trying hard, not in "real life".