Activity › Forums › Creative Community Conversations › MacPro: More RAM or GPU
-
Colin Mcquillan
December 16, 2013 at 7:18 amHow well is C4D using the GPU these days? I’m a couple versions behind and it is heavily CPU dependant for rendering and really only touches the gpu for previews.
Colin McQuillan
Vancouver, B.C. -
Helmut Kobler
December 16, 2013 at 9:03 amI was wondering that about C4d as well!
Did a quick search, and didn’t see too much about Open CL support….
——————-
Los Angeles Cameraman
Canon C300 (x2), Zeiss CP.2 lenses, P2 Varicam, etc.
http://www.lacameraman.com -
Scott Williams
December 16, 2013 at 9:10 amI did find this while hunting around on Firepro graphics cards which suggests they’re starting to use them.
https://www.fireprographics.com/ws/mae/maxon/resources/AMD_Maxon_SolutionSheet.pdf
-
Walter Soyka
December 16, 2013 at 2:29 pm[Scott Williams] “I regularly use After Effects, Premiere Pro, and Cinema 4d if that helps.”
I disagree with a lot of the advice in this thread. All these apps will benefit from more CPU and more RAM. Pr is the only one of these three apps that makes heavy use of GPU acceleration via OpenCL. Graphics/mograph/compositing is still very CPU-reliant in general.
I think that Helmut’s advice to go for GPU now and upgrade RAM as soon as you can (because RAM will be easier to upgrade) is sound, but if you plan to purchase this machine and not touch it again, I’d go for more RAM and maybe consider holding out for the 8-core CPU configuration.
Ae renders entirely on the CPU, with only a few significant third-party plugins doing any internal rendering on the GPU. Most of these use OpenGL which will not benefit from dual GPUs and which will see diminishing returns on OpenGL performance on the higher Mac Pro GPUs (although the larger VRAM quantities could be beneficial for large textures on detailed models in Element 3D).
C4D also does final renders entirely on the CPU. The AMD brochure referenced here refer to viewport speed (using OpenGL or Enhanced OpenGL), not final rendering speed. I am puzzled by the references to GPU-accelerated dynamics calculations; I haven’t heard that anywhere else, and my own experience with dynamics-heavy R14 projects has been that the dynamics simulation caching was a slow, single-threaded, CPU-bound process. If anyone knows more, I’d love to learn.
Some words of wisdom on Enhanced OpenGL from the MAXON help:
Tip:
Enhanced OpenGL offers numerous relevant material channels and lighting properties that can be displayed fluidly in the editor view without having to render the image.Enhanced OpenGL can be activated separately in each view (Display menu) also with regard to effects such as shadow, transparency, etc.
Enhanced OpenGL, including corresponding options such as transparency, shadow and post effect must be activated in the respective view (Display menu). CINEMA 4D supports OpenGL post effects (e.g., the Cartoon Renderer post effect) only to a limited degree.
Don’t expect miracles, though. Of course it is impossible for any graphics card to display all CINEMA 4D lighting or material properties (current graphics cards just are not capable of this type of performance) and there is no guarantee that the OpenGL display in the editor view will reflect the actual rendered result (OpenGL and CINEMA 4D renderers simply differ too greatly from one another). Enhanced OpenGL serves primarily to offer instant visual feedback for everyday situations without having to constantly test-render your scene, in particular with regard to the depiction of shadows.
Tip:
Make sure you always use the most current driver for your graphics card.Also make sure that the graphics card driver lets you enable and disable certain graphics card options. For example, antialiasing for transparent objects can be disabled, which, of course, would result in making it impossible to render smoothed, transparent objects in CINEMA 4D.
The following properties are, for the most part, supported and can therefore be seen directly in the editor view:
- Material channels: Color, diffusion, luminance, transparency, environment, bump, normal, alpha, specular, (these can be activated / deactivated in the Editor channel, exclusively for Enhanced OpenGL display.
- Light sources: The most important options will be implemented (e.g., type, inner angle, outer angle (Spot Light), a Spot Light’s color gradient (Details tab: Use Gradient).
- Shadows: Hard Shadows only (any shadow type can be selected). All other shadows will only be displayed as Hard Shadows. Inaccurate shadow depiction can result when working with larger scenes.
- Shaders: Many shaders will be calculated natively by the graphics card (color gradient, Fresnel, bitmap (i.e. loaded textures), color, checkerboard, noise (exception: All Veronis, Ober and Zada) falloff, colorizer, filter, etc.), which will result in a relatively accurate depiction of the scene in the editor view. Other shaders such as noise, etc. will be emulated: Internally, the (relatively small) shader preview (bitmap) will be used and projected onto the object with UVW projection. These bitmaps will be saved with the scene, which can result in vastly larger (scene) file sizes!
