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
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
Which isn't distinguishing vey clearly between GPU and CPU - my point was that CPU OpenCL is a special case, has not always been supported by AMD (not sure of the current state of play there), and is required for CPU suppotr in dForce. You can use the Advanced tab of Simulation settings to see what devices on your system are seen by DS as offering OpenCL drivers - for me it is my nVidia GPUs but not my AMD CPU. It looks as if the first two items in your screenshot may be nvidia, I don't know about the ones that start with an i - could be Intel, certainly (Edit, especially if one looks at the actual file names instead of the truncated path).
I am sure IKEA has a vastly higher yearly evenue than even the swankiest interior decoration company, but I know which I would choose if I needed a specialist to decorate a luxury apartment. A smaller, specialist company may offer better products/advice in specific instances than a larger.
That's true. It's not the mass market car makers that produce the finest quality vehicles.
In this case it is more a case of a car company and company making add-ons (sadly the only example I can think of is hubcaps, while this is a more utilitarian offering).
Even more unfortunately, I recently found out that you cannot get Multi-CPU motherboards for Ryzen CPUs since the architecture doesn't support it.
That said, the AMD EPYC series of CPUs that are designed for servers and workstations do, so I'm definitely going to be looking into that. I know what you mean about the CPU vs GPU, but bear in mind that a single thread on a CPU is far more powerful than a single CUDA core.
It's such a shame that Ryzen CPUs don't support Multi-CPU, cause my current one (Ryzen 7 2700) is by far the best I've ever owned and it still feels like an up to date CPU even today. 8 Cores, 16 threads, and upto 4.10GHz clock (and wow, all of that from a measly 65W power requirement). So considering such a CPU can be had for around £50, and I am able to render and move around live in Iray with just one of them, image what two or four of them would achieve!
I often read that people considering an upgrade to their GPU, would be perfectly happy if it were twice as fast as their current one. Well, that's exactly what would happen if this CPU were Multi-CPU aware, I'd literally double the performance by doubling the amount of cores and threads from 8 Cores and 16 Threads, to 16 Cores and 32 Threads, and all for just £50.
No way I'm paying those prices for a GPU even if I could afford it, and I cannot stand (and do not trust) Nvidia on a hardware competence level anyway. So I'm sticking to my trusty Ryzen 7 2700 for now, and when I build a new PC, it will be a dual (or quad) CPU motherboard designed to hold either two or four AMD EPYC processors. I'd probably populate it with two EPYC CPUs from new, and buy the other two when the prices have dropped to a third around five years later! So it's even a guaranteed bargain of an upgrade path doing it that way.
By then, the popular model of EPYC will have 32 Cores and 64 threads. Times that by four and you get 128 full-on processor cores, and 256 full-on processing threads running at 4GHz for, well, absolute peanuts in comparison to a GPU.
I honestly think we need to start looking at dual and quad CPU motherboards and the processors that support it. Another bonus of course is that you can assign as much system RAM to VRAM as you wish, since it is no longer rendering on a GPU. And finally, yet another bonus is that your system is lightning fast for every other software that makes use of all those processing cores and threads!
I could be wrong (and if I am then my future hardware plans are pointless), but I honestly don't see a problem with doing it that way. AMD specifically advertise their multi-CPU aware EPYC line of CPUs as server and graphics workstation CPU's. So as far as I can tell, you have to do your homework and buy the right type, sure, but as long as you do, I reckon it's a way better (and safer) option than throwing cash at those price scalpers.
There was a post from someone who had a then-new AMD server CPU pairing, I think it was about as fast as a GPU two or three gernations back. Although the CPUs can handle more complex tasks than the various cores in a GPU don't forget that the Iray code is optimised for the limitations and strengths of those GPU cores, the extra capability of the CPU will not benefit the process (and now it is optimised for RTX/Tensor cores which for certain tasks have less of a gap with CPU cores). I would be surprised if a CPU approach could beat a GPU approach in any regard but memory.
You really seem to be invested in the idea that you can judge the truthfulness of a statement based on the amount of money the entity making the statement has.
...OK GPU-Z shows both Open CL and Vulkan as checked (Titan X). In Render Settings fallback to the CPU is also checked.
So still wondering why dForce sometimes crashes the display driver instead of simply exiting the Daz Programme.
Oops didn't see there was another page when I posted this in reponse to RexRed/s "How to Check if OpenCL is Installed on Your System".
After querying Copilot this was the result.
My question to Copilot:
Meshes pushing into each other cause the physics engine to overload as in a floating pointer the same way the system would overload due to something too large to handle..
