Quadro vs. Geforce for large scene rendering

ZurokSlayer21

Reputable
Dec 20, 2015
12
0
4,510
Hello experts, I am trying to choose a GPU for my computer. This computer is set up to be a one-size-fits-all type of computer, that can be good at gaming, while also being able to render large scenes effectively in Blender, Zbrush, and Unreal Engine 4.

I am looking at either a high end Geforce, or Quadro. My question is which card would be better and baking and rendering a scene with a couple hundred million polygons, with 4k to 8k Targa textures for each material? I understand that the high end Quadros have more VRAM, but I have also seen situations where Geforce Titans excel at rendering. I also understand that VRAM will not increase performance, but I am using a ton of memory, or that Geforce cards will excel at gaming. Again this is a one-size-fits-all workstation, and would spending more for a Quadro really be worth it?
 
Solution
Eh. I would look into Quadro SLI, its not the same as Geforce SLI. only certain apps use it and that generally is Big company Oil and some speciality autodesk software. That second card won't be doing anything expect CUDA support, in which a Tesla is needed. I believe NVIDIA disabled CUDA on a second card unless its a Tesla for Workstations. Reason being is to sell more cards.

Also Quadros have the same shelf life as a reference geforce card. It made with the same caps and mosfets. Nothing special other than being tested for longer periods of time. Which is why they all have lower clocks speeds and boost turned off to lower the heat output, which is turn keeps the card going longer.
Hello. I´m not an expert because I don´t have Quadro, but I do renderig too (CPU mostly).

Quadros are excepcional with the visualization/modeling windows on programs like 3Ds MAX, The GForces are not too good when the scene is filled with a lot of objects (1070 and 780ti owner).

GForces and Quadros can do Render in GPU, if the render motor is compatible with the newest GPUs (Pascal).


GForces will do rendering faster because usually they are OC (Windforce, FTW...). For exemple Quadro P6000 is equal hardware to the new Titan XP, but this has more MHzs.

Quadros have more Vram, and it is the good point to have a Quadro, that 24GB of the P6000 can load very complex scenarios.

About using 8k textures, I´m not sure if you will fill too fast the vram.

Having 2 GPUs for render don´t plus the Vram, if you have 2 GPUs with 24GB, you still will have 24GB, but will render x2 faster (when I render with GPUs, I´m limited by the 3GB of the 780ti (we can force the render on 1 GPU if that one have more vram)).

Quadro P6000 costs about 5000€, TitanXP 1200€ and 1080ti about 800€.

If you want a good GPU for gaming and rendering, then go for a TitanXP (12GB) or 1080ti (11GB but with custom heatsink=more OC).

(If you don´t use CUDA for the rendering, AMD have the Firepro w9100 32GB and Firepro S9170 32GB)
 
If you are only using Blender, Zbrush and Unreal than a Geforce card is gong to be the better option. Blender is OpenGL, Zbrush up until a year or two ago was CPU only and Unreal is DirectX. Only Blender would benefit greatly from using a Quadro. However Unreal will be unusable at 8K in the Quadro family without a Quadro P6000. Using a 1080Ti will get the job done with 11GB of VRAM. 8K resolution is very demanding and resource hungry. Not sure why you are working in 8K instead of 4K.
 

I also us CPUs for rendering, but I mean't real-time rendering in the viewport, such as retopologizing an asset from Zbrush, or working with millions of polygons in edit mode without lagging. I have had some scenes exceed 36 GBs of RAM, so I know that will crash any GPU out there. I have read that Quadro's can handle billions of polys.

When I said 4K and 8K, I'm talking about the textures I use. My monitor is only 1080p. Each material I use has about 3-5 4096x4096, or 8192x8192 textures saved as a lossless TARGA format, which can be hundreds of megabytes large.
 
You have to work smartly with Zbrush. I know the version I was using was a few back, but they have the scaling options so you work with millions of polygons in small areas than view the whole model in "low poly" mode. Of course your export is high-rez texture maps, but your modeling is sectioned. I don't think any computer can handle billions of polygons without layering.

Quadros being able to handle billions of polygons is false. Its the same GPU as the consumer Geforce cards. the only difference is they generally have more VRAM and drivers are optimized for OpenGL and wireframe with AA. Even with a P6000 you would run out of memory if the scene is too big. 24GB isn't a lot when you think about everything being textured in 8K. Best thing to do is work with smaller scenes and combine them together with loading screen or realtime scene merge.
 

You have a point, and I keep forgetting that I'm having a performance problem, not a running out of memory problem. I also may have over exaggerated how many polys I work with, sorry. This is my dilemma, in Object Mode of Blender, which shows my object, but I can't edit the vertices, a 1.25 million poly model will orbit fine with no lagging, but when I go into Edit Mode, where I can edit the individual vertices, it will lag like crazy. Will the geforce cards be enough to handle this, or would I need those special drivers on the Quadro to edit this model without significant lag. I think the reason I mentioned Zbrush is I wasn't sure if a better GPU would help or not since it is CPU based. Also sorry.
 
I believe Zbrush is still 100% CPU based unless you go into the view-port mode for scene setup. As for blender this will benefit from a Quadro because they are optimized in the drivers for vertices and wireframe. Same GPU, but NVIDIA basically crippled he consumer Geforce cards on purpose. For example the P1000 is a GTX 1050, but can drive more frames in wire-mode than a 1080TI. 1.25million polygons is nothing for Pro card.

You don't need to spend $6000 on a video card, when a P4000 would run circles around any consumer card.
 


I think I am actually revolving more towards 2 P5000s in SLI. That in theory should give more power than the standard 1080, but still have all of the optimizations of a Quadro. This way I can view and edit dense character models in Blender, but also be able to see decent performance in more consumer applications. I am OK with losing a few FPS in games and UE4 if that means my life in Blender will be easier. I have also heard that typically Quadro's can handle higher loads for longer periods of time, having more durability than consumer cards.
 
Eh. I would look into Quadro SLI, its not the same as Geforce SLI. only certain apps use it and that generally is Big company Oil and some speciality autodesk software. That second card won't be doing anything expect CUDA support, in which a Tesla is needed. I believe NVIDIA disabled CUDA on a second card unless its a Tesla for Workstations. Reason being is to sell more cards.

Also Quadros have the same shelf life as a reference geforce card. It made with the same caps and mosfets. Nothing special other than being tested for longer periods of time. Which is why they all have lower clocks speeds and boost turned off to lower the heat output, which is turn keeps the card going longer.
 
Solution