PDA

View Full Version : Is there any benefits to using a workstation gfx card in Lightwave 2018/19



MichaelBeckwith
02-22-2019, 12:37 AM
I just wanted to know if there are any benefits to using a workstation gfx card in Lightwave 2018/19 as opposed to a gaming gfx card with regards to both Modeler and Layout. At the moment I have 2 gfx cards, one being an Nvidia Titan X 12gb and the other is an AMD WX 5100 8gb card. If there are huge gains to using the WX 5100 in either Modeler OR Layout then I would be happy to put the WX 5100 in as opposed to the Titan X (Maxwell not Pascal).

I have been trying to find benchmarks on the internet to try to figure out which is the best card and am wondering that the WX 5100 even though it isn't as powerful as a Titan X, it might be vastly optimised for OpenGl for Layout and Modeler. Any foeedback welcome. :)

BeeVee
02-22-2019, 01:58 AM
In terms of OpenGL performance, there really isn't much difference. Where you'll benefit is if you use a GPU render method. For the time being, there is only Octane, which requires an NVidia GPU and that counts count the AMD card.

B

next_n00b
02-22-2019, 03:50 AM
Lightwave =>2018 using more and more Nvidia only features. Last added is GPU de-noising. You will be disappointed with AMD in LW, as I was. In LW =<2015 you can benefit greatly with Quadro in OGL, but now with 2018 consumer GPU is as fast in OGL as Quadro.

pixym
02-22-2019, 05:29 AM
I have never seen any sensible benefit from "Quadro" graphic cards… In layout or Modeler displays.

RPSchmidt
02-22-2019, 06:50 AM
If you buy the newest quadro cards and use a GPU renderer like Octane, there is a huge benefit. Of course, it costs what you might expect to pay for a low-end used car.

BUT if you have the bucks, the performance pretty much speaks for itself, as you can see in the performance chart here Quadro RTX 8000 specs (https://www.techpowerup.com/gpu-specs/quadro-rtx-8000.c3306)

You could also consider an astronomically priced quadro card like the GV100, which uses the Volta architecture

The RTX 8000 and GP100 are about the same price, although using different architectures (RTX is Turing and the GP100 is Pascal).

If you are really rolling in the dough, though, all three (RTX 8000, GV100, and GP100) cards support NVLINK which allows them to not only share their GPU power, but also their VRAM.

Check that here NVIDIA NVLINK (https://www.nvidia.com/en-us/design-visualization/nvlink-bridges/)

Two of those beasts connected with two NVLINK bridges would be monstrously powerful.

And, of course, pretty expensive.

pixym
02-22-2019, 06:57 AM
For the price of this RTX 8000 Quadro card, you can buy 10 RTX 2080 TI that will be almost 8 times faster on GPU rendering than one RTX 8000 I think…

RPSchmidt
02-22-2019, 07:42 AM
For the price of this RTX 8000 Quadro card, you can buy 10 RTX 2080 TI that will be almost 8 times faster on GPU rendering than one RTX 8000 I think…

Not exactly; the average price of a RTX 2080 ti is $1200 and the average price of an RTX 8000 is $5600; so you could buy 4 RTX 2080 ti cards for the price of on RTX 8000.

Also, the RTX 2080 ti would only give you 11gb of GDDR6 maximum for rendering regardless of how many cards you had.

The RTX 8000 would provide 48gb of GDDR6 on a single card. If you used NVLINK between two RTX 8000 cards, it would essentially give you 96gb* of GDDR6 for rendering.

I'm not sure what the trade-off would be between dual RTX 8000 cards in full NVLINK and 8 RTX 2080 cards. That would be a very interesting test.

*corrected because ma brain didn't want to add this morning

pixym
02-22-2019, 08:05 AM
Ah ! I though RTX 8000 average cost was $9900

Greenlaw
02-22-2019, 08:55 AM
Not natively but some third party physics plug-ins for LightWave (like Turbulence, Deep Rising FX, and HurleyWork's UP) get a significant performance boost from compatible graphics card.