How to speed up your Ansys Simulations / Nvidia GPU Acceleration

[@2Pcharnal]

Great! I followed your steps, and now my RTX45000 GPU is working. Thank you very much.

[@JorgeMoralesAv]

Great to hear! I do also have an RTX A4500 but I never used for ansys specfically because I have an A6000. Let me know if it helps as well!

[@deepak_nigwal]

@@JorgeMoralesAv hi, can you compare rtx a-series cards with 40-series cards? i am trying to choose between A6000 and 4070 Super... which one would be better (for Ansys/open source FEA + Deep Learning models)

[@AmitkumarSingh-gr5rl]

Same here ​@@deepak_nigwal

[@remcodejong8033]

That's a lot to wade through for: To utilize a NVIDIA GPU device that is not on the recommended list of cards, set the following environment variable: ANSGPU_OVERRIDE=1

[@markglovatski3658]

Thanks a lot! Thanks to you, I was able to use my RTX 4070Ti, it reduced the calculation time by more than a half! Great)

[@cornevermaak6334]

Hi Jorge. Thanks for the video, I found it extremely informative. I’ve been using Ansys and Creo for many years now. Ansys has a plug-in for Creo and they work seamlessly. I can launch Ansys from Creo and the model is pulled in directly. Recently I’ve been using Fluent in simulating flow, velocity and head loss in my valve designs. I cannot find a multi gpu simulation anywhere on KZbin. You would be the first to my knowledge if you can run 2 GPU’s for us to view online. It would be interesting to see whether there’s a reduced runtime from the extra CUDA cores and VRAM. Please show us how to set this up and configure properly.. Many thanks and regards Corne

[@Gonzanoza]

Thank you for the great video!

[@gary4530]

Hi Jorge, thanks a lot for your video. I am running with RTX 3060 (though not supported officially, your way works). I have run the same model by different core numbers, GPU. Once GPU involved, the elapsed time is getting longger. I want to check with you on the CUDA settings. I have installed Cuda 11.8. while Ansys Result Information told me that the Cuda version is 11.0. And Nvidia Driver told me that the Cuda version is 12.2. What can I do to really take advantage of the GPU? Thanks a lot

[@realracing3specter295]

you found the solution? are you able to use gpu in ansys ???

[@gary4530]

@@realracing3specter295 Not really. Good luck

[@realracing3specter295]

@@gary4530 so far i learned, Ansys only allow high end professional grade graphics cards, whether they are single or double precision compute. It never support consumer level geforce cards which are widely available in the gamer market.

[@Belsen85]

I have two questions: 1. Was it linear or nonlinear solution (large deflection ON)? 2. How many Degrees of Freedom were there if you still remember?

[@sadiqnureliyev7646]

Thank you for detailed explanation. However it didn’t work well for me. I have multi-GPU which are GPU-0(NVİDİA GTX 1660ti) and GPU1(AMD Radeon) and number of cores in CPU are 6. What should I do exactly ?

[@ogbodocollinsjnr.3412]

Thanks man, this was super helpful.

[@JorgeMoralesAv]

Glad it helped!

[@e_gekko]

Hi Jorge I ordered a new Lenovo P7 but whithout GPU. But now i‘ve to choose a GPU and i‘m not sure which one. I have 3 in my mind, RTX 4090, RTX A4500 (20GB) or the new RTX A4000 ADA (20GB). Which one would you recomment? And do simulations tools (Ansys Discovery Live) benefit from the ADA versions? Thanks for the support. Your doing a great job!

[@JorgeMoralesAv]

Hello! the RTX 4090 will probably not fit in the Workstation. If you only want Ansys Discovery, i Would go with a RTX 4080. If you install the nvidia Studio Drivers, not the Game Ready, it will be now officially supported for Ansys Discovery. Thanks for watching!

[@joachimsrensen653]

Thank you for the video! Does Ansys Mechanical support multiple GPU acceleration?

[@calvintseng7283]

I’m curious about the version of the ANSYS ? And I want to ask that the others like transient-thermal can boost GPU like the way you did?

[@ilyesouadah7299]

hey awsome video, but i really wanna know does the bypass work for an amd graphics card?

[@deepak_nigwal]

hi, can you compare rtx a-series cards with 40-series cards? i am trying to choose between A6000 and 4070 Super... which one would be better (for Ansys/open source FEA + Deep Learning models)

[@luisc.sosamanzo4731]

Jorge, muy buenas noches y te mando.un caluroso saludo desde México. Una pregunta, en caso de tener 2 GPUs en la misma PC, pero ambas son diferentes modelos de NVIDIA, ¿pueden trabajar ambas en paralelo o deben ser mismos modelos?

[@JorgeMoralesAv]

Hola Luis, ambas pueden trabajar en paralelo. Pero si alguna es muy diferente en cuanto a memoria VRAM y CUDA cores a la otra pueda ser que el rendimiento no mejore tanto a como si trabajara solo la mas poderosa. Por que el trabajo se tiene que paralelizar desde el CPU y luego pasar por dos tarjetas . Saludos desde alemania y un placer ayudar

[@vacuna1014]

great tips

[@JorgeMoralesAv]

Glad it was helpful!

