Hello guys, finally got to make a direct comparison of CPU vs CUDA vs RTX on regular stuff I work with. Rhino 6 and V-Ray Next 4.3.
CPU (8700K, 11 threads) - 43:15
CUDA (RTX 2070) - 4:24 (10x faster!)
RTX ON (RTX 2070) - 2:55 (14,5x faster!!!)
Sorry for the massive denoiser. Render times were probably influenced because I was photoshopping the whole time. There was a tiny visual difference between CPU and GPU render.
Great differences!
I have purchased a Radeon RX 5700 XT because of a discount, mainly for gaming at home. Just recently have I discovered that I can only render using my CPU in V-Ray Next, because there are no CUDA cores 
There is a great price gap between a 5700XT and a 2070 Super which are very close in gaming performance. I don’t know what to do 
Thanks for sharing this. I’m constantly trying to learn more about the rendering options available these days. It’s hard to find good, easy to understand information (for a non techie like me) .
I’m still using an ancient Quadro K1200 and have been thinking about an upgrade - choosing between a gaming card or a so called professional one is probably first thing to work out.
Hi Ufuk,
You can either simply stick with CPU rendering or sell your AMD card. Definitely don’t jump to conclusions now, Nvidia is introducing new cards next month.
HiMisterB, that card definitely leaves you with CPU render for now. We still have those in the office. : D
If you’re not willing to spend 2000+ USD for a card, the choice is pretty straightforward: GeForce cards for rendering, Quadro cards for a smooth viewport performance. Let’s wait what next month brings.
Cheers
Jonas
Hi Jonas, my whole system is due for an upgrade - it’s six years old now (yikes - seems like yesterday is arrived). I tend to buy complete machines ready to go, I used to love tinkering inside PCs but have very little tolerance for that these days. I’ll see what Dell Outlet has when the time comes, either another Precison or perhaps an Alienware if the discounts are good enough. I didn’t know about the new cards so I’ll wait awhile - thanks for the heads up.
Hi,
Just take your time and double check what you’re buying. I’m responsible for choosing hardware configs at our company so I’d be more than happy to give you some pointers or feedback before the purchase. These branded computers especially tend to lag one or two years behind.
Cheers
Jonas
Thanks very much, Jonish. That would be very helpful indeed.
I don’t do that much rendering, perhaps because my hardware isn’t up to the job (it is painfully slow as you can imagine). Finding simple info like your speed comparison isn’t easy so thanks again for sharing it.
I did pop along to Dell Outlet yesterday and the Alienware R9s looked pretty good, with RTX 2080 cards. I have always tended to go for the simplicity and reliability of Dell, but could be swayed. But I’m not quite ready to buy yet - I keep putting it off until the next rendering job comes along. I’ll check back before I do - thanks again.
I used to own a 5700 XT. At first I was thinking like you, but I discovered ProRender and it has been very helpful in Blender and 3DS. It is really good quality and supports Ray Tracing.
I am using 2070 Super right now. There is no difference in quality, just the render time is a little lower. Usage area is much more Nvidia, many programs are not supported on AMD side.
Which is the best option for doing a high quality render? Or are they all the same, just have different speeds?
I believe GPU and CPU renderers are separate, but you can definitely achieve good quality with both of them.
I did a test of RTX vs. Cuda for my current project - a large train interior with pano output:
Cuda 13min33s
RTX 14min24s
I use 2x2080ti + 1x3090. I don’t understand why I never got a big speed improvement per RTX - some times it’s a little bit faster, some times it’s slower.
then something with your setup isn’t right. rtx/optix is significantly faster compared to cuda. I remember when I had a 2070S and optix was first activated in rhino-cycles. going from cuda to optix was not just a little bit faster, it was a lot. haven’t ever tried using cuda on my current 3080, just optix.
maybe your multi-gpu setup gets cycles confused somehow. maybe its just using the RT cores of just one of your three GPUs for optix but all the cuda cores of your three GPUs for cuda. then your result might make sense.
generally they just have different speeds. there might be a difference in some effects, not really sure. @nathanletwory should know for sure.
Sorry, I don’t know about V-Ray.
i meant rhino-cycles CPU vs CUDA/OPTIX
Right, this is a V-Ray thread, though.
…aaand I’m realizing this just now.
sorry
I found something interesting: My AMD 3900x 12-core rig is almost exactly as fast as my aging GTX 1080. I’d like to replace the video card, but now is not the time.
The 3900x just edges it out. Let’s see, the video card renders quieter, but, if I have a very large project with large renders, the video card will not have enough memory.