Note:
Since this functionality emulates shaders, opening a scene that contains many shaders can require a lot of memory.Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events -
Julian Bowman
December 16, 2013 at 2:55 pmHow about if I only use FCPX with 1080 footage in ProRes and motion (mostly templates). money is a consideration so for me it seems like the best choice would be the 6 core with a 500 (mid range) graphics card (RAM i will max from elsewhere). But can I save myself a grand and just use the 4-core with the 500 card? I don’t do masses of rendering or exporting so if the CPU is export and render only on these apps, would I benefit from the 6 core?
Cheers in advance.
-
Walter Soyka
December 16, 2013 at 3:07 pm[Julian Bowman] “How about if I only use FCPX with 1080 footage in ProRes and motion (mostly templates). money is a consideration so for me it seems like the best choice would be the 6 core with a 500 (mid range) graphics card (RAM i will max from elsewhere). But can I save myself a grand and just use the 4-core with the 500 card? I don’t do masses of rendering or exporting so if the CPU is export and render only on these apps, would I benefit from the 6 core? Cheers in advance.”
My suggestion to anyone buying a Mac Pro for FCPX and trying to figure out how to best allocate a limited budget across CPU, GPU, and RAM is to wait until the unit and the new software actually ship and we can see some real-world performance reports.
Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events -
Marcus Moore
December 16, 2013 at 3:21 pmThis is the best advice that no one wants to hear- WAIT.
Wait for someone to line these machines up and test them on the software you use. Not benchmarks. No guesswork.
It may mean you won’t be able to order a machine for a month- but you won’t be sweating your decision later.
-
Shawn Miller
December 16, 2013 at 6:26 pm[Walter Soyka] ” I am puzzled by the references to GPU-accelerated dynamics calculations; I haven’t heard that anywhere else, and my own experience with dynamics-heavy R14 projects has been that the dynamics simulation caching was a slow, single-threaded, CPU-bound process. If anyone knows more, I’d love to learn.”
Maybe they mean third party plugins like Turbulence 4D? I have to admit though, T4D is the only plugin I know of that uses GPUs for dynamics calculations.
Shawn
-
Walter Soyka
December 16, 2013 at 7:20 pm[Shawn Miller] “Maybe they mean third party plugins like Turbulence 4D? I have to admit though, T4D is the only plugin I know of that uses GPUs for dynamics calculations.”
Maybe. But I think you’re being generous and I wonder if they’re just, you know, wrong. Here’s the text in question:
New levels of computational performance with AMD Graphics Core Next (GCN) GPU architecture
Take advantage of exceptional performance in intensive non-3D tasks such as effects processing and video rendering, without having to buy a second board. AMD FirePro cards based on the latest GCN GPU architecture use a 28nm design and deliver increased compute performance to handle such advanced effects and are the perfect match for the new physics tools in CINEMA 4D R14. This makes the effects of aerodynamic forces, springs and breaks look more authentic than ever. In the case of soft shadows, not only memory utilization increases but also the GPU is required to perform more complex calculations to add realism to your scene. The AMD FirePro W5000 offers over 1.2 TFLOPS of single-precision performance and packs the performance of a super computer into a single slot.“This makes the effects of aerodynamic forces, springs and breaks look more authentic than ever.” What? That sounds like a bad marketing paraphrase of a claim MAXON makes in the C4D R14 press release [link]: “R14 now offers greater levels of realism and control with increased versatility over dynamics with aerodynamic forces, plastic springs and breaking connectors. Users can now create much more believable animations of flying or breaking objects.” Maxon never mentions any kind of GPU acceleration, and with all the buzzwordiness of GPU today, they’d be shouting that from the rooftops.
Soft shadows are rendered in the viewport using Enhanced OpenGL, and they are rendered for final render on the CPU, so they seem dreadfully out of place in the compute section.
These wouldn’t be the only errors on that PDF. Look at the Cinebench R15 benchmark results (fps) chart — the K2000 and the K4000 have the same reported frames per second, but bars of different length against the same axis.
I just don’t give much credence to this sheet.
Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events -
Rick Lang
December 16, 2013 at 7:57 pm[Walter Soyka] “Look at the Cinebench R15 benchmark results (fps) chart — the K2000 and the K4000 have the same reported frames per second, but bars of different length against the same axis.”
Just a typo I assume for the K2000 results, but I agree with your sentiments on all the hype that isn’t saying much clearly.
Rick Lang
iMac 27” 2.8GHz i7 16GB
Reply to this Discussion! Login or Sign Up