Hello,

We are yet to discover and learn the possibilities of cloud computing and AWS, so we did some research.

Taking a look at all possible AWS EC2 instances, we have found that the Accelerated computing P3 “p3.2xlarge” instance seems to be a good choice. This uses the NVIDIA Tesla V100 GPU which is close to the recommended RTX 3080 in performance.
The memory seems to be more than enough (16GB GPU and 61GB system). IO bandwidth (network card<->cpu<->gpu) is critical but since it is a server-grade computer, there will be no problem most likely. (there is not enough information to be able to decide based on numbers)
Unfortunately, the  CPU could be a problem since we recommend (at least) 3.6Ghz on all cores and the server-grade CPUs usually have a super low frequency (2.3Ghz in this case). However, it might work if you don't have to render at 60 fps.

The best option is to test it and see how it performs.

Please note that we only mentioned the “p3.2xlarge” instance, since we only support 1 GPU configuration. Based on GPU performance needs, the P4 instance’s NVIDIA A100 GPU  would work as well since this is a newer version of the V100, but there is only 1 instance option in P4 that has 8 GPUs.

Also, we have found that AWS offers the usage of RTX workstations for their users.
“Customers can configure virtual workstations with access to NVIDIA RTX Workstations at no additional cost.”
But we could not find any more information about this option, since we suppose this is only available for registered users.
So another route could be that you ask Amazon which EC2 can be upgraded and with which RTX card.

Best regards,

Thanks Ahmed, that is a really useful and detailed reply.  We have an AWS account with EC2 (T4 GPU) for backup vMix use already, so should be able to find out more information regarding possible RTX deployment through that.  All of our work is HD 29.97p so I am hoping the CPU clock speed will not be an issue.  I will post updates and information to this thread as we find out more in case others on the Forum are interested.  Phil.

Update on putting Aximmetry on AWS:

Unsuccessful so far.

Because of non-availability of p3.2xlarge instance in our zone, we installed it on to a g4dn.2xlarge machine (NVIDIA T4 GPU) just to test the concept.  It seems the errors we have are related to installation rather than CPU / GPU.  Things to note:

During installation we received a number of errors involving failure to load various plugin .dll files.  The same errors pop up when we load Aximmetry (at the Parsing Plugins stage). See attached example.  I can see that the files are present, and I have tried copying files from my functioning local machine, but the errors still continue.

These are the plugins that fail to load when we open Aximmetry on AWS:  AximmetryEngine.Plugins.Astro.dll;  /DMX.dll;  /Mapping.dll;  /Meteo.dll;  /ModularShaders.dll;  /NVGameWorks.dll;  /Standard.dll;  /Standard3d.dll;  /StudioFull.dll

 

Also, even though we selected to install UE with Aximmetry, it did not do so.  I installed UE manually and seems to work fine, I can cook a project but cannot load it into Aximmetry without a 'Object reference not set to an instance of an object' error message.

Also, if I try and use a UE tutorial set e.g. 'Basic', it cooks fine, but cannot load into Aximmetry.  Errors are:

- hundreds of 'Unknown / not licensed module type: xxx' errors

- Object reference not set to an instance of an object

 

Having looked online, it seems that AWS is not used for Aximmetry, but having spoken to some users, they all agree it would be fantastic to be able to do this.  

We are very keen to try and move forward on this idea, and any help would be gratefully received.  This message will also be copied directly to Aximmetry tech-support.

Thanks very much, Phil.

Hello Phil,

Thank you for sharing your experience with us. We are currently exploring AWS and trying to somehow find a usable instance among their lineup. We will update you on this regard as soon as we have a full answer.

We also tried the Paperspace cloud service provider. Unfortunately, the latency was excessive. So, a Broadcast quality production is not achievable unless you are connected to the backbone internet of the servers. In this case, they are located in Amsterdam. Well, experiences may vary. So, you could give it go to see how it performs for you.

Best regards,

Hi Phil,

As it turns out, there can be two reasons for the failed to load plugin library errors:
The Wireless LAN feature is not installed on Windows.
Or Visual C++ redistributable 2012, 2013, 2015-2022 and their x86 and x64 versions are not installed. Also, after installing those you have to restart the computer.

After fixing both on your Amazon Web Services server, you should be able to run Aximmetry on your server.

Warmest regards,

Thanks Eifert, I will try again with AWS sometime in the future, watch this space!

Phil.

Interested in any developments on this

Hi Darren,

In which part are you interested? For example, how to set up and run Aximmetry on an AWS NVIDIA graphics card instance? Or you are interested in general about how to run Aximmetry in the cloud?

Warmest regards,