Sapphire w/ mocha mask - very slow performance in Avid

I have an S_EFFECT over some 1080p DNxHD 145 material (the effect is the X Reveal S_EFFECT preset). This effect plays in real-time on my timeline (even with the Avid cache purged).

Now I add a mocha mask from within the X-reveal effect (just a small area of the image to be masked) and I find that playback is unbelievably slow. I checked my GPU and it is not being used at all and of my 28 cores (14 core x 2 hyperthreading) I noticed that only a single physical core being pushed during playback.

Would be nice if Sapphire could leverage the power of a modern workstation. Is this an Avid issue or a BorisFX issue?

System is an i9-7940 @ 4 GHz, 64 GB RAM, P4000 GPU (drives are M.2 and SSD)


Can you quantify what you mean by slow rendering? E.g. how many frames per second?

The mocha mask part of the rendering happens primarily on the GPU so it would not show up as utilizing the CPU cores. The GPU should show up as being used depending what diagnostics you are using. It is possible that the mocha rendering is encountering a problem on that GPU however and falling back to CPU. Could you provide your mocha log file? You can find it by opening the Mocha prefs pane, selecting the tab with logging controls and choosing the option to open the log.


Hi Jason - I mean that if you play the green-dot timeline it is only playing at 12 fps until Avid caches the rendered frames whereas it plays full 29.97 fps without the mocha mask.

I guess this isn’t all that bad (it seemed to be worse when I tested it earlier). The log from Mocha indicates that it is using the GPU and Windows does show a bit of GPU usage (I was mistaken that it wasn’t using it at all).

The thing is, I have very fast disks, a fast CPU and fast GPU and yet somehow only a small percentage of all of this is being taxed - seems like a single thread is the bottleneck.

Here’s the mocha log:

Apr 26 18:05:42 (mocha-sapphire) Commenced logging at Fri Apr 26 18:05:42 2019
Apr 26 18:05:42 (mocha-sapphire) Build Date Feb 6 2019
Apr 26 18:05:42 (mocha-sapphire) Version 6.0.2 build 185.g3358571b54e9 win_x86_64.
Apr 26 18:05:42 (mocha-sapphire) Windows 10 64-bit
Apr 26 18:05:43 (mocha-sapphire) Style sheets directory: :/stylesheets
Apr 26 18:05:43 (mocha-sapphire) Locale: default
Apr 26 18:05:43 (mocha-sapphire) Setting up 226 actions for usage logging
Apr 26 18:05:43 (mocha-sapphire) Setting up GL widget WITH vertical sync
Apr 26 18:05:43 (mocha-sapphire) **************** GPU Information ****************
Apr 26 18:05:43 (mocha-sapphire) Platform:
Apr 26 18:05:43 (mocha-sapphire) Profile: FULL_PROFILE
Apr 26 18:05:43 (mocha-sapphire) Version: OpenCL 1.2 CUDA 10.1.120
Apr 26 18:05:43 (mocha-sapphire) Name: NVIDIA CUDA
Apr 26 18:05:43 (mocha-sapphire) Vendor: NVIDIA Corporation
Apr 26 18:05:43 (mocha-sapphire) Extensions: cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_fp64 cl_khr_byte_addressable_store cl_khr_icd cl_khr_gl_sharing cl_nv_compiler_options cl_nv_device_attribute_query cl_nv_pragma_unroll cl_nv_d3d10_sharing cl_khr_d3d10_sharing cl_nv_d3d11_sharing cl_nv_copy_opts cl_khr_gl_event cl_nv_create_buffer
Apr 26 18:05:43 (mocha-sapphire)
Apr 26 18:05:43 (mocha-sapphire) Device:
Apr 26 18:05:43 (mocha-sapphire) Name: Quadro P4000
Apr 26 18:05:43 (mocha-sapphire) Vendor: NVIDIA Corporation
Apr 26 18:05:43 (mocha-sapphire) Version: OpenCL 1.2 CUDA
Apr 26 18:05:43 (mocha-sapphire) Driver version: 430.39
Apr 26 18:05:43 (mocha-sapphire) Global memory size: Not Available B
Apr 26 18:05:43 (mocha-sapphire) Local memory size: 48.0 KiB
Apr 26 18:05:43 (mocha-sapphire) Number of computing units: 14
Apr 26 18:05:43 (mocha-sapphire) Maximum work group size: 1024
Apr 26 18:05:43 (mocha-sapphire) Maximum work item dimension: 3
Apr 26 18:05:43 (mocha-sapphire) Maximum work item sizes: (1024, 1024, 64)
Apr 26 18:05:43 (mocha-sapphire) Maximum clock frequency: 1480 MHz
Apr 26 18:05:43 (mocha-sapphire) Available for tracking: yes
Apr 26 18:05:43 (mocha-sapphire) Available for remove: yes
Apr 26 18:05:43 (mocha-sapphire) *************************************************
Apr 26 18:05:43 (mocha-sapphire) Using at most 80 percent of system memory


Hi, Steve,

At the moment I think that is just where we are at in terms of performance when rendering mocha masks in Continuum or Sapphire. There is definitely a performance hit from the mocha mask rasterization. It is GPU enabled already, but I know this is something we have been hoping to improve upon further in a future update.


Thanks, Jason.

Hi, Steve,

As announced at NAB, you should see performance gains with the mocha mask rendering in the 2019.5 updates slated for release early this summer.


Cool. Looking forward to it :slight_smile: