Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2026 Daz Productions Inc. All Rights Reserved.
Comments
note for me copied from Google AI
To switch from BF16 to FP16 in pre-compiled applications without touching code, you will primarily rely on environment variables, command-line arguments (flags), or configuration files.
Because you cannot change the underlying code, your success depends entirely on whether the app's developers built in a toggle for precision formats.
Here is how you can check for and apply these changes across different types of applications:
1. Command-Line Arguments (Flags)
Many pre-compiled AI tools, loaders, and command-line apps accept flags when you launch them.
2. Environment Variables
Some frameworks (like PyTorch or Hugging Face utilities running inside compiled packages) check your system's environment variables before launching.
3. Configuration Files (.json, .yaml, .ini, .toml)
Look inside the application's installation folder for configuration files.
4. Graphic User Interfaces (GUIs)
If the app has a visual interface (like WebUIs for Stable Diffusion or LLM runners):
⚠️ Important Compatibility Warning
If you are changing this because your hardware does not support BF16 (e.g., older NVIDIA GTX 10-series or 16-series cards), be aware that some older GPUs suffer from a "black image" or "NaN" error when running certain tasks in FP16. If your app crashes or produces corrupted outputs after switching to FP16, you may need to drop down to full FP32 precision instead.
I have actually tried this in notepad myself, my little brain thought of it
I have used notepad to edit other stuff in the past, it's my favourite go to for duf files in DAZ Studio
....yeah still rendering on my trusty old Maxwell12 GB Titan-X (well really about 11 GB as it doesn't have RT cores). Still, far better than CPU rendering. I have a 3060 12 GB stil in the box as the BIOS on the system is so old (and the final update for it came out many years ago) it doesn't recognise it,
Also it is not just the rising costs and availability of GPUs, but memory (to support the card's VRAM during rendering), which has skyrocketed by about six times what it was a year ago. So it's sort of a "double whammy", particularly if you need to upgrade for whatever reason
Your mobo must be really old if there's no more bios updates. The 3060 was only released 5 years ago.
Do some reading about INT8 quantization and test run any INT8 checkpoint (image or video) if you haven't already. A quick google of your specs and what quantization to use got this response from Gemini:
If the checkpoints don't fit in vram, they often get offloaded to ram, then disk, each at least an order of magnitude slower than the last. If the model barely fits in one, you can run it but it'll probably spend a ton of time swapping blocks back and forth from these memory locations, making each generation take much longer than if you could find a model that fits wholly in vram or ram. If your motherboard has integrated graphics you could run your main monitor from that to squeeze every bit of performance out of the dedicated gpu. Ideogram 4 for image generation, for example, has an INT8 quantization that should fit in 12 GB vram if nothing heavy is competing.
I already run both my monitors off my integrated AMD Vega graphics
I know nothing about coding so limited in what I can figure out
annoyed I updated the program as it worked properly an earlier build
still trying to find one that does, they do have some earlier versions but apparently update some
couldn't find anything to do with this to edit in notepad either
I recommend ask chatgpt or google ai, just explain what's going on and what's your goal, and ask it to elaborate when it's not clear or understood. Local image and video ai checkpoints are often fp8 or int8/convrot, your card would likely run better with the int8 checkpoints. fp8 isn't optimal for your card, it spends extra time crunching numbers, but your card has acceleration features for int8. A model like Ideogram 4 will give you separate ~8GB downloads for fp8 or int8/convrot and whichever one you choose has the number format baked in. Getting set up to run it in comfyui or pinokio is a whole other process.
Edit: I don't know why I recommended ideogram as the test model, it's cool but Anima is easier to prompt and even smaller (4GB) with a 1 click pinokio install. It doesn't really matter which model to use as a test, just whatever's easy to see if int8 makes it fast enough to be fun.
I have been asking AI
thats what the quote in my previous post was
I do have things that work in Pinokio but its hit and miss
and using ComfyUI honestly would be even harder
I know my skill limits
if the situation of this thread title was not an issue
I would do the simplest thhing and buy new hardware
...Asus P6T X58 with LGA 1366 socket, PCIe 2.0 expansion slots, 6 DDR3 DIMM slots (max 24 GB) and TPM 1,2 headers. The latter is why I need to replace the MB (along with the CPU, CPU cooler, and memory and boot drive) to move to Win 11.
The MB uses Legacy BIOSm not UEFI with SecureBoot without which running 10 could result in instabilities (including boot crashes) due to other (older) hardware issues..
The MB was originally released in 2008, I purchased the one I have in 2011.
Wow! 2008! DDR3 too. It's a museum piece
A half decent budget upgrade could be the MSI A520M-A PRO (around 50-60 Euros and supports TPM 2.0) along with some DDR4 RAM. Given current prices, I'd suspect a newer mobo and DDR5 would be a stretch on the wallet.
I'm still rockin a 2012 macbook pro as one of my daily drivers. No, it can't run Daz Studio. :) It can't even run OSX, so I put Linux on it. Unironically, cloud ai increases the value of such an old laptop since I can still get real work done.
That's odd, I have a 21-inch 2011 iMac and a 2012 27-inch iMac, both running DAZ Studio with OSX; I use them for doing any DAZ Studio lip sync. Your 2012 MacBook Pro can't run the latest OSX, but I've been running OSX since 2001 so it should come with OSX installed when you bought it and I believe it can run up to Catalina (10.15).
As to Cloud-based AI, heck ya, I use it to make images and animations on my very low-end iPad.
I cannot help but marvel at the idea that hardware prices are through the roof because of the ai build out, and yet it seems the bulk of my audience here wants cheaper hardware to more efficiently do local ai stuff. I'd like cheaper hardware to render my Daz stuff faster and also to make sure I have a path forward for hard drives and such when needed and for a new system when my current system eventually fails. Bother!
I want both actually
I generally do both, 3D rendering with AI editing
I figured that.
Yea, these big boys that have basically left us in the dirt will hopefully regret their decision to dump most of their revenue and tech knowledge into the big data centers when they will all of a sudden realize that the masses can't use any of it because their own systems aren't up to snuff and can't afford the outrageous prices of getting better parts for their rigs or even a whole new system. I hope my current system holds out until this glass ceiling shatters.
No, they're hoping you'll pay to rent a render farm to stream call of duty to lightweight game system. Then you don't need a system. You just need high speed broadband and a monthly rendering bill.
Doing more with less! Sounds like good advice when times are tough.
Cheers!
Easy for you to say, but there are folks on this board in far worse position than I am and this pricing issue is squeezing them right out of the game. It's not doing me any favors either.
As you surely know, 3d software drives a treadmill that constantly demands more powerful hardware over time. AI models clearly have the same hunger and the effects of that have been discussed here already.
You might consider that people really don't appreciate being told to do more with less, especially when the price of everything keeps going up and up. And the little smiley face does not make it more acceptable.
Let's say they make $50 for every consume GPU sold. There would need to be 20,000,000 lost sales for them to lose $1 billion, which is less than their existing contracts for data centers. The everyday consumer's "squeeze" is less than nothing to them.
...I put a configuration using 64 GB DDR4 and even re memory alone has shot up in price from about 100 - 120 USD to over 600 USD for what is considered legacy hardware, Also most of the major "reputable" manufacturers have moved . to DDR5 leaving companies I never heard of to fill the void.. So even a "stopgap" solution is pricey
They make a lot more than that per GPU sold. Try 40%.
I was being conservative to make my point. They don't care about the consumer market. And it doesn't make business sense if ROI is your goddess.
I think folks should look at this:
https://isaiprofitable.com/
It tells its own story. I have more to say but it'll have to wait.
Add one item: go to your favorite search engine and type in amount nvidia has loaned to ai companies
Good morning! Ohayou gozaimasu! Buongiorno! Magandang umaga! Bonjour!
Nothing like a good night's worth of sleep to rebuild that energy ... er, positve energy.
I'll have to take a rain check on this discussion for now. There are other folks in the Daz Forum I'm having an interesting conversation with. No, its not about the the smiley face
and whether the emoji is kosher or not. Let me reply to them first.
Cheers!
I programmed some stuff in C for DOS, just a simple machining utility for calculating the plunge depth of a tool on a lathe to follow an ogive shape (in all honesty it's wierd to see when you use it on a lathe). With all the libraries I needed, the program was 27kb and it outputted to screen and a text file..
I then used C++ Builder 4 to do a simple C++ program to do the same calculations and output the result into a listbox and save the text file. 32 bit Program size 2.3Mb
I used C++ Builder 10 to recompile the 32bit C++ program code into 64bit using the updated libraries. 64 bit program size: 27Mb. Or 1000x bigger than the DOS program that did the same job.
Regards,
Richard
Wow, C++ Builder. So long ago.
Remember when 640 KB wa all anyone should ever need?
I was just thinking about my first desktop...1mb ram and a 20mb hard drive. I bought it for second hand in 1992 for NZ$1200 with a b&w vga screen. Today I can buy four Dell refurbished pcs (8gb ram, 256 ssd, i5) for the same amount. Four! That $800 I'm planning to spend on a new 8tb hard drive doesn't sound so bad.
Yeah, mybad, I meant latest/modern OSX. The last version it could run is 7 versions behind the current. It used to have Yosemite before I put linux on it, that's 12 versions back. I could probably put Catalina on it to get Daz Studio running, but it wouldn't be much use for iray stuff due to system specs. I'd eventually like Daz Studio running in Linux but I'd use a newer workstation for that. (I read some folks have got it working, at least!)