[sefde]

I want to build an ultrafast CPU/workstation for CFD that can be used for analyzing entire aerodynamics of an F1 car.

Please show me the link of the computer components such as motherboard, processor, ram, graphic card, anything.

I need ultrafast ram such as 2133 mhz or even faster.

Just show me the link.

[stuart23]

Sefde, I think you underestimate just how much computing power it takes to resolve the flow field around an F1 car. On the UTMOST lower limit of resolution, you are looking at a 100 million element RANS simulation. This would be considered low-fi, and well below what actual teams do.

In order to solve a case like this, you will need a lot of computing power and memory. I would suggest going with a DMP cluster. You should be able to pick up old servers for relatively cheap.

If you are still wanting to build your own machine. I would recommend a blade chassis (supermicro, dell, hp etc etc.) filled with blades with 2 x Intel Xeon 5xxx processors, 64GB Netlist RAM, and QDR Infiniband. Then add a Lustre Filesystem (you could stripe across the blades or have it separate) and throw an nVidia Tesla on one or two of the blades for post-processing. Unfortunately you will not get much change from USD100,000 for this machine.

I know some people who have built CFD workstations with gaming hardware, and they work OK. But what you need is another 10 levels above that...

[scipy]

Way to crush his dreams, Stu!

[stuart23]

Too harsh?

I just read this thread here http://www.cfd-online.com/Forums/har...-analysis.html looks like I'm not being too unreasonable. 1 Blade enclosure typically has 16 boards, with 2 processors per board and 4 cores per processor this gives 256 cores; which is the same as what Abdul recommended

[scipy]

Hehe, I know. It's just that poor sefde (as some of us do too initially) planned to simulate F1 aerodynamics on a desktop PC and now u are being a neigh-sayer! ))

I just realized that sefde is the same person who started the other thread I guess he didn't like the cluster answer on September 1st, so he thought to try a week later and see if things had changed..