[shang2]

I am currently a graduate student running OpenFOAM simulations. I am using an old i5-3570. My personal laptop has an i5-7300HQ, although I do experience overheating if I do not undervolt. As my grids are getting more complex, the convergence times are getting a bit long. I estimate my current simulation will take 20 days (it's a 2D simulation with just 90,000 grid points).

I was thinking about building a workstation. In particular, I've spec-ed out a machine (PC Part Picker link here) that uses an i7-7820X.

Looking at some synthetic benchmarks, in multicore tasks, the i7-7820X seems to be 2.5X faster, and that's before accounting for faster RAM and a potential overclock.

My question is: what kind of performance boost can I expect from this upgrade? I would be paying for this out-of-pocket.

[flotus1]

My question is: what kind of performance boost can I expect from this upgrade?

Quite a few variables and unknowns here...maybe something in the order of 2-4x compared to the desktop.
The focus of your purchase should be memory. Get 4 DIMMs that are as fast as possible. 16GB total would be enough since the cell count of your model is very low.

Are your simulations steady-state? 20 days of computing seems like a lot considering the problem size. Maybe some settings can be tweaked to speed up convergence?

[abhishadyy]

Originally Posted by shang2

I am currently a graduate student running OpenFOAM simulations. I am using an old i5-3570. My personal laptop has an i5-7300HQ, although I do experience overheating if I do not undervolt. As my grids are getting more complex, the convergence times are getting a bit long. I estimate my current simulation will take 20 days (it's a 2D simulation with just 90,000 grid points).

I was thinking about building a workstation. In particular, I've spec-ed out a machine (PC Part Picker link here) that uses an i7-7820X.

Looking at some synthetic benchmarks, in multicore tasks, the i7-7820X seems to be 2.5X faster, and that's before accounting for faster RAM and a potential overclock.

My question is: what kind of performance boost can I expect from this upgrade? I would be paying for this out-of-pocket.

You shall be experiencing substantial boost here and the parts you have selected looks fine.. else you can also try consulting system integrators like Ant PC or Boxx for their simulations ranges.

In between, I have considered here that your applications depend on CPU Cores and processing power.

[ashokac7]

Originally Posted by shang2

I am currently a graduate student running OpenFOAM simulations. I am using an old i5-3570. My personal laptop has an i5-7300HQ, although I do experience overheating if I do not undervolt. As my grids are getting more complex, the convergence times are getting a bit long. I estimate my current simulation will take 20 days (it's a 2D simulation with just 90,000 grid points).

I was thinking about building a workstation. In particular, I've spec-ed out a machine (PC Part Picker link here) that uses an i7-7820X.

Looking at some synthetic benchmarks, in multicore tasks, the i7-7820X seems to be 2.5X faster, and that's before accounting for faster RAM and a potential overclock.

My question is: what kind of performance boost can I expect from this upgrade? I would be paying for this out-of-pocket.

If you want this for simulation the you could get 16 core AMD Threadripper in this price. I am considering you have to change the motherboard for i7 also, as it will not support.

For simulation if you use parallel solver then 16 cores make lot more sense than 8 core in same price. And i7 7820X looks more like for multi-tasking. And I will not go for over clocking as simulations can last for 10-20 hours and more. Keeping your processor at this high temps and voltages will reduce the durability drastically. And there are more chances of crashes in over clocking.