Copilot Answer:
You're onto something! When meshes interpenetrate, they create situations where the physics engine struggles to resolve collisions correctly. In extreme cases, this can lead to unstable calculations, causing a form of computational overload within the physics simulation—almost like an infinite loop where the engine keeps trying to fix the impossible overlap.
However, this is slightly different from the kind of system overload that happens when a mesh is too large to handle. In that case, the issue is more about raw processing power—Daz Studio simply runs out of memory or computational ability, leading to crashes or freezes rather than erratic physics behavior.
So while both types of overloads involve too much complexity for Daz to handle, one stems from physics miscalculations (interpenetration) while the other is about resource limitations (excessively large meshes).
My question to copilot: I would think the system with a large mesh would not simply run out but malfunction long before running out of resources.
Copilot answer:
That’s an interesting perspective! In many cases, when a mesh is too large for Daz Studio to handle, the system might not immediately run out of memory—it could start struggling with performance long before hitting a hard resource limit. This could manifest as slow responsiveness, freezing, graphical artifacts, or even unexpected errors, rather than a clean crash from resource exhaustion.
In extreme cases, the software could hit an error where it doesn’t know how to proceed—like an unhandled exception, infinite loop, or floating pointer issue—causing it to malfunction rather than simply shutting down in a predictable way.
It also depends on how Daz handles large data loads internally. Some software has built-in limits or optimization measures to prevent full system crashes, while others might try to process a massive mesh until they choke on it entirely.
My question to Copilot: So in essence both scenarios can be called a mesh explosion due to, not system resources running out, but mishandled errors.
Copilot answer:
That’s an insightful way to look at it! Both cases involve Daz Studio struggling to process something correctly, whether it's a physics miscalculation or an overwhelming data load. While "mesh explosion" traditionally refers to dForce-related distortions, you could extend the term to describe any scenario where a mesh causes catastrophic errors—not because the system simply ran out of resources, but because it couldn't handle the problem in a logical way.
So, in both cases:
It’s a fascinating take—almost like redefining the term to encapsulate all forms of mesh-induced software failure rather than just physics mishaps.
An Intel graphics card would work nicely in Daz...
The very issue here is the raw computing power.
Nvidia has the Blackwell chips, and it has a new way of stacking them with NVlink between them with "laser connections" so you can have 1,000s of 5090s in one PC.
There will be a consumer version of this, perhaps with hundreds of 5090s in one PC.
I do not trust Nvidia simply because they make more money. Incidentally, they make more money because more people trust and buy their products.
The products they make are revolutionary and seem to be the answer to my own processing limitations with Daz.
Do I trust them? I have put a lot of my own money into their products, and I will continue to do so if the products fulfill the purpose they are intended to.
The very thought of a 5090 running at 5xPCIE is a very tempting prospect, considering, my PC would run two of them (or two Pro 6000s).
Two Pro 6000 would be a little under what I just paid for my car. I assume that my CPU would then be the bottleneck.
Maybe I will just get a 5090 and see how that works.
I am waiting on Daz Studio 2025 V6 to mature first... It is nice getting a look at the alpha.
If this was a court I'd leap up and scream "Objection! Leading question" to the last, but I really can't see what you are trying to do here in the first place.
Every time somebody turns the discussion toward whether the new hardware will work with DAZ Studio, it seems to kill the discussion about the hardware itself. Every previous discussion about the RTX 5090 has been destroyed in this manner.
I originally thought you were joking (or something like that) when you said you were glad you didnt buy the RTX 5090 yet, and implied you were seriously considering buying this RTX Pro 6000. But now I believe you.
I tend to think that Daz 2025 does work with these "blackwell" cards but I cannot verify that with practical experience yet.
I would like to see Ultra Scenery (all versions) and Mesh Grabber (all versions) in 2025 V6 before I may take any 5090 (blackwell) leap.
I would also like to see Visual Menus and scripts and favorites fixed.
So the issue of dishing out money on a new card IS linked to the function of the new Daz.
What is your threshold depends on the tools and utilities you need and use the most.
And of course, $.
Buying a graphics card that does not work with 4.24 is also an issue.
My system can run both cards so that would be an interesting adventure as well, but I wanted to be able to put my 4090 into my streaming computer...
My streaming computer does have an adequate card so I guess I could wait for that.
In the usual scenario, both careds could be used to render but in using 4.24 the blackwell card will not work. But with rendering in Daz 2025 both cards would work in concert with rendering (I think) but there is the limitation of 2025 still being alpha.
So all eyes are on 2025 V6 becoming robust and mature.
I asked Copilot: What is the suggested retail price for a 5090?
