GTX 1080 no compatible with OCTANE V3!??

[zeigg]

Care to post some test results of 1080 (ideally at 2GHz clock) vs 780Ti? I have 780Ti at work and was looking at 1080, even possibly 2, so i am really curious how they compare.

When I run the octane_22_benchmark.ocs scene on pathtracing I get between 5.35 - 5.41 Ms/sec.
This is using a single GTX 1080 Founder's Edition (From NVIDIA, not a partner card) with no overclocking or modifications.

Compare this to your GTX 780 Ti setup by running the same test.

For more GTX 1080 tests see here:

Ok each one took a while but... All at 16000 Max Samples:

DL = ~37 Ms/sec
PT = ~6 MS/sec
PMC = ~9.5 MS/sec

Here's a vid of me running it... unfortunately I didn't let it render all the way, I mostly wanted to test workflow feedback and not final render times.

Please remember that this is still on the unoptimized hybrid CUDA 7.5/8.0 version of Octane 3.
I hope this answers any questions you have.

[mbutler2]

So, slower than a 980TI.

Not optimized, I understand of course.

[yalbaiz]

Hello everyone,

I'm back with some questions for those of you who have a lot more experience with Octane than myself.

A couple of weeks ago I posted a question about hardware upgrade options and asked for your opinions. I received a few responses, and they were helpful and greatly appreciated. Now I am here to report that I have upgraded my hardware as follows:

CPU: i7-6900K 3.20 GHz (8 Cores - 16 threads)

GPUs: 2x 4095MB NVIDIA GeForce GTX 1070 (EVGA)
1x 1535MB NVIDIA GeForce GTX 590 (EVGA) (2 internal GPUs so total VRAM is 1535MBx2)

System RAM: 64 GB DDR4

Storage: 1 TB SSD

I was wondering if having my old GTX 590 in there has any negative effects like slowing down renders or any other issues. And does anyone have any suggestions on the best way to configure the above setup to best work with Octane.
Note: (I am planning on replacing the GTX 590 with another GTX 1070 soon)

Thank you all in advance for whatever feedback you might give,
Yarob.

[bepeg4d]

Hi yalbaiz,
1070 should have 8GB of VRAM, not only four, while 590 has only 1.5GB of VRAM. 590 has a very good score of 100 in OctaneBench, but it is also very hungry in power consumption, roughly a 590 is comparable to 3x 1070
My suggestion is to use the 590 for display only, for now, so you can use all the 8GB for Octane, and if the scene is very simple and fits in few VRAM, you can add the 590 on the fly in the Devices preferences
ciao beppe

[yalbaiz]

Hi yalbaiz,
1070 should have 8GB of VRAM, not only four, while 590 has only 1.5GB of VRAM. 590 has a very good score of 100 in OctaneBench, but it is also very hungry in power consumption, roughly a 590 is comparable to 3x 1070
My suggestion is to use the 590 for display only, for now, so you can use all the 8GB for Octane, and if the scene is very simple and fits in few VRAM, you can add the 590 on the fly in the Devices preferences
ciao beppe

Thanks for the info and response....
One strange thing I noticed... In Octane it only shows that I have 1.5 GB of VRam (At the bottom right of the render view), even though I set the Priority check marks for the two 1070s. Iam not sure if that is just a display bug or it only actually sees 1.5 GB VRam.

To confirm what you mentioned about GTX 590, I used the Benchmarking tool provided by OTOY (which only recognized my GTX590, the 2 1070s were greyed out), I did the test and I got a 109 score.

But still in the Standalone version of Octane, the VRam indicated while rendering (bottom right of render view) shows only 1.5 GB. In the devices list in settings however, all GPUs are listed with proper VRam indicated.

I tested the Benchmark scene called "Idea" in the standalone version, and below are the times it took to render the scene:

Idea Scene - DL method - 1m 49s
Idea Scene - Path Tracing method - 2m 35s
Idea Scene - PMC method - 6m 11s

Does that sound right to you?
PS: Can you please let me know how to setup my GTX 590 to be only for display? Or do you mean just connect my display to that GPU's output and not to any other output?

Thanks.

[nuno1980]

I tested the Benchmark scene called "Idea" in the standalone version, and below are the times it took to render the scene:

Idea Scene - DL method - 1m 49s
Idea Scene - Path Tracing method - 2m 35s
Idea Scene - PMC method - 6m 11s

Were 2x GTX1070 benchmarking? At how many samples per pixel? 16000?

[mbutler2]

I did some testing of a slow card (970) combined with a few fast cards (4 x 980TI). Here's the thread:

viewtopic.php?f=9&t=54121&p=274605#p274605

In my setup, basically the slow card isn't even worth having, it adds so little to the overall processing. (Slow card: "Hey guys, Can I come too?! Please, please?") It didn't slow anything down, but having little brother tag along didn't add anything significant. Only reason to have it was to drive a monitor, but it was taking up a slot I could put a faster card in.

So I sold that 970 before the price of them starts to drop too much.

As far as best card going right now, I'm still a fan of the 980TI for price/performance/ track record. At this moment its still faster than the 1080 for Octane, and cheaper.

[yalbaiz]

I tested the Benchmark scene called "Idea" in the standalone version, and below are the times it took to render the scene:

Idea Scene - DL method - 1m 49s
Idea Scene - Path Tracing method - 2m 35s
Idea Scene - PMC method - 6m 11s

Were 2x GTX1070 benchmarking? At how many samples per pixel? 16000?

Hi, nuno1980,

As I mentioned in my post, Octanebench_2_17 does not allow me to test GTX 1070s (they are greyed out in the GPU selection and I can not use them for the test).
Maybe its just the Benchmarking tool that has this limitation i am not sure, but they are detected and used in the standalone version 3.03

The test results I did using the standalone version (not the benchmarking tool) using the same scene (Idea) are as follows:

(All GPUs - 2x GTX 1070 + 1 GTX 590) - Max Samples 16000 - Idea Scene - DL method - 2m 49s - 56.5 Ms/Sec Avg.
(Only 2x GTX 1070) - Max Samples 16000 - Idea Scene - DL method - 4m 00s - 40 Ms/Sec Avg.

I do not understand why it took longer to render the same scene in the standalone version (2m 49s) than it did in the benchmarking tool (1m 49s). Maybe it actually did use the GTX 1070s in the tool but indicated they were not used.

I also noticed that in the C4D plugin, if I turn off the GTX 590 and choose only the two GTX 1070s, I get an error message saying "There is no CUDA Device which is selected..."
So I guess that in the C4D plugin the new Pascal architecture is still not supported (I'm using plugin version 3.02 - R2)

Thank you,
Yarob.

Thank you,
Yarob.

[yalbaiz]

Just a quick update, I upgraded the C4D plugin to 3.03-R5 and now it works with only GTX 1070s selected.

Just thought I should mention that,

Thank you.

[mbutler2]

Yeah, the benchmark scores do seem to differ from real world significantly.

The benchmark scores say that a 980 TI is 1.5 times faster than a 970, but I found it to be way more. 2.5 times faster in my own personal test scenes.

Surely benchmarks are using a broader range of tests, mine are for my specific usage scenarios.