Looking around Octane Render again ... For Poser.

jorge dorlando

Hi,
Ok, I know there are threads here about Octane, and much discussed.
I've also passed on some topics, I had answers about octane, and at the time I gave up to acquire Octane, because of the issue of memories video card, which in a broad number of characters, is tight for a videocard with 3 or 4GB of VRAM.

But what's making me re-consider Octane, is this:
Zotac Nvidia Gtx Titan Black 6GB DDR5 384bits
"6gb of ram and Cuda Cores: 2880"
2880 Cuda cores, is = 360 CPUs, intel i7 quadricore ('ve already included virtualized kernels)
between poser render a scene in Vue, using only 8 cores of intel, and render the same poser scene in Octane using 2880 cores ... It's very ambitious thought!

Note: I did not bought the Gtx Titan Black 6GB DDR5 384bits, not yet.
also not bought the Octane for poser, no.
Well ... I would like to see your suggestions

Mr Gneiss Guy

Don't make the mistake of equating an intel "core" with a cuda "core". They are NOT the same thing and can't be directly compared. An intel core is a general purpose processor to handle whatever code it sent it's way in an "execution pipeline" , in other words, give it a complex task, it starts at the beginning and runs till that instruction is ended. These instructions can be anything that your computer can do, so it has to be capable of handling whatever instruction it gets. A cuda core or any video card core is designed to solve a chunk of math that the processor and operating system/driver hands it, and that it all it does. That is why the individual video card cores are so much smaller, they are MUCH simpler processors. Also, general instructions are not particularly easy to parallel (spread over multiple cores) and are hard to program for multi-threading. The types of math instructions that video card cores get tasked with are very easy to multi-thread.

jorge dorlando

Mr Gneiss Guy said:

Don't make the mistake of equating an intel "core" with a cuda "core". They are NOT the same thing and can't be directly compared. An intel core is a general purpose processor to handle whatever code it sent it's way in an "execution pipeline" , in other words, give it a complex task, it starts at the beginning and runs till that instruction is ended. These instructions can be anything that your computer can do, so it has to be capable of handling whatever instruction it gets. A cuda core or any video card core is designed to solve a chunk of math that the processor and operating system/driver hands it, and that it all it does. That is why the individual video card cores are so much smaller, they are MUCH simpler processors. Also, general instructions are not particularly easy to parallel (spread over multiple cores) and are hard to program for multi-threading. The types of math instructions that video card cores get tasked with are very easy to multi-thread.

Ooh, no! The comparison I made is only as to render.
I'm aware that I need the processor (CPU) to keep the poser running, and keep the characters to the area of the scene poser, for then Octane able to access them, to render GPU ...
The system also will not run without processor ...

jorge dorlando

Now ... I'm thinking of buying this card here:
http://produto.mercadolivre.com.br/MLB-556745919-placa-de-video-gtx-titan-hydro-cooper-signature-_JM
and make her a aircooler, instead of watercooler. replacing the + box cooler, this gtx780:
http://produto.mercadolivre.com.br/MLB-554714441-caixa-cooler-com-dissipador-gtx-780-evga-_JM
Because The PCB is the same as GTX Titan too, with an 8-pin and 6-pin PCI-E power connector

It’s comfortably the most covetable stock GPU design we can recall, but seems likely to irk some of Nvidia’s GPU partners; we highly doubt we’ll see the same array of third-party coolers as usual and other than the radial fan’s motor there’s little-to-no space to put branding stickers. With the huge variability in after-market coolers though, we’re pretty happy Nvidia has started to invest in its stock models to such a degree, a trend that looks set to continue into the rest of the 7-series. All that said, though, as we found in our recent look at after-market graphics card coolers, there is often head room for reducing operating temperatures and noise levels, so we'll be interested to see if any partners do attempt some more exotic solutions.

With a similar GK110 GPU and the same PCB and cooler as GTX Titan, it’s not surprising to find that GTX 780 matches it with power requirements. A TDP of 250W is 55W higher than that of GTX 680, but with the significantly uprated cooling it’s unlikely to cause problems.

So, GTX 780 is in truth, a slightly neutered GTX Titan, but considering Titan’s monster performance, this still makes GTX 780 a formidable card on paper

Note: source: http://www.bit-tech.net/hardware/2013/05/23/geforce-gtx-780-review/1

DustRider

jorge dorlando said:

Mr Gneiss Guy said:

Don't make the mistake of equating an intel "core" with a cuda "core". They are NOT the same thing and can't be directly compared. An intel core is a general purpose processor to handle whatever code it sent it's way in an "execution pipeline" , in other words, give it a complex task, it starts at the beginning and runs till that instruction is ended. These instructions can be anything that your computer can do, so it has to be capable of handling whatever instruction it gets. A cuda core or any video card core is designed to solve a chunk of math that the processor and operating system/driver hands it, and that it all it does. That is why the individual video card cores are so much smaller, they are MUCH simpler processors. Also, general instructions are not particularly easy to parallel (spread over multiple cores) and are hard to program for multi-threading. The types of math instructions that video card cores get tasked with are very easy to multi-thread.

Ooh, no! The comparison I made is only as to render.
I'm aware that I need the processor (CPU) to keep the poser running, and keep the characters to the area of the scene poser, for then Octane able to access them, to render GPU ...
The system also will not run without processor ...
The direct correlation of performance from CPU to GPU is actually quite complicated. It is more dependent on the number of instructions per clock cycle than per core, but there are other things that go into the mix that really make a true 1 to 1 comparison difficult. But, regardless of the exact correlation, your right, with a Titan you are using an insane amount of computing power for rendering. It's like having a render farm in your PC.

That;'s why Octane is able to get biased render performance with unbiased rendering (or in the case of the Direct Lighting kernal, nearly unbiased rendering). I know of one person that is using a rig with 4 Titans with insane render speeds. But, like anything else in 3D, when you can render faster, your start making more complex shaders and scenes, then render speeds slow down again.

With a Titan your biggest limitation would probably be the number of texture maps you can use (144 RGBA textures, 68 grayscale textures, 10 HDR RGBA textures, and 10 HDR grayscale). The plugins and Octane do a great job of texture/memory management. I've had scenes that took over 5Gb of RAM in DS use less than 2GB of RAM to render in Octane.

jorge dorlando

dustrider said:

jorge dorlando said:

Mr Gneiss Guy said:

Don't make the mistake of equating an intel "core" with a cuda "core". They are NOT the same thing and can't be directly compared. An intel core is a general purpose processor to handle whatever code it sent it's way in an "execution pipeline" , in other words, give it a complex task, it starts at the beginning and runs till that instruction is ended. These instructions can be anything that your computer can do, so it has to be capable of handling whatever instruction it gets. A cuda core or any video card core is designed to solve a chunk of math that the processor and operating system/driver hands it, and that it all it does. That is why the individual video card cores are so much smaller, they are MUCH simpler processors. Also, general instructions are not particularly easy to parallel (spread over multiple cores) and are hard to program for multi-threading. The types of math instructions that video card cores get tasked with are very easy to multi-thread.

Ooh, no! The comparison I made is only as to render.
I'm aware that I need the processor (CPU) to keep the poser running, and keep the characters to the area of the scene poser, for then Octane able to access them, to render GPU ...
The system also will not run without processor ...

With a Titan your biggest limitation would probably be the number of texture maps you can use (144 RGBA textures, 68 grayscale textures, 10 HDR RGBA textures, and 10 HDR grayscale). The plugins and Octane do a great job of texture/memory management. I've had scenes that took over 5Gb of RAM in DS use less than 2GB of RAM to render in Octane.
I did not know that.

jorge dorlando

I’ve had scenes that took over 5Gb of RAM in DS use less than 2GB of RAM to render in Octane.

is very good to know that octane, reduces GB Scenes