|
[Sponsors] |
July 12, 2018, 13:43 |
Workstation for around 1 million cells
|
#1 |
New Member
Andrzej
Join Date: Jul 2018
Posts: 1
Rep Power: 0 |
Hello All,
I have read a couple of dozens threads here, but I still have trouble deciding on a "perfect" workstation for my use. I hope you will be able to help me. Currently I've got old i7-3770k, 16GB of RAM, some SSD, and it takes almost a day to finish my 700k cell simulations. It's a side project, a couple of times a month, just to get some cash. That's why I don't have the budget or need to invest in Xeon Gold server grade units. I have three choices, listed below, that I think would be best, but I need to confirm with you what is the best choice. 1) AMD R7 2700X ASRock X370 PRO4 or MSI, or Asus X370 with NVMe. G.Skill Ripjaws V DDR4, 2x8GB, 3200MHz SSD Samsung 970 EVO 250GB PCIe x4 NVMe GPU 1050 4GB or Quadro P600 SilentiumPC Supremo L2 V2 550W 2) Intel I5-8600k Some Z370 motherboard 2x16GB DDR4-3200 SSD Samsung 970 EVO 250GB PCIe x4 NVMe GPU 1050 4GB or Quadro P600 SilentiumPC Supremo L2 V2 550W 3) Intel I7-8700k or 7700k some compatible MOBO, priced is similar to the Z370 one 2x16GB DDR4-3200 or faster SSD Samsung 970 EVO 250GB PCIe x4 NVMe GPU 1050 4GB or Quadro P600 SilentiumPC Supremo L2 V2 550W So, which would be the best for my case? And how much improvement can I expect in simulation time? Also, do I need graphics card? Should I get 2x8GB, or 2x16GB, or 4x4(4x8) set? Would populating 4 slots on Dual Channel processor/MOBO actually lower the performance vs 2 slots with Dual Channel? And what type of benchmark should I look for in popular processor tests that would reflect the power for CFD simulation in the best possible way? Thank you for your response, Andrzej |
|
July 13, 2018, 05:48 |
|
#2 | ||||
Super Moderator
Alex
Join Date: Jun 2012
Location: Germany
Posts: 3,427
Rep Power: 49 |
Quote:
Provided of course you are not limited by Disk-I/O in which case the new system will be much faster and investing in a larger SSD might be worth it. Quote:
Quote:
Just look at memory usage on your current system. If you have enough memory there, 16GB will be enough on the new system. My rule of thumb despite all of this: use as few memory modules as possible, there are no downsides as long as you populate all memory channels. The only downside to using 2x8GB vs 2x16GB: you will most likely get single-rank DIMMs which perform a little worse for the same frequency. Quote:
|
|||||
Tags |
2700x, 8600k, 8700k, budget |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
[snappyHexMesh] SnappyHexMesh running killed! | Mark JIN | OpenFOAM Meshing & Mesh Conversion | 7 | June 14, 2022 02:37 |
[ICEM] Problem with prism cells | sidharath | ANSYS Meshing & Geometry | 0 | September 1, 2015 08:09 |
[snappyHexMesh] problems using snappyHexMesh 2.1.0 on a supercomputer | Sunxing | OpenFOAM Meshing & Mesh Conversion | 9 | September 20, 2014 10:30 |
[snappyHexMesh] external flow with snappyHexMesh | chelvistero | OpenFOAM Meshing & Mesh Conversion | 11 | January 15, 2010 20:43 |
Computer resource for 4 million cells | Mark | FLUENT | 10 | June 8, 2005 03:13 |