GTX680 vs. HD7970 ... same performance ?

[JuanPabloCuervo]

Hi,

both cards in the same PC,
GTX680 & HD7970

Indigo Renderer can only use 1 card at a time, Bug#1.
...
Bug#2 ...
they measure exactly the same spp & ks/s.

CUDA vs. OpenCL1.2
8 compute units vs. 32.
1050Mhz vs. 925Mhz

Caterpillar Image looks different, at 6minutes.
or Nvidia OpenCL 1.1 vs. AMD OpenCL 1.2

in other softwares like LuxMark 2.0 or ratGPU, i can select any, and they measure different.
so i guess something is wrong with Indigo>Renderer.

Thanks.

[Zom-B]

Indigo Renderer can only use 1 card at a time, Bug#1.

No Bug, Indigo simply only supports one GPU in GPU mode!

Bug#2 ...
they measure exactly the same spp & ks/s.

Maybe no Bug, but your CPU is maybe way older, and can't feed any of your cards with data fast enough, so they are having the same speed since your CPU is to slow to max out the GPUs.

Caterpillar Image looks different, at 6minutes.
or Nvidia OpenCL 1.1 vs. AMD OpenCL 1.2

Catapillar is buggy scene, don't use it

in other softwares like LuxMark 2.0 or ratGPU, i can select any, and they measure different.
so i guess something is wrong with Indigo>Renderer.

No, something is wrong comparing apples with pears here!
Indigo atm only renders in hybride mode and I simply hate it to explain that twice a week

[JuanPabloCuervo]

Indigo Renderer can only use 1 card at a time, Bug#1.

No Bug, Indigo simply only supports one GPU in GPU mode!

Bug#2 ...
they measure exactly the same spp & ks/s.

Maybe no Bug, but your CPU is maybe way older, and can't feed any of your cards with data fast enough, so they are having the same speed since your CPU is to slow to max out the GPUs.

Caterpillar Image looks different, at 6minutes.
or Nvidia OpenCL 1.1 vs. AMD OpenCL 1.2

Catapillar is buggy scene, don't use it

in other softwares like LuxMark 2.0 or ratGPU, i can select any, and they measure different.
so i guess something is wrong with Indigo>Renderer.

No, something is wrong comparing apples with pears here!
Indigo atm only renders in hybride mode and I simply hate it to explain that twice a week

#1. my CPU is Rank# 53 stock.
http://cpubenchmark.net/cpu.php?cpu=Int ... GHz&id=840
& Top12 Overclocked.
http://cpubenchmark.net/overclocked_cpus.html

#2. i think it´s got something to do with the Render Size.
but...
#3. when i test my GPUs with ratgpu.com benchmark, balls scene.
i can select the render size to 320x200, with high rays 128 & high bounces 128, it does eat 100% of both GPUs.
also with LuxMark2.0 Benchmark, is small render size, but eats 100% of the GPUs.

so is not the size...

#4. call me crazy, but a free software that supports any amount of GPUS, vs. a $1000usd. software that only supports 1, is not a Bug.?

#5. if "Catapillar is a buggy scene," why is included in the Indigo Render ? instead of your Diamond?

#6. i don´t get the apples vs. pears comparison...
is...
Hybrid CPU+GPU path tracing Benchmark apples?
Hybrid CPU+GPU path tracing Renderer pears?

#7. just a side comment;
LuxMark benchmark has that "Endless Render", like Indigo
ratGPU path tracing render has a user defined quality level, "end".
i like more the user defined end render.

if i select 8 rays 8 bounces ends fast, looks dark & grainy.
if i select 128 rays & 128 bounces it ends slow. looks ok to me.
if i select 8192 rays + 256 bounces, it ends, but don´t know when.

if there was a same scene in both platforms to test render quality vs. time.

[Zom-B]

#6. i don´t get the apples vs. pears comparison...
is...
Hybrid CPU+GPU path tracing Benchmark is apples?
Hybrid CPU+GPU path tracing Renderer is pears?

The problem is the term hybrid here. The Lux benchmark initializes each GPU and CPU and whatever as a own rendering device that renders the whole scene for itself, like network rendering between the components and then merging to a final image.
Indigo uses the GPU only for some parts of the rendering process and for example even don't put any textures in GPU RAM. The GPU (afaik) only does the splatting of samples.

#4. call me crazy, but a free software that supports any amount of GPUS, vs. a $1000usd. software that only supports 1, is a Bug.

Then use the free software if it is that great, there are reasons why people buy and use Indigo

[CTZn]

Both devices reach the same performance because they are using the same CPU base power indeed.

Hybrid as per Indigo means that CPU and GPU are doing different tasks wich happen serially if I'm understanding correctly. The GPU would idle while waiting for the CPU data, then process that data faster than the CPU would before sending it back.

I don't qualify to discuss technical choices though. Only thing I know is that Indigo using 1 device was marked as being a temporary measure.

About comparing softwares, there has been a valuable initiative happening, check http://www.indigorenderer.com/forum/vie ... =7&t=12020. The project would deserve more support I believe.