SolidWorks - CPU Clock vs Core Count


#1

A loooooong time ago when the X99 boards were released, I jumped on an MSI X99S-SLI PLUS for Christmas. I didn’t know anything about what I wanted CPU-wise beyond that it was going to be a ‘5820k or higher’ in performance. I’ve also got an older MSI Z87-G55 from 2 Christmases before that. Needless to say, I’ve got a bad habit.

A few months ago, I decided on a reasonable list for the money (X99, 32GB, GTX 780Ti for under $2k). After scouring eBay and recommendations/argument threads for used Xeons, I settled on the Xeon E5-2690v3, a 12-core, 2.7GHz chip that still cranks numbers for its speed. I found it for sale in multiple places, from $400-600, which is quite reasonable for a $2000+ chip. Granted, the clock speed will be lower by .1GHz due to it being a ES variant. At the time, I was still a bit wary of the clock speed, so I checked performance charts. It still pulled like a champ in individual core performance, placing about the same as the 5930k at stock speeds. Of course, data on paper will never matter when two chips are placed side by side, but data gave me a general idea of how well the individual core performance was.

I made sure to keep the 5820k/5930k as a backup plan in case the Xeon didn’t provide a safe path for where I wanted to go.

Many may immediately think that speeds trump core count, but what matters to me is usable performance for my money. I plan to ‘multitask’ quite a bit (in my case, have CAD open on one monitor, multiple data/chrome tabs on another monitor, and potentially another monitor), so there’s that.

Now that I know where I’ll be going in the career field, I know I’ll be using SolidWorks/Inventor/other CAD programs a ton. I saw on a few threads in places saying that CAD programs, especially SolidWorks, favor high clock speed over core count due to the prioritization of calculations on one core. I couldn’t find much info on favoring for GPUs (eg. Quadro vs GeForce equivalent), so that would be reserved for a different time (unless you can provide insight at this moment).

My questions are:

  1. Do you think an individual core performance equivalent to that of a stock 5930k would be sufficient for SolidWorks loads?

  2. If so, would a $400-600 12/24-core, 2.7GHz chip provide a benefit over a $400-500 6/12-core, 3.1-4.5GHz chip? Do you think the extra 0.4-1.5GHz in speed would make a significant difference in performance?

Thanks in advance!


#2

My daughter uses a program that need high single core speed. I don’t know if it is solidworks, or what.

What you are looking for is the best single core performance, and overclocking would be a bonus.

> 2.7GHz vs 3.1-4.5GHz chip

Performance is fairly linear when comparing clock speeds (all else being equal). You probably won’t notice 2.7 vs 3.1 (15% difference), but you will notice 2.7 vs 4.5 (67% difference).

Cache is also important for Solidworks. If the 2.7 has more cache than the 3.1, then go for the Cache.

> I couldn’t find much info on favoring for GPUs

Get the fastest card you can afford that is Solidworks Certified compatible.