What role does the GPU play in UE4 when compiling?

Well besides FPS gameplay smooth etc

What role does the GPU play in compiling etc? or is this all just CPU based?

I have a HD 4870 512MB card and looking to develop 2D games to port to android etc so I am assuming I don’t need any fancy GPU. I forced DX 10 mode with the -d3d10 extension in UE4 shortcut. So the crashing seems to be gone for now.

I see the core i3 haswell 3.5GHZ has dropped to a all time low of $109 on Amazon.

My current chip is a haswell pentium G 3220 but stinks. Since the core i3 does 4 threads maybe it would make a huge difference?

Kinda regret not buying AMD because a fx 6300 AMD is $98 on amazon right now. And I bet its double the speed of the $109 i3 haswell 3.5GHZ

GPU does nothing for compiling. It’s only for realtime graphics calculations. I don’t know how those specific processors compare, but at this point Intel is doing much better CPU’s so I’d have to go with the i3, though that’s still on the low-end.

^ hi my HD 4870 plays reasonably well still to this day and it plays Battlefield 3 aswell on the 40 fps side of things. Pretty ok if you ask me.

So would you say I can use it to develop 2D games or a basic 3D FPS game? should I invest in the i3 for this purpose? its hobbyist / student projects really.

with that said I wonder how a haswell i3 @ 3.5 GHZ with 4 threads perform compared to a FX 6300 AMD?

From a comparison here:]=2020&cmp]=2252&cmp]=1781

The i3 will be faster than what you have now, getting a better CPU won’t improve your gameplay as much as getting a better graphics card, but there’s a point where the CPU can slow everything down so you kind of want to keep them at a similar level.
As far as that AMD processor goes, it’s slower on single-core operations and since it’s a completely different system you’d have to upgrade your motherboard as well to be able to use it.

A good CPU is good for developers who need to compile materials, compile games, bake lighting, or perform any precomputed stuff. If you’re using lightmass and you’re lighting a level, you do NOT want to wait 20 minutes for lightmass to do its business, only to find out everything you need to change, then to wait another 20 minutes to see how that works. I have an i7, but I deeply regret it because more things are moving off the CPU and into the GPU this gen, including dynamic GI.

So an i3 is fine for most games and OK for developing if you want to go the dynamic route. If you want to work on more precomputed things, then an i5/i7 is great. But I would invest in a better GPU. It’s the thing that will run your games (and the entire engine) much faster. A GTX 970 is only $330, and that’s for the cream of the crop. A decent GPU does not cost much nowadays.

^ yeah because I removed my HD 4870 and used on integrated intel HD graphics and when it says “compiling shaders” it takes 100 times longer. This suggests to me the GPU seems to do a lot more calculations than the CPU.

I will get a new GPU and a i3 thanks don’t want to waste money on a i7 and no GPU

ok so wait the thing that says “Compiling Shaders” that takes forever. Is that CPU or GPU dependant?

Because I think as far as I remember when I used my IGP it was far worse.

Shaders are compiled on the CPU, as well as physics/collision simulations are on the CPU. Everything else (rendering) is done on the GPU.

OOoh I see. Hmm then it would seem my best bet is upgrade my CPU eh.

right now AMD would seem like a way better purchase than Intel. LOL their 8 core is so much better than intel’s dual core i3 for the same price. Really regret not going AMD originally lol

The reason removing the HD 4870 made compiling shaders take longer, is the Intel HD Graphics is built into the CPU. It now had to perform rendering, and compiling shaders on the same chip, that is what slowed it down. :slight_smile:

AMD processors have some pretty high stats, but it doesn’t perform to that level. Again, also remember that to get the AMD processor you would have to replace your motherboard as well, while the i3 is compatible with your current system.

aah I see well this explains a lot LOL

I wonder the performance difference between the FX 6300 AMD and the 4130 i3 haswell @ 3.5ghz
in UE4?

FX 6300 scores a 6,356 on PassMark - CPU Mark
4130 i3 scores 4,813

I’m looking at getting a i7-5820K myself. In straight up multi-threaded applications like rendering and baking AMD processors perform pretty well and are cheaper, but when it comes to single or double threaded applications or games, I’d take a i5 over even a higher clocked 8-core AMD processor.

But @darthviper107 and @DotCam , I assume it is currently technically possible for the CPU to load half of the tasks into the VRAM, let GPU calculate and give back results into RAM (or CPU) right? Or haven’t today’s CPU/GPU/Microsoft engineers thought about this so far, in 2016? o_O

The CPU and GPU do not perform computing in the same way. You would have to specifically write a program to be able to utilize a GPU. Most software companies have a hard enough time writing software that properly scales across multiple cores, especially when it’s more than 2 or 4. CUDA might be the easiest, most accessible way to take advantage of GPU power to perform computing, but it’s specific to Nvidia, and you still have to make tasks run in parallel.

You can’t load things in and out of memory like that, since you don’t know what might affect the lighting of something else, if you only load part of the scene into memory and render the lighting for that you may miss something that affects those objects.

Getting data out of a GPU is typically a pretty slow operation, too. It’s often not worth it.

Oh I was wandering since it takes soooo long for me in UE4 to “finalize load of” and “initialize world” even though my i5 6600K (stock) is at 100 % load it takes like 10-20 minutes to open a scene/level in UE4… :confused:

I’d recommend the 5820k to anyone running UE, great value. Clocks to 4.2ghz on standard cooling. Chews through an engine compile in around 15 minutes.