Iray Rendering on CPU Only, and It's Overheating
I have a bad problem! Iray is rendering on my CPU only, and that's causing superheating. This just started two days ago.
My video card is a GTX-1060.
I'm using Studio 4.10.
My test scene is three G3Fs, nude, with hair, Anagenesis 2 skin system, and an HDRI. Low resource requirements.
Windows diagnostics shows the video card is working perfectly. No sluggishness with my browser, games, or videos. Only issue is with DS.
I installed the latest nVidia driver, with no effect to GPU non-usage.
Video card monitor and Windows Task Manager both show the GPU is not being used for the render, and Task Manager shows CPU is at 100% usage.
Originally, I had all the boxes checked in Render Settings > Advanced. I unchecked CPU, with no effect to GPU non-usage.
This is what the log is showing me:
2018-04-18 01:44:31.866 Iray INFO - module:category(IRAY:RENDER): 1.0 IRAY rend info : Rendering...
2018-04-18 01:44:31.866 Iray VERBOSE - module:category(IRAY:RENDER): 1.9 IRAY rend progr: CUDA device 0 (GeForce GTX 1060): Processing scene...
2018-04-18 01:44:32.066 Iray VERBOSE - module:category(IRAY:RENDER): 1.6 IRAY rend stat : Geometry memory consumption: 889.6 MiB (device 0), 0 B (host)
2018-04-18 01:44:32.316 WARNING: dzneuraymgr.cpp(307): Iray ERROR - module:category(IRAY:RENDER): 1.6 IRAY rend error: OptiX Prime error (Device rtpModelUpdate TL): Unknown error (Function "_rtpModelUpdate" caught exception: Encountered a CUDA error: cudaMalloc(&ptr, size) returned (2): out of memory, [15466617])
2018-04-18 01:44:32.316 WARNING: dzneuraymgr.cpp(307): Iray ERROR - module:category(IRAY:RENDER): 1.9 IRAY rend error: CUDA device 0 (GeForce GTX 1060): Scene setup failed
2018-04-18 01:44:32.316 WARNING: dzneuraymgr.cpp(307): Iray ERROR - module:category(IRAY:RENDER): 1.9 IRAY rend error: CUDA device 0 (GeForce GTX 1060): Device failed while rendering
2018-04-18 01:44:32.316 WARNING: dzneuraymgr.cpp(307): Iray WARNING - module:category(IRAY:RENDER): 1.9 IRAY rend warn : All available GPUs failed.
2018-04-18 01:44:32.316 Iray INFO - module:category(IRAY:RENDER): 1.9 IRAY rend info : Falling back to CPU rendering.
2018-04-18 01:44:32.316 WARNING: dzneuraymgr.cpp(307): Iray ERROR - module:category(IRAY:RENDER): 1.9 IRAY rend error: All workers failed: aborting render
2018-04-18 01:44:32.316 Iray INFO - module:category(IRAY:RENDER): 1.0 IRAY rend info : CPU: using 8 cores for rendering
GPU monitor showed that at the point the render began (the point that Iray switched to CPU) that of 6G available VRAM, only 2.8G had been used. That seems to be a massive amount of memory usage for three nudes with hair and no set. Even so, there is more than 3G VRAM available to handle the render.
I rendered much more complicated scenes with more characters, full clothes, full sets, and tons of metal only several days ago, on the GPU, with no issues.
How do I fix this?
Comments
A Kaby Lake i7.
Your log includes "OptiX Prime error". Try unchecking the box for Optix Prime in Render Settings.
Quad.
Son of a gun! That seemed to do the trick. It's rendering on GPU. A little slow, but my CPU is happy. Any way to fix the OptiX?
You know, I take back what I said about it being slow. Three G3Fs are finishing in 20 minutes. That's fast enough. Thanks for the help!
It sounds like you found what prevented rendering on the GPU, which is good. But you still might want to investigate your CPU cooling situation. Are you using the stock coolers from Intel (or AMD)? Or maybe I should say "coolers".
Last year I built a new FreeNAS file server. I followed the recommendations to thoroughly burn in the cpu and test the memory before handing it a bunch of disks to possibly corrupt if the hardware were bad. I ran the Prime95 test from mersenne.org. There are few things that will put such an instant and non-stop load on all your CPU cores as this test. My CPU went instantly from 34°C to 80°C with the stock Intel cooler. Not good. I got a Cooler-Master Hyper 212 evo with dual fans, and the temps stayed in the 50's. I just finished putting together a new desktop PC to use with Daz. This time I went with the Noctual NH-D15. It's also a beast. I tested the new system and got similar good results. As it happens, I'm building a small mini-ITX system for my daughter, and I'm stress testing it right now. Her case does not have room for one of these coolers, so I'm trying a low-profile Noctual. Temps instantly jumped up to the 50's, then stabilized in the low 60's. That's pretty good.
If you have a stock cooler, you might look into upgrading it. If you case is tight, one of the low profile coolers might still help. And of course, water cooling is pretty popular, and may be even better. But I don't have any experience with that.
Well, your card only has 6GB... (Not that it is "shy on memory"... but...)
If you are on windows 10, you may only have 4GB available, not including what it takes to also drive your video-display. (WDDM v2.3 issue. It reserves, explicitly, up to 2GB of VRAM. Which, when Daz asks IRAY, "how much VRAM is available", it reports back 4GB. Which Daz and IRAY use to "setup a scene to fit into".)
If you use OptiX, that may be consuming more memory, which may not be "accounted for", pushing you over the edge, leading to the "out of memory" error.
Are you rendering with unusual settings?
- High image threshold values (512 and 1024 is default.) {Leading to the use of uncompressed images across the whole project}
- Unusual texture-map images in your model, or excessive image-maps used in multiple areas of the shaders
- High-detail models, with high Sub-Division levels... (Each Sub-D level is almost like tripling the number of models in the scene.)
- High "quality levels" or image-sizes... Going from 1080p size, to 4K is 3x the volume of pixel-data.
- Did you turn-on any of the advanced settings for extra output images?
- Are you running another "GPU" memory-hog, like Chrome-browser or another art program, or 3D program, while trying to render...
As for the CPU issue, I would look into that. You said six-core... Is that one of the new i9 CPUs? They run hotter than normal, because they don't have thermal-epoxy or thermal-weld between the chip and the case. They kind-of depend on self-throttling, to remain cool, even with the best coolers. There is honestly little chance of it "overheating". It will just keep throttling-down and running slower, to sustain a "safe temperature". It will push the 84-90C limits. It wouldn't "fry" until it pushes past 95-105C... That is the limit of the TIM used in the chips. At that point, the heat begins to fold-in on itself, and cascade beyond control. If it hit those temps, and was totally "throttled-down"... The chips can throttle-down (self-underclock), to 5%... If you overheat at that speed. There is something wrong with your cooler, for sure. (It would be clogged with dust or would be poorly "set", to get to that point.)