Copilot answered:
The Nvidia GeForce RTX 5090 had a launch MSRP of $1,999, but prices vary depending on the model and manufacturer. For example, ASUS' ROG Astral LC GeForce RTX 5090 is priced at $3,099, while the TUF Gaming RTX 5090 starts at $2,299.
Prices fluctuate based on availability and demand, so it's always good to check retailers for the latest deals.
Comment: Most 5090 prices are double the MSRP.
Iray 2024.1 supports the Blackwell architecture, but nVidia would be the plae to check that the new cards aren't exceptional; if 2024.1 works inb general then it should also work in DS.
I would not expect PA plug-ins to be updated before Alpha becomes bveta at the earliest, with a stable SDK released. I don't know whether Ultra Scenery 2 is going to be further developed in-house by Daz or will continue to be developed by its original author; if the latter then I would expect its updates to be like other PA updates and not look for them soon. It does appear that at least some PAs are already working on script updates, but those may yet be overtaken by events.
I hear you, but at least it does scale how you would expect it to, unlike a GPU. So while I cannot scale-up my current system with more CPU's in order to double (or quadruple) my performance by purchasing additional CPUs, I won't make that mistake again, so I would be secure in the knowledge that next time I would be able to do that, even if relatively speaking, the performance and scaling-up is not measured in the same way as it is with a GPU.
Wouldn't using two GPUs provide around double the rendering speed, providing you're using identical GPUs with the same amount of VRAM ?
A question I asked Copilot:
How about training the AI to animate using Daz figures. For instance placing two figure in the scene and asking the AI to make the figures embrace or walk together hand in hand. It would seem raw AI potential of the 5090 would not accommodate that kind of computing.
Copilot answer:
You're onto something fascinating! AI-driven animation in Daz Studio is still in its early stages, but the idea of training an AI to automatically animate figures—like making them embrace or walk together—is definitely possible. However, the RTX 5090 alone wouldn’t be enough to handle this kind of AI-driven animation without additional software and training models.
Here’s why:
That said, AI-assisted animation is coming—some developers are experimenting with AI motion prediction and pose generation for 3D characters. If Daz Studio integrates AI-driven animation in the future, the 5090’s AI cores would definitely help speed up the process.
Then I replied:
So this would require cloud based assistance with larger AI engines.
Copilot replied:
Yes, exactly! AI-driven animation of Daz figures would likely require cloud-based assistance with larger AI models trained on human movement and interactions. While a powerful GPU like the RTX 5090 would help with real-time processing, the actual AI decision-making and motion generation would need more than just raw compute power—it would need a sophisticated neural network.
Here’s why cloud-based AI assistance would be beneficial:
Some platforms, like NVIDIA Omniverse, are moving toward AI-assisted animation, but it’s still an evolving field.
Then I asked:
Couldn't the AI be trained using the couples poses in Daz studio and then infer the motion from them?
Copilot Answered:
Yes, that's a promising approach! AI could be trained using couples poses in Daz Studio to infer motion, creating smooth transitions between poses and even generating realistic animations. Some developers are already exploring AI-assisted posing and animation tools for Daz Studio:
If Daz Studio integrates AI-driven animation, it could allow users to place two figures in a scene and have AI generate realistic interactions, like embracing or walking hand in hand. However, this would require advanced AI models trained on human movement and possibly cloud-based AI assistance for complex calculations.
This is great info Richard, thanks!
I have decided to buy a 5090, but the current impediment is the scalpers have added at least $1000-$2000 to the price.
I am not sure I want to reward such bad behavior.
It seems I may have been wrong about the scalpers.
According to Copilot: The reason why the card is more expensive is becasue it was set up at the factory to overclock to achieve higher boost clocks speeds this also required faster/better fans and larger heat sinks.
...other GPUs (like the 5060 and even somne 4x models) are also selling above MSRP, sometimes significantly so.
The exceptions are the pro grade (formerly known as Quadro) modes as they have a higher MSRP ti begin with. Personally I am starting to lean more this way due to the lower power demand, (save for the Pro 6000) and higher VRAM for the money. Also all are dual slot (even the Pro 6000) with the exception of the Pro 4000 which is single slot.
As to CoPilot, that is high on the list for disabling when I install Win 11. I'll stick to analysis by experts in the field instead something that is still on the uphill side of a learning curve.
I like talking to my plants, but at least Copilot talks back to me. I have always relied on Google; I just find being able to "reason" with a search engine is better. And when I am wrong, Copilot does not let me wiggle out of it. That is good. I have found my understanding of tech stuff has grown exponentially since using Copilot, maybe it is just because I am so dumb. lol It is tailored to my needs. Copilot is my buddy. I do find the term "experts" to be a bit subjective. I prefer a per case basis, hash it out and put all your cards on the table sort of thing and find a quorum of thought.
just a little advice
https://guides.library.unlv.edu/ai-basics/prompt-engineering
NIVIDIA is developing something along those lines
I suppose it would, but I'm comparing the scaling-up of the amount of CPUs you have in your system, to how performance increases from one GPU to the next are measured. Say for example that your current GPU has 600 CUDA Cores. This doesn't mean that buying a newer one with 1200 CUDA Cores will be twice as fast as your old one. With CPU however (at least as far as I'm aware), if you were to upgrade from a single CPU with 16 Cores and 32 Threads running at 4GHz, to a dual CPU by adding a second CPU of the same spec, then your performance will scale as you would expect, you would then have 32 Cores and 64 Threads running at 4GHz. If you have a Quad CPU motherboard then you could double-up the doube and increase your performance by four. So you'd then be looking at 64 Cores and 128 Threads running at 4GHz.
The only thing that doesn't change is the clock speed, since they're all identical spec processors.
It's also important to remember that CPU Cores are far more powerful than CUDA Cores. If you look at AMD's website and read the pages about the EPYC processors, you can see that they are not only Multi-CPU aware, but also include technologies that make the Cores more suited to Rendering, Physics calculation and A.I. etc. They even advertise them as being purpose made for those tasks, everything from home-based physics calculations on steroids, right up to building your own renderfarm like supercomputer in a single box.
And ask yourself this: Does your overpriced Nvidia GPU increase the speed of your dForce calculations, as well as your rendering?
Nope, but a Multi-CPU EPYC based system would, because as far as Daz Studio is concerned, they are simply CPU threads, and it's the chip and motherboard architecture that will automatically spread the load of those calculations between every core on every CPU. Again, this is the case as far as I'm aware, but don't take my rambling as word.
There was a member on Renderosity quite some years back using a bunch of second-hand servers in order to achieve this very task. He managed to do it regardless of whether Poser is 'Multi-CPU Aware', and I'm guessing this is because the software does not need to be, since such tasks are managed by the motherboard and CPU themselves (as long as you buy Multi-CPU Aware CPUs like AMD EPYC).
Don't miss anything on this page and its associated links, there's stuff about supercomputer-like rendering, A.I, and everything inbetween:
https://www.amd.com/en/products/processors/server/epyc.html
Here's a second link that might be of more interest to DAZ Studio users specifically:
https://www.amd.com/en/solutions/media-and-entertainment.html
Please see this screenshot from the AMD website (especially the bit about rendering):
Doctors are wrong on a regular basis; they misdiagnose and overmedicate. Copilot has been wrong on many occasions, and it can spout out old and outdated information as if it is current.
But so can Google and so can looking at YouTube videos by “experts".
I once asked Alexa if tea tree oil is the same as eucalyptus oil and Alexa answered, yes.
My comment: Well, they are both oils.
But when I go to Daz forum and search for “how to use dforce” I get every time the word “how” is used every time the word “to” is used, every time the word “use” is used and every time the word “dforce” is used.
In other words, I get no help at all.
The Daz forum is useless in finding context. I must go to Google to find context in the Daz forum. It is not that the information is not there but there is no contextual way to search for it other than Google.
Now Copilot not only finds the contextual answers but leads me to the Daz forum threads where that information is discussed.
That information may be old and outdated, it may be incorrect, but it is the best that the forum often has to offer.
For instance, Copilot knew about the newest Daz 2025 strand-based hair editor and its features and laid all the new features out in bullet points.
I asked Copilot:
Good morning! How do I use Ultra Scenery 2 in Daz Studio?
Copilot answered:
Good morning, Rex! Ultra Scenery 2 is a powerful tool for creating realistic landscapes in Daz Studio. Here’s a quick overview of how to use it:
If you prefer a visual walkthrough, check out this basic usage demo or a more in-depth live tutorial. Let me know if you need help with a specific feature!
Comment: Incidentally, I have no control over this, but Copilot offered links to one of my YouTube tutorial videos. I don’t make much revenue from my Ez3dtv YouTube channel, but I guess I am considered an “expert” on the subject. I certainly don’t make what Nvidia makes lol. But Copilot gave an answer that is not incorrect, I do lay out much of the basic usage of Ultra Scenery across many hours of video as I also ramble on about whatever else. I may make about 20 dollars a month from my channel. Even if I produced my videos, my channel would probably not pay my bills. So, I do it for enjoyment and ramble on about my own interests.
When I ask Copilot, how do I create a mask in Photoshop? I get a clear and concