Neat Video & Neat Image community

Hello,

in the announcement of the new Neat Image version 7.0, you are talking about "CUDA support" and about "GPU support". My question is: Is there a speed improvement for NVIDIA cards only (CUDA) or will I, as an owner of an AMD/ATI card (Sapphire Toxic HD 4870), experience improved speed, too?

Thanks in advance!

Cheers,
Jan

Currently Neat Image uses CUDA-capable GPUs that happen to be provided by NVidia only. So, unfortunately, an ATI card will not be able to help in this situation.

I hope we will be able to support ATI cards too, in the future updates of Neat Image.

Still, you will observe speed improvements in this new version, thanks to additional optimizations and improved algorithms. Also, do not forget to look in the Performance tab in Neat Image Preferences. There is that Optimize button, which will help you optimize CPU core utilization. That is especially useful on multi-core processors (like i7).

Kind regards,
Vlad

Thank you very much for your quick answer!

NITeam wrote:Currently Neat Image uses CUDA-capable GPUs that happen to be provided by NVidia only. So, unfortunately, an ATI card will not be able to help in this situation.

I hope we will be able to support ATI cards too, in the future updates of Neat Image.

That would be great. Is this planned for 7.x versions or only just for later major versions?

NITeam wrote:Still, you will observe speed improvements in this new version, thanks to additional optimizations and improved algorithms. Also, do not forget to look into Performance tab in Neat Image Preferences. There is that Optimize button, which will help you optimize CPU core utilization. That is especially useful on multi-core processors (like i7).

Okay, I guess I'm convinced. As I happen to use a Core i7, a speed improvement should be noticeable.

Cheers,
Jan

We plan to test the ATI GPUs and if they will show significant speed up as compared with modern CPUs in our noise reduction algorithms (in some types of calculations, GPUs are not faster than CPUs) then we will try to employ ATI GPUs as well. Hopefully in one of 7.x versions.

Regarding i7, please try the Optimize function, you may be surprised. Please post the log from that test.

Thank you,
Vlad

In the meantime I purchased the upgrade to the v7 Pro Photoshop plug-in, and I'm very content with the new version.

NITeam wrote:Regarding i7, please try the Optimize function, you may be surprised. Please post the log from that test.

Sure, here it is (tested while a bunch of other programs was open):

Detecting the best combination of performance settings:
running the test data set on up to 8 CPU cores

1 core: 2.78 MPix/sec
2 cores: 5.47 MPix/sec
3 cores: 6.17 MPix/sec
4 cores: 7.95 MPix/sec
5 cores: 8.18 MPix/sec
6 cores: 8.32 MPix/sec
7 cores: 8.8 MPix/sec
8 cores: 8.64 MPix/sec

Best combination: 7 cores

I set it to 4 cores - that should be enough performance and will leave sufficient power for other applications.

Cheers,
Jan

The Benchmark for an AMD 1055T:

1 core: 2.67 MPix/sec
2 cores: 5.02 MPix/sec
3 cores: 6.89 MPix/sec
4 cores: 7.73 MPix/sec
5 cores: 8.89 MPix/sec
6 cores: 9.75 MPix/sec

Interesting to see that it scales quite differently compared to the i7

Hoogo wrote:Interesting to see that it scales quite differently compared to the i7

Well, as I said a bunch of other programs was running while I performed the test, and the values differed when repeating it, so I wouldn't attach too much significance to this benchmark result.

Cheers,
Jan

Same here. I just retried with a music player and a video running in background.

But I just remembered that the i7 has 4 real + 4 HT-cores. I guess that's why it's only slightly faster with more than 4 cores.

On 64-bit Linux, Neat Image seems to be looking in the wrong location for libcuda.so and libOpenCL.so. I'm running Neat Image 7.3 64-bit Linux version on a Linux Mint 14 machine with an nVidia GTS250 card running the latest nVidia driver. The Performance tab of the Preferences dialog tells me "No computation-capable GPU devices found", and the detection log shows:

Looking for NVIDIA CUDA-capable devices...
Failed to load CUDA driver ("/usr/lib64/libcuda.so")
If you use an NVIDIA card, please install the latest video driver with CUDA support.

Looking for AMD OpenCL-capable devices...
Failed to load OpenCL library ("/usr/lib64/libOpenCL.so")
If you use an AMD card, please install the latest AMD driver with OpenCL support.

Both these library files are installed and are in /usr/lib/nvidia-current. There is no /usr/lib64 directory.

GPU processing works fine with Autopano Pro on the same machine.

Nick Payne wrote:On 64-bit Linux, Neat Image seems to be looking in the wrong location for libcuda.so and libOpenCL.so.

Please contact us at support [at] neatimage [dot] com.