Thea Edition v1.2

[choppedporks]

Congratulations GPU looks great.

Doesn't work on my gtx 670 though
EDIT: After some attemps it actually works just not in every case. Sorry for misleading. I'll make some more tests.
BTW: Is this a normal performance? http://www.thearender.com/forum/viewtop ... 444#p86444

But I've tested adaptive BSD - is way better now
I wish you all the best!

[CVN-76]

I have to admitt I'm disappointed by the move to CUDA.
I've been waiting to purchase Thea Render based on this from the blog in 2011:

"There is already some work on a prototype and such solution will be based on the OpenCL standard..."

Is OpenCL rendering still on the slate or is it a lost cause?

Thanks.

[giannis]

Congratulations GPU looks great.
Doesn't work on my gtx 670 though

But I've tested adaptive BSD - is way better now

I suggest first to make sure that the graphics driver is updated. Also, the whole scene has to fit inside GPU memory.

Best wishes

[giannis]

I have to admitt I'm disappointed by the move to CUDA.

I am also disappointed with OpenCL (based on vendors' implementations).

And no, it is not a lost cause.

Best wishes

[artedesigner]

@cvn-76: what is the main reason you would prefer OpenCL over CUDA?

Besides that you shouldn't take a post of 2011 to complain over something which is not even sure when it tells you that only a prototype version will be based on OpenCL. As many people may know, OpenCL has got more limitations than CUDA and the OpenCL drivers are simply not mature yet. This could be one reason they choose CUDA and honestly, I'm happy they did.

Just an example:
Where CUDA comes in strong and OpenCL needs a lot of catch-up is with what they’ve built on top of the language. CUDA has support for templates, which brings nice advantages. Then there is a math-libary which comes for free:

cuFFT – Fast Fourier Transforms Library
cuBLAS – Complete BLAS Library
cuSPARSE – Sparse Matrix Library
cuRAND – Random Number Generation (RNG) Library
NPP – Performance Primitives for Image & Video Processing
Thrust – Templated Parallel Algorithms & Data Structures
math.h – C99 floating-point Library
For most there are alternatives or you can easily build them yourselves, but there is nothing alike. This will of course come in time for each architecture, but now this is the big win for CUDA.


Resuming it:

CUDA

+ is marketed better.
+ has developer-support in one package.
+ has more built-in functions and features.
- only works on GPUs of NVIDIA.

OpenCL

+ has support for more types of processor architectures.
+ is a completely open standard.
- Only AMD’s and NVIDIA’s OpenCL-drivers are mature – Intel and IBM are not yet there (grade of maturity)
- is supplied by many vendors, not provided as one packet or centrally orchestrated.

[leminilab]

Congratulations GPU looks great.
Doesn't work on my gtx 670 though

But I've tested adaptive BSD - is way better now

I suggest first to make sure that the graphics driver is updated. Also, the whole scene has to fit inside GPU memory.

Best wishes

Hi Giannis and again, my sincere congratulations to you and the team on the new release!
I wish you all the success you deserve!

When you say the graphic card needs to have enough memory for the scene to fit in, what do you mean exactly?
The file size needs to be smaller than the total amount of memory available in the graphic card?

[giannis]

Hi Giannis and again, my sincere congratulations to you and the team on the new release!
I wish you all the success you deserve!

Thank you very much

When you say the graphic card needs to have enough memory for the scene to fit in, what do you mean exactly?
The file size needs to be smaller than the total amount of memory available in the graphic card?

I don't mean something special, I just want to point out that the scene should fit in the video memory same as
the scene should fit in RAM for CPU rendering. But usually video memory is less than RAM, so it could become
a limiting factor more easily.

Best wishes

[niki]

more success to you Giannis. thank you very much for creating thea render.. im very happy im a thea user

[naudofr]

I haven't enough time these days to explore this new Thea edition and play with it unfortunately but as few as I see until then it's a fantastic work you have done Giannis and your team.
GPU was the first thing I waited, but I realize that render presets and UI improvements are excellent too.
Congratulations again !

[choppedporks]

Congratulations GPU looks great.
Doesn't work on my gtx 670 though

But I've tested adaptive BSD - is way better now

I suggest first to make sure that the graphics driver is updated. Also, the whole scene has to fit inside GPU memory.

Best wishes

Hi Giannis,
I've installed the latest driver before installing Thea so it is something different. Also I performed the test on a scene with just one cube in it to make sure it doesn't overload GPU RAM - it's not it either.

Best Regards,
Kuba