Aximmetry can definitely support very complex scenes. Like you said, performance of the low resolution rendering in the viewport isn’t comparable to rendering the scene in full quality. There’s nothing Aximmetry can do about this, it’s strictly related to UE5 performance. Rendering to a Decklink card, keying, light wrap, color correction and the occlusion mask rendering also have a performance overhead, but that overhead would be as high or higher if you did them in UE5 as well. Regarding the CPU, you have to differentiate between the number that Task Manger is showing, which is the total performance available across all cores and the CPU meter in Aximmetry, which shows the performance available to render the scene in Aximmetry. Just because you have spare CPU across all cores it doesn’t mean that UE5 and Aximmetry can spread the workload across all cores. Game engines are typically built for high single core speeds and don’t do multithreading very well. This means that for real-time rendering, most of the cores will not be used. This isn’t related to Aximmetry either, it’s just how game engines like Unreal works. Hope that makes sense?

Thank you for your reply,

Well, from my experience, the number of core did leads an increase in ue performance. .

I explain why: I bought a 16 cores Threadripper cpu and then decide to sell it and buy a more powerfull 32 cores cpu (same model) cause I was not fully satisfy of the performance. The speed of the single core is the same but the number of cores is the double, and I saw a significant increase in unreal performance. This also gave me a feeling of relief and pleasure since I spent almost 4,000 euros on the new processor, and I lost about 600 euros in reselling the previous one (which I had bought a few days before).

That is for sure, the fps is  higher, everything run more speedy and smooth. The machine is exactly the same with same components, I only replaced the cpu.

For what concern the view port resolution, it is set to 3840 2160 in the Level Editor section of the Editor Preferences in UE.

The frame rate is higher when running ue for aximmetry, than in the aximmetry composer, even if I only insert the ue module and connect to the out.
I dont know, anyway, what really matter is find the solution : ) Do you know if I add a second remote render machine this could help the workload on the main machine? Or, as I imagine, the second machine only process the video signal coming from a camera but does not affect the workload in unreal?

Again, thank you for your support.

G

What I’m trying to explain is that performance doesn’t scale linearly with additional cores. Sometimes you’ll even get lower performance with more cores due to added heat generation. While some things in Unreal might utilize the extra cores, others will not. Some tasks needs to run on a single core, and if a single core can’t handle the task you’ll get frame drops, even if the total CPU usage is just 10%. For real-time rendering an Intel 13900K will probably perform better than a Threadripper. The Threadripper will of course be much faster for other tasks like shader compiling, but that doesn’t help when you need the highest possible single core clock speeds.

I’m assuming you’re doing green screen? There’s no way to split rendering between two workstations for this scenario. You can render two different camera angles for two different cameras, but how would you ever be able to render the same perspective in sync with a single camera input? 

Thanks for your comment. In unreal I have no problem and my workstation is working perfectly with every kind of softwares (and I am very happy with it). The trial period is ending and aximmetry did not meet my expectations in terms of performance. As much as I consider it a beautiful program, I'm moving towards other solutions.

Thank again and wish you the best for your work and your projects.

Ciao

Hi Giancarlo,

Thanks for your message.

I dont think it is a cpu problem, many studios using similar technologies prefer professional workstations, similar to mine, for their setups. For a whole series of reasons, (numbers of pci-lines, memory capacity, reliability, computing power in a large number of tasks).

Puget systems also uses threadripper for several setups, and for some situations it detects a SLIGHT increase in performance with higher speed cpu and fewer cores.

I'm almost sure it wouldn't change anything or it would make the situation worse.

Let's not forget that unreal runs great on my machine, I work at 60 fps on very complex scenarios and with the viewer resolution in UHD.

In any case, I have seven days left, I want to do one more test before abandoning the idea.

Have a good day!

G

Hi Giancarlo,

If you don't have a UHD monitor, then it is very hard to force the normal Unreal Editor to render in UHD. Otherwise, you should make sure you start New Editor Window:

And that the output log writes out that "LogViewport: Scene viewport resized to 3840x2160, mode Windowed."

I don't know how you compared the CPU speeds.
For example, what Aximmetry writes out in the ProcessorLoad is very different compared to other applications' monitoring tools.


Looking at Puget Systems computers, they are recommending Unreal and Virtual Production workstations. They are not recommending Unreal render computers. In their blog, they write about compiling, where of course the more cores a CPU has the better it is (linearly). This is also true for many professional video or image editing programs, the more cores the better.
But in the end, you would use Unreal for rendering images in real time. This means for example that Unreal can not just put one core to help render this frame and the other core to help render another frame. All cores will have to work for the same frame at the same time. While in the case of for example a video editing program, the cores can work on any frames while you make changes or when doing the final render. The same is true when compiling or cooking an Unreal project.
Actually, parallel computing in game engines is way more complex than that, and there are many other issues.
One other thing is that Unreal is a game engine and it was designed to render games for users who will use gaming stations.

But in the end, as you said, probably there is not much difference between Threadripper and Intel 13900K and it is not the source of the performance issue you are experiencing.

Warmest regards,

Hi

some updates about my situation.

I have rebuilt my unreal project from scrath because my session needed to be optimized, and also changed the graphics card (from RTX A5000 to 4090).

Now my problem with the CPU seems solved, it works around 15%, the value of the GPU however is high, it works around 70/80% with peaks over 100 and more when adding the video player signal to test a green screen. In short, I'm happy that the CPU isn't under strain now, and surprised about the performance of the GPU. Hower the system seems to work well and I hope that everything will be ok when I will  have to record the scene. Thank you for the support.

G

Hi,

Some of the GPU load when using the video player module could be because GPU decoding is turned on.
You should make sure GPU decoding is turned off in the Edit->Preferences:

Only turn it on when you are absolutely sure your project will be CPU-bound instead of GPU.

Warmest regards,

Thank you,

GPU decoding was already turned off but thank you for the info, I will try other codecs for the video, and then I still have to try with the SDI signal coming from the camera.

The GPU works heavily also before adding the video player, but maybe its just because its a large unreal scene (large world, many trees, foliage etc) and the resolution in 4k.

I could try the Nvidia DLSS plugin in aximmetry  unreal editor (I have to search for it cause I have read that a special version of the plugin for aximmetry is needed).  

Apart from the indicator reporting high GPU loads, the system seems to work well. 

Thank you.

Giancarlo