[@Gonzanoza]

Could you please advice if RTX A4000 can significantly improve amd 7950x3d?

[@prashantsaini6372]

Hi Jorge, I am using system coupling to couple the Ansys Mechanical and Ansys fluent to solve my complex problem; the problem I am facing with the system coupling is that it only takes 2 cores to solve the problem. Please let me know how I can increase the number of cores in the system coupling so that the simulation speed will increase significantly. My system is a rack server with two CPUs (each CPU has 24 cores). Thank You

[@JorgeMoralesAv]

Hello, first thanks for watching! What kind of licensing you have for Ansys? For system coupling there are three menus that you should take a look for multi core solutions. In workbench, in the System Coupling Analysis tool, there you can also specify how many cores. Second, in Mechanical as i showcased in my video and in Fluent again you can specify the cores. Take a look to the three options and should work.

[@BenjaminPeric-ql5lj]

Hi Jorge, thank you very much for you video! Can you do the same video for Ansys Fluent? I have the problem, that my Ansys Fluent (Research Licence) wont use the GPU at all. Even though it is enabled (*) and supportet (Quadro RTX A4000). Thank you in advance!

[@erickvaldespinolopez117]

Hola Jorge, gracias por la información que compartes con tus videos, me podrías ayudar con una duda por favor, quiero adentrarme en el mundo de la simulación con ansys , estoy armando una pc, mi presupuesto es bajo y estoy entre adquirir una tarjeta gráfica rtx 3060 o una quadro p2000 ¿Cual recomendarías adquirir? o me podrías que recomendar una mejor opción por favor

[@acreaterz]

Hello, I have Nvidia Quadro K2200 and I know it's not tested by Ansys for GPU acceleration, but I wanted to try it myself. Im getting error message "NVIDIA Library Requested but not Enabled for GPU acceleration". I've created system environmetal variable and installed CUDA Toolkit 12.3 Update 1, but still getting the same error.

[@simulationbros]

isnt it that in mechanical FP64 performance is the biggest factor?

[@JorgeMoralesAv]

Before it was a really huge deal the FP64 Precision. You have always the option to compute in single precision mode. Then the GPU FP32 will be fully utilized. Which is where the newer RTX GPUs really shine. Thanks for watching!

[@SeattleShelby]

​@@JorgeMoralesAv I have an RTX 3090. According to NVidia, it runs FP64... and it works great in Ansys Discover. Unfortunately, I can't get it to work with Ansys APDL Mechanical Workbench for some reason. *EDIT* Figured it out. You have to change the solver to "Direct Solver" in your structural analyses. It's now utilizing the GPU. Funny that Ansys Discovery does this by default -- you don't even have to enable an environment variable to get it work in Discovery.

[@MiG82au]

@@SeattleShelby Everything runs FP64 to some extent but it's so slow due to the lack of double precision execution units that a modest CPU is faster. For example, the FP64 speed of my 6 core CPU is higher than my 3080, because the 3080 has 64 times fewer FP64 units than FP32. If you're getting good speedups your solver must be using single precision.

[@phlizzow]

Thanks 🎉

[@digambarkumar8355]

hey brother my gpu usage is vey less i have a rtx 4050 and i512450 with 8 core 12 thread and 16 gb RAM 4800mhz here my gpu usage remains less than 30%

[@virenshah5387]

Hi, I have 2019R2 ansys mechanical, however New RTX 4000 16 GB GPU doesn't work. I tried environmental variable as you shown...pls help 😢

[@realracing3specter295]

you found the solution?

[@akshaymanoj4377]

does it work with gtx 1650?

[@petshub3752]

please can i use a gtx 1650 for gpu acceleration

[@RumbaniNdhlovu]

Hey Jorge, I've been watching your content since the HP Z440 upgrade. Finally got an HPZ440 and now I'm debating what to get for a GPU upgrade between RTX 2070 Super 8gb and the Quadro M6000 24gb. Which would ypu recommend for graphics designer, 3d animation and video editing work? Is the 24gb a plus or has RTX got a better impact on workflow?

[@JorgeMoralesAv]

Hello Rumbani, i would definitely recommend the M6000. But, with the prices nowadays of RTX dropping, I would strongly recommend you get a used RTX A4000. With the newest RTX features, more than 6000 CUDA cores, 16 GB vRAM and very power efficiency, I think today there is no better choice than that one. Here in Germany, are now ranging from 400 to 550 €. Thanks for watching !!!

[@RumbaniNdhlovu]

@@JorgeMoralesAv Thank you so much for responding. I didn't even consider the RTX versions. I'll look put for those. Wow. Thank you again.

[@valr1658]

I have only 4 cores Core i5 and gtx1680