As I mentioned In the title, I have 3070 with only 8gb vram, it seems to be not enough.
If I add another GPU with more vram, will unreal engine 5.1 use vrams of both graphics card?
Incredibly unlikely, if you can get multi-gpu support to work at all it will be most likely using SLI which just mirrors the memory in both cards.
Hey there @ErenJaeger2.0! The engine itself is not capable of being run in parallel normally, recently we got support for multi-GPU path tracing, and technically SLI/AFR support has been in the engine for a while, however that’s for packaged projects, not so much the engine itself. While a very talented developer could work that into the source, it’d likely be a massive undertaking and as Ark referenced, it’d likely be SLI only.
NVIDIA SLI Alternate Frame Rendering | Unreal Engine 4.27 Documentation.
so the only solution to make a professional product is to buy a 2300$ graphics card with 24 GB Vram hahaha
Most developers are running a rig roughly similar or weaker than yours in most cases. It’s usually dependent on the type of game you’re making and the optimizations in both editor workflow and in game. UE5 is decently heavier than 4 in many regards, especially graphically. The shift to DX12/SM6.6 only (for nanite/lumen) is a bit of a tech hurdle. Sorry to hear the performance hasn’t been acceptable, if you’re hitting any specific snags I may be able to help mitigate them.
1 Like
Not sure where you’re located, but the 3090 is $1350 (usd) on Amazon right now.
Though I would not recommend it at all, technically if you’re really pinching pennies, you could also just buy an old machine learning card, like the Tesla M40 for $150 on ebay, but given that it has no video outputs on the back of the card you will have to be creative in getting it to work. Also it has no cooler so you’ll also need to get creative cooling it. Unfortunately I don’t think it supports hardware raytracing.
I was really hoping the lower end 30/40 series cards would see some sort of memory increase, but alas here we are. Apparently stuck in 8-12gb land for eternity.
1 Like
my performance is going well, but it’s a vram issue, but I can work on optimizations.
I have a different question, I’m currently working with nanite objects, can I do scaling settings later, like if I want to release a game, can I scale performance to lower resolution so everyone can run my game?
Absolutely! Scalability is built in to the core of the engine, and (most) features are scalable by default. Nanite itself is only applicable for DX12/SM6.6 however if you have to force your client back to DX11 for any reason the nanite meshes will use their fallback mesh, so no worries there as long as the fallback isn’t too heavy as a conventional mesh.
Disclaimer: One or more of these links are unaffiliated with Epic Games. Epic Games is not liable for anything that may occur outside of this Unreal Engine domain. Please exercise your best judgment when following links outside of the forums.
The game settings nodes for BP are here if you’d like to skim through them:
A tutorial from Matt Aspland that goes into using the game settings to create a main menu:
Note: the game settings changes won’t apply in editor so you have to try them in test builds if possible.
1 Like
THANKS A LOT !
since i can start with high settings and scale them down, can i start with low settings like 2k objects and automatically upgrade them to 8k later (nanite), or is that not possible?
You’ve got the right idea. During development you’re going to want to work with the lightest assets possible while you get your pipeline and mechanics laid out. Once your game is a game and is playable, that’s when you want to start adding assets that you can show off as a vertical slice or minimum viable product down the line. Starting with the biggest assets first make source control far more work as well as more resources to load/unload whenever you’re just implementing and testing things. For reference, on a commercial product for a team I worked with, we didn’t stop art grayboxing until 70% of the mechanics were in.
Basically Nanite is built so that you could technically do that from the start, but yes you could just swap out the meshes relatively easily and even change them from normal SM (with LODs) to nanite in batches, but if you have tons of assets you may need some knowledge of asset management using Utility Blueprints or batching, but these skills you will need later in your development career so best to learn them early!
1 Like
Hi! I have a question. I’m working on a fan art project in unreal engine 5.1.1. And it’s like every time when I was the megascan foliage of grass, flowers, and bushes with 2k resolution. I keep getting the gpu d3d error crashes constantly. Because of my dedicated gpu memory passes throught 6.8 gpu ram. And I just have gotten the rtx 3060 ti with 8gb ram, and 16gb ram with. Is it because my project is too large for the ram not to take?. Do I have to wait for myself to get a need vram card and new rtx 40 series or something?. Because I’m trying to lower the resolution textures by using the virtual textures streaming. I had all of the other application apps closed (but kept only one opened). Tried out the tdr delay a few times before. Didn’t work much. And I tried to lower the triangles percent for the static meshes a bit. While even when I enabled nanite on all of them. So I don’t know what else to do. And I’m really having a stressful, struggle, and depressing hard time with this issue. And BTW, my performance is running fine. My framerate is 60fps on the editor. So if you have any happier ideas of something can cheer me up, without anything involving not to disable ray tracing, lumen, virtual shadow maps, making the graphics quality to extra lower to look doll in the editor. Just let me know. Thank you
Hey there @Joshuab_artist! It’s a bit dependent on if the crash is related to your VRAM or if it’s a specific driver issue that’s occurring. Often, 8gbs of VRAM is more than enough to work on most projects.
You’ve probably run across a thread somewhere similar to this one where I run down a bunch of things that can be done. This first one lists most of the standard ones, but some users have added more over time:
This second post is the nuclear option, but has proven effective for a number of users with persistent crashes that resetting Nvidia settings, reinstalling drivers, and etc. fail: