CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > General Forums > Hardware

Doubt about possible bottle neck

Register Blogs Members List Search Today's Posts Mark Forums Read

Like Tree4Likes
  • 2 Post By Simbelmynė
  • 2 Post By flotus1

Reply
 
LinkBack Thread Tools Search this Thread Display Modes
Old   October 19, 2018, 10:59
Default Doubt about possible bottle neck
  #1
New Member
 
Miguel Molinos
Join Date: Dec 2015
Location: Madrid
Posts: 14
Rep Power: 8
migmolper is on a distinguished road
Good afternoon,


We are a research group and we are assembling a workstation for SPH calculations.

Our idea is to put two graphic cards NVIDIA RTX 2080 ti in order to do ours GPU calculus. Our CPU is a i7 8700-k. It is possible that this configuration generate a bottle neck when both GPU are working simultaneisly? If the answer it is yes, wich CPU would we buy?



Cheers
migmolper is offline   Reply With Quote

Old   October 19, 2018, 11:23
Default
  #2
Senior Member
 
Simbelmynė's Avatar
 
Join Date: May 2012
Posts: 524
Rep Power: 14
Simbelmynė is on a distinguished road
It should probably depend on the SPH implementation. There are many algorithms that need almost no CPU interaction. Mining is one such example, where you can use as many GPUs as you have PCIe slots on your motherboard, while at the same time have the cheapest possible CPU.


This may or may not be the case for your SPH implementation. My gut feeling though is that if you have to worry about the 8700k being a bottle neck, then the implementation is not very well suited for a dual GPU setup. What other CPU except perhaps 9900k could possibly be better? (I am assuming that you will not do any actual calculations on the CPU, only use it for non-GPU bound activities)
migmolper and tin_tad like this.
Simbelmynė is offline   Reply With Quote

Old   October 19, 2018, 12:10
Default
  #3
Super Moderator
 
flotus1's Avatar
 
Alex
Join Date: Jun 2012
Location: Germany
Posts: 3,071
Rep Power: 43
flotus1 has a spectacular aura aboutflotus1 has a spectacular aura about
Depending on the implementation of your code, there are 2 potential bottlenecks caused by the choice of CPU.
1) as Simbelmynė said: not enough CPU performance. This includes both the CPU itself being too slow in parallel workloads due to its 6 cores and lacking memory bandwidth with only 2 memory channels.
2) PCIe transfer bottleneck: The CPU only has 16 PCIe lanes for GPUs. So with 2 of them, each GPU is connected by 8 lanes instead of the full 16 lanes. Again, it highly depends on the implementation of your code if this will cause an actual bottleneck.
migmolper and tin_tad like this.
flotus1 is offline   Reply With Quote

Reply

Tags
gpu processors, sph method

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
[Other] Changing from Fill Bottle to Rinse Bottle anraw OpenFOAM Meshing & Mesh Conversion 3 August 21, 2017 07:03
Melt Glass Blowing for Bottle (Glass Distribution in Bottle oozcan Fluent Multiphase 0 May 5, 2017 15:03
[Other] Using reconstructParMesh on a cluster, bottle neck and out of memory KTG OpenFOAM Meshing & Mesh Conversion 2 May 5, 2017 12:48
[ICEM] Mesh between bottle neck splines 2D mikaido ANSYS Meshing & Geometry 0 February 10, 2016 09:38
Coke Bottle ram Main CFD Forum 13 September 29, 2000 04:17


All times are GMT -4. The time now is 08:59.