Response from Plaxis support for Plaxis 2D 2017:

We did not do true benchmarking to compare performance on different (recent) machines. So I am not sure what to say about the actual performance differences. Based on our previous benchmarks:

-- Prefer higher clockspeed over more cores

-- Always store your temporary folder on a SSD

Your problem will probably have a lot of plasticity involved, and in that case, calculations might end up slow to converge, but this is due to the nature of the plastic calculation in Finite Elements. The graphics card seems more that sufficient. Output can indeed be slow when using large models. Probably storing your data on a SSD can improve it a bit, but please note that Output does calculate a lot of derived Output parameters when changing to a new step or phase.

The extra 2 Mbytes of cache will more than make up for the extra 0.1 Ghz of processor frequency (3.5 versus 3.6). My experience with the i7 processors is that the ability to use the "turbo boost" frequency for a sustained period is very unpredictable. (I am sure it makes sense to the electrical engineers, but from a user perspective it seems random) I am guessing it would help you with the interactive responsiveness of your computer, but not as much for the sustained computing of a big model that is using all the cores.

One difference that I note regarding these two processors regards their RAM support. The Xeon processor will support much more RAM and uses 4 channels. This is a product of it being designed as a server processor. Is there any chance you would use more than the 64 Gbytes of RAM that the i7 supports? The newer i7 supports slightly faster RAM, but I think that is offset by the superior path to the memory that the Xeon uses. All that said, why you looking at these relatively old processors? For example, have you looked at the i7-9700?

As for SSDs, in theory they are internally tracking reads and writes to the individual data units and will simply retire them as they are worn out. Thus, the usable size of the drive will should begin to shrink as it "wears out" but you shouldn't experience data loss. Theory doesn't always work in practice, but I would think that the odds of failure-with-data-loss are not any higher with an SSD than they are with a spinning disk hard drive. The only solution for catastrophic failures of either disk type is conscientiously backing up information and retiring drives to less critical roles at about half of their expected life. In the case of model computation, all that should be lost in a drive failure is the time required to run the model, right? I was suspicious of SSDs initially, but have had really good fortune with them so far and I have become a big fan of their speed. Yet my sample size is just barely into the double digits so far.

Murphy's Law leads me to be expecting something bad to happen in the future.

The answer for your question is founded in Plaxis Manual (Plaxis -general) in the name (Hardware specifications).

By using tolerated error more than 0.01

One more component which really matters the output time is the RAM. You should have 8GB RAM. The time required for getting output also depend upon the type of analysis. If you are using time dependent analysis like consolidation or coupled flow analysis which is also a transient type of analysis then model takes considerable amount of time to run and give output in spite of good hardware configuration.

Dr. Mustafa Jafari ,

i think this system : 1- Intel® Core™ i7-7700 Processor, 8M Cache, Base freq. 3.60 GHz, max turbo freq. 4.20 GHz is good

but new generation processor series 8 Ultra! are very fast and they have cooling system

Both of them are outdated machine (newer generations of component are available), so if you are planning to buy either of those, I would reconsider.

Currently, Intel is on the 9th generation of processor (10th to be here soon), but 8th and 9th gen are really similare.

the xeon machin is even older, v3 xeon are from 2014 or so.

If you already have both of then available, i would go for the i7 7700.

I will probably try to do the comparison between my current plateforme that use an i7 5960x (8c/16t clocked at 4.2ghz) from 2014, and the i7 9700k (8c/8t, that i will clock at 5Ghz) and check the difference, if I do i'll keep you updated

Matthew Mabey Concerning your interrogation about the turbo boost behavior, the response is that is mostly have to do with power consumptions and thermals.

The cpu use a certain amount of power to run, and the amount of power required if a function of the clockspeed (to be precise, it depend on the voltage applied to the chip, but higher clockspeed imply highter voltage). generally speaking, the cpu have a "budget" of 95W to draw, but for a short period of time it can exeed that limit, for example up to 125W, resulting on higher clockspeed for while. But, when this period of time is over, or when the temperature is too high, it will throttle down to the base clock, that is more sustainable.

The other part of the clockspeed behiavor, is about the number of core being used : it will consume less power to drive one core à 4.2ghz than 4 of them, so if the task is using only one core, the boost will be higher, and will last longer.

When you combine all of those consideration, you have the kinda "erratic" behiavor of the frequency