Make Blender use ALL of my ram and cpu

Or at least more than 178 out of two gigs !

Please tell me how to make it allocate more usage.

Make bigger and more complex scenes. Usually, people want the resource usage to be as low as possible, so you are lucky one!

Blender will only use what it needs, and can only use what’s available.

CPU: will use full amount when rendering except for AO (and maybe SSS too?). If your system is multi-threaded or multi-core, you can increase the number to be used in the render settings.

RAM: on 32-bit OSs, applications can only use about 1.5 gigs. And these apps only get what’s left over after allocating memory to the OS itself, and the graphics display. To get more, you would have to 1) add more RAM, 2) enable a setting called the ‘/3GBswitch’ if you’re on Windows, and 3) use a special build of Blender that is ‘large address aware’ and which can be found at graphicall.org.

Hope that helps!

1 Like

if you’re posting this 'cause you think things are slow, perhaps your video card needs more oomph. if slow, tell us your system specs (CPU model, speed, gfx card, OS, etc).

Hmm, well I sure am not getting that at all. I have two threads. 2 gigs of ram 2, Dual 1.8 gig ppc for a standard def ntsc size image
and this is the output, it’s taking a minute and a half a frame for a simple world spin. I have 600 to go! This wont work, I have always had this issue with Blender. A composite does not take that long, it’s just an image over an image, I do this crazy fast in Photoshop all the time, and that takes seconds.

Fra:99 Mem:143.15M Sce: Scene Ve:52608 Fa:52224 La:1
Fra:99 Mem:143.15M Sce: Scene Ve:52608 Fa:52224 La:1
Fra:99 Mem:184.56M | Part 1-4
Fra:99 Mem:173.75M | Part 2-4
Fra:99 Mem:183.95M | Part 4-4
Fra:99 Mem:157.87M | Part 3-4
Fra:99 Mem:123.97M | Compositing 12 Render Layers
Fra:99 Mem:123.97M | Compositing 11 b
Fra:99 Mem:123.97M | Compositing 10 RGB Curves
Fra:99 Mem:123.97M | Compositing 9 RGB Curves
Fra:99 Mem:134.52M | Compositing 8 Blur
Fra:99 Mem:134.52M | Compositing 7 Vector Blur
Fra:99 Mem:145.07M | Compositing 6 Blur
Fra:99 Mem:145.07M | Compositing 5 Mix
Fra:99 Mem:145.07M | Compositing 4 3 RenderLayer
Fra:99 Mem:150.34M | Compositing 3 Sharpen
Fra:99 Mem:155.61M | Compositing 2 Screen
Fra:99 Mem:160.89M | Compositing 1 Composite
Saved: /tmp/0099.jpg Time: 01:32.07

With the RAM, it is quite simple: the more things you have, the more will be used. So try to make some high poly (100 of thousands polys) models, texture them with 4096x4096 texture in different channels and that should need some memory. Considering there are some optimization structures that use memory too, it should use a lot? But why did you mentioned that 178MB limit?

For the CPU, the thing are a bit more complex. Not every algorithm can be threaded, so some will use only one core.

Do you have this problem also with other applications? Like when create an image with 10 layers in PS/GIMP of high resolution and run some filter on them?

EDIT
Ok, after you posted the previous post, I have some advices. Increase the number of parts. This is essential to fully utilize the potential on multicore processors, because in blender on part is one thread. Use at least 16, better is to use 8x8 at a price of some overhead, but the processor will be used more efficiently. Depending on the render resolution the composting part might take long, but from the memory usage it seems that you don’t have a hight res picture. Do you use Save Buffers (in Render buttons)? If so, turn it off, you have enough memory handle it and the hdd is much slower than RAM. Don’t forget, you have only 50k poly, that isn’t very much, in modern games there are much much more, so if this scene took a lot, then what about something more complex?

ahh…power PC…I have an iMac 1.8 G5 w/ 1gb ram. Unless I manually set the thread priority fairly high, I don’t get 100% cpu usage even for rendering. This is a little bizarre for compositing though…how long does the actual render take without the comp? I’m assuming you’ve been watching activity monitor or something similar. does it use all cpu for rendering the parts at least? can you show us the scene or something, or at least tell us what type of scene you’re rendering with what settings?

Nope, Photoshop happily eats all of my ram if it wants and makes things go fast. Same goes for all Adobe apps, other apps like Modo and games are fine as well.

I call bull however that all Blender needs is 178 ram, more ram = faster, I want it to just take it all and render faster, I have enough to go around. Putting more things in the scene wont do me any good, I don’t need anything else in the scene, I dont need a farm yet, besides Blender wont use more than 178-200 megs as it seems.

Run more calculations, render two frames at once. Whatever, it seems I am going to have to make a duplicate Blenders and render these out in chunks “on the same computer” again, I hate doing that, things get mixed up fast

Yeah I can use the computer just fine to do other stuff cause It wont take more, otherwise I would just leave it alone if it would behave

As for the scene’ it does not matter I have always had this problem with any project I do with Blender

Apologies for the attitude, stuff like this urkes me is all

edit delete

edit delete

You don’ have there enough data to use more RAM, and using more RAM doesn’t mean speed of arithmetic operations done on CPU. More RAM is good only if run out of it, then the missing bits are stored on hard disk, which is slow. Do you checked the CPU usage in task manager? It should be high of most time while rendering this, less only when loading from disk is happening.

Running another instance of blender won’t help I think. You have just a lot of RAM for this project, be happy that it isn’t the other case (my hdd broke because of swapping).

EDIT
Imagine this: You want to transport 30 passenger and you have a Boeing 747 Jumbo. They will have a lot of room, but that doesn’t mean that the aircraft will fly 3x faster. But you want to have have a Concorde, that is faster, but can carry only 1/4 of what can Jumbo…

CPU 166 Thats nothing ! While good usually, in this case it is not. And yes running four Blenders to render the same scene at the same time works just great, I had really just forgotten about it :stuck_out_tongue: since I hate doing it.

Really? You cannot expect a process to use the CPU for 100%. If you are talking about the overall usage, then it’s strange indeed.

myn.pheos - yes, my overall usage is in the range of 70-80% generally when I’m rendering things. really frustrating.

youngbatcat - how long does it take to render when not doing a comp? As was said, if the program doesn’t need ram, it won’t use it. If it’s not swapping to HD, then it doesn’t need more ram. PS uses lots of ram, as that’s the way the program has been developed - since it can only do operations (effectively) on one pixel at a time, it has to store the rest of those pixels in RAM. (someone correct me if I’m wrong). Obviously blur and such filters use more than one pixel, but usually not the entire image at once.

yes, my overall usage is in the range of 70-80% generally when I’m rendering things. really frustrating.

I agree, this is frustrating. I don’ have PPC, so I don’t know, but this seem like a big reserve for other applications. When rendering, or running any other computation heavy programs it is better to use the hardware fully.

I misread your first post . . . I thought you meant 1.78 gigs out of 2, not 178MB.

Realizing now that you’re on a Mac, I don’t know if the memory advice I gave applies in any way or not.

That being said, it wouldn’t really matter. Using more RAM would not speed up render time, only faster/more CPUs would do that. RAM allows you to store more data . . . a bigger/more complex scene would increase the usage, but of course there’s no point in doing that except to test this concept.

RAM will not help you process more data, as is done in the rendering stage . . . it’s all on the CPU. Running multiple instances of Blender will not speed things up. The multiple instances will increase RAM usage, but the multiple render processes will be competing for CPU time (if multiple instances is the only way to utilize 100% of the CPU, this would actually help).

It doesn’t do much good to compare other apps either . . . Photoshop is a totally different kind of program, and even other 3D apps have different/optimized methods for speeding up renders.

Then how can I make it use all of my cpu ? I have dual chip system, and running two threads, so how can I make it just compute as much as it wants with no limits other than the OS itself?

Like they said before, blender will only work on a section per thread, if you’ve moved it to one X part by one Y part, then only one thread is used.

I may be misreading your posts, are you just compositing two images, or two renders. From what you write, it sounds like you are compositing images using nodes, but from the output you posted, it looks like you are rendering and then compositing using nodes.

Also, the ram used is all that blender needs to make the calculations. If you were to add say particles, or something that needs more ram, blender would up that, but since it doesn’t need it, it won’t eat it up unnecessarily. Even Photoshop will only use as much ram as it needs. If a picture is so big, then it uses so much ram, if you have a certain number of undo steps, then again, only a certain amount of ram is used, the ram only speeds up hard disk seeks, not the CPU, blender works the same way. Also the CPU utilization is not only blender, but also on your operating system and how it prioritizes tasks, that’s why you may have to set it to a higher priority to use more CPU, other programs are using the CPU too, so they are part of that automatic prioritization by the OS.

To use your computer most efficiently, I would use two threads, have the X Parts set to 1 and the Y parts set to 2, or vice versa. That’s probably the best performance you will get. You may also try an optimized version of blender if there is one for your system.

Posting a sample blend may also help with others troubleshooting the optimization.

I’m afraid this isn’t true. This means, that when one thread (where the scene is much simpler) finishes, the other (say, where are reflective and transparent materials, and many soft shadows) will be still rendering. There will be no unrendered part, so only one process is needed to finish it.

So best is to use finest grid, so the processor will be used fully as long as possible. But the more parts, the more times thread are launched, the output is merged and that creates overhead. It’s good to use at least 4x4 parts, but better is 6x6 or higher. For quad core systems, even higher. Optimized versions are fine.

youngbatcat: You can try to use 8 threads, in this case there will be running 8 computation heavy threads that want CPU time, so this might help. I don’t know if this (the low usage of CPU) is caused by its architecture (I doubt this), but it might be how processes and threads are launched on Mac OS and how the task manager assigns CPU time to processes (which is a part of OS).