Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Dedicated GPU for architectural rendering?
- Thread starter keithp57
- Start date
- Sort by reaction score
The rendering is done on the CPU, so the mid Mini blows away any off-the-shelve iMac since the imacs now come with i5's by default.
For modelling a better 3D card is welcome, but the problem with todays videocards is that 90% of the transistors works on shaders. That is totally useless when building models. An old 6600GT probably beats any modern card on wire-frame rendering. I really wonder how Sketchup drafting performance compares between the top iMac and the HD4000. The difference will be less noticable then with Games I bet.
For modelling a better 3D card is welcome, but the problem with todays videocards is that 90% of the transistors works on shaders. That is totally useless when building models. An old 6600GT probably beats any modern card on wire-frame rendering. I really wonder how Sketchup drafting performance compares between the top iMac and the HD4000. The difference will be less noticable then with Games I bet.
While a few renderers do have some gpu options, they're not always intended for production quality renders. You absolutely have to know the software you're using and fully understand the hardware. Nothing Apple ships is capable of more than draft renders at larger sizes even if you picked something that renders on the gpu. The exceptions would be maybe photoshop, after effects, nuke, etc. If we're talking about particle rendering and raytracing some objects being comped, the CUDA options can handle that. If we're talking about your typical offline renderer, that isn't gpu bound. That you asked means that you need to better understand your own workflow. By the time you load that up with textures that will hold up at A4 or A3 sizes, you will not be able to hold them in video ram anyway with cards that are available under OSX, even on a mac pro (maybe Quadro 4000 or some of the flashed cards). Mapped textures require a lot of memory. As blanka mentioned your speed of navigation in such a scene is more gpu dependent.
Thank you both for your help. You are right, I am not very knowledgeable about how the hardware works nor how my software works. I found these on the Vectorworks (my CADD software) website:
"Integrated Graphics Accelerators are available on some laptop and motherboard combinations. These types of media accelerators (Intel HD 3000, Intel GMA, etc) should be avoided as they are acceptable for basic use but often provide poor results when used with a graphics intensive applications like Vectorworks."
And:
"The GeForce series provide good performance and excellent feature support when used with Vectorworks."
And:
"Larger display should use a graphics card with at least 512MB of VRAM, with 1GB or more recommended."
Though they don't specifically site the Intel HD 4000 (which is used in the Mini) I assume that the HD 4000 is not optimal and I should hold out for the iMac.
Thanks again.
"Integrated Graphics Accelerators are available on some laptop and motherboard combinations. These types of media accelerators (Intel HD 3000, Intel GMA, etc) should be avoided as they are acceptable for basic use but often provide poor results when used with a graphics intensive applications like Vectorworks."
And:
"The GeForce series provide good performance and excellent feature support when used with Vectorworks."
And:
"Larger display should use a graphics card with at least 512MB of VRAM, with 1GB or more recommended."
Though they don't specifically site the Intel HD 4000 (which is used in the Mini) I assume that the HD 4000 is not optimal and I should hold out for the iMac.
Thanks again.
You didn't mention what software you were using before. It should be noted that this likely relates to real time results within the program, not so much offline rendering of images to present to a client. The imac isn't that great either for CAD use. That really is more of a Windows market. As for the imac, if you want 1GB of more of vram, you probably can't pick the cheapest one. They skimp out on video ram.
Another positive for the HD4000: VRAM is no longer an issue. Since the access to main RAM is as good as previous VRAM access, the whole RAM is available for the GPU as well. In theory you could use say 16Gb of "VRAM" if you program smart.
Remember that most PC video cards need a load of VRAM since the PCI/X slot is a real bottleneck, only 5-10% the capacity of the HD4000 system bus.
Remember that most PC video cards need a load of VRAM since the PCI/X slot is a real bottleneck, only 5-10% the capacity of the HD4000 system bus.
Do you mind explaining that last comment in a little more detail. I'm sure more than one person would find this info helpful/knowledgeable.
Bleh that's not true at all. PCIe is much more common than it was a few years ago. How many are really stuck on PCI-X aside from maybe a single slot in the configuration. PCIe broke backward compliance, which was why some of them left legacy slots. We aren't talking about video games that rely on fill rates. For something like this, the gpu loads large amounts of information at once.