Learning Forum

[Mark165165]

I'm upgrading my computer to run the Discovery Live. I'm having a
rather lively discussion with my hardware vendor regarding the system
requirements.

I read the specs and was under the impression that the ASUS (my
motherboard manufacturer) Nvidia GTX 1050 was enough to run DL. He is
adamant that only the 1070 will do.

I'm not looking for "recommended" requirements, even more so that I'll
not be running CFD but mechanical simulations, just "minimal" and the
1070 seems a tad overpowered.

Could you please clarify this? Will Discovery Live run on a NVIDIA GTX
1050 minimally or do I really need a 1070?

[Mark165165]

PS A list of viable models  would be really useful. Specs only mention builds (Pascal etc) and this is rather vague and unpractical.

[Naresh Patre]

Hello mark56154

I got below details about NVIDIA GTX 1050 graphics cards from NVIDIA website.

 

The above specs indicate that if you are referring to NVIDIA GTX 1050, it has only 2 GB graphics memory which doesn't meet minimum hardware requirements (4 GB graphics memory) of Discovery Live. The only card from the above mentioned GTX 1050 cards that meets hardware requirements is GTX 1050 Ti.

We have a Verification and Benchmark manual posted on this forum which lists several verification cases solved in Discovery Live using 4 GB graphics card (not NVIDIA GTX 1050 Ti but Quadro M2000M). I would request you to check out this manual to get an idea about the results obtained using 4 GB graphics card.

Though you mentioned that you are not looking for recommended requirements, I would advise you to go for graphics card having greater than 4 GB. A higher memory card will allow you to simulate larger problems with more detail.

[Mark165165]

Hello Naresh, thanks for the input.

I see in the benchmark that the Quadro M2000M 4GB has a rather poor performance, but is the Quadro M2000M not an older model?

I'll not be running CFD, but rather mechanical motion (not structural static) simulations. Are these less GPU intensive, or just as bad?

I was looking at
https://www.computeruniverse.net/de/asus-geforce-gtx1050ti-o4g-lp-brk-low-profile-4-gb-oc-mid-range-grafikkarte

or
https://www.saturn.de/de/product/_asus-geforce%C2%AE-gtx-1050-ti-cerberus-advanced-4gb-90yv0a75-m0na00-2388283.html?rbtc=ide%7Cpf%7C2388283%7C%7Cp%7C%7C

They are Pascal, run on a CUDA core, and have 4 GB... what do you think?

Many Thanks

[Mark165165]

PS

When you have the time, partnering up with Nvidia and running a more complete line of tests to get a real list of recommended models could be useful, considering how hardware-dependent (and innovative) DL is.

[Naresh Patre]

Hello mark56154

I won't say that Quadro M2000M 4GB has a poor performance since the results from Discovery Live are heavily dependent on graphics memory - higher the memory better are the results. Though this card is older, it still meets the minimum hardware requirements of Discovery Live (in fact my laptop has this graphics card).

You mentioned that you'll not be running CFD, but rather mechanical motion (not structural static) simulations. My apologies but I am not able to understand what do you mean by mechanical motion simulation. Can you please provide some detailed information in this context? In case you are not aware, the types of mechanical simulations that can be currently performed in Discovery Live are Static Structural analysis and Modal analysis. You may want to check out this post detailing the types of simulations that can be performed in Discovery Live.

Lastly, regarding the 2 Geforce GTX 1050 Ti links that you shared, both cards have 4 GB graphics memory which meets the minimum requirements for running Discovery Live.

[Mark165165]

Hello, Naresh!

Yes, reading the benchmark report, I realize I may not be using DL at all. I need to simulate a system with several flywheels, both axial and radial clutches and sprag clutches, and how they would all interact (i.e. several rotation variations at system input, what the output rotation would be). Much, much later structural integrity, but at first just system input and outputs to dimension each component. I was looking for something to bypass the need to physically prototype. Would you by any chance have any idea what the right tool for this would be?

Best Regards

Mark

[Naresh Patre]

Hello mark56154

Thanks for providing details about the simulation. I am checking with the team and will let you know.

[Mark165165]

That would be wonderful.

It's a study for an ISO 50001 compliance solution, for a factory floor that has several thruster service disk brakes, and my client's boss wants to know whether if he put regenerative breaking (the system I'm evaluating would have the function of consolidating several inputs form such regenerative breaks) if it could move a specific machine.

Thanks

[Subashni Ravichandran]

Hello mark56154

Sorry for the delay in response. As Naresh is unavailable, I'll be looking into your query. I will be forwarding your email id as registered to this Forum to our sales team. They will be able to help you find a suitable product after evaluating your needs. 

[Mark165165]

Thanks,  Subashni.

I wanted something a little nicer than Simulink. It's mostly a mechanical multi-body simulation.

Best Regards,

-M

[Subashni Ravichandran]

mark56154

I contacted our team internally and they suggested that you visit the Ansys website and fill out your details by clicking on "Contact Us" at the bottom of this page. 

Our sales team will verify your details at the earliest and ensure that the appropriate sales executive for your region will contact you.

I'm sure you'll find the best product for your applications, with ANSYS.

[Mark165165]

Thanks Subashni, I'll do just that. Thanks for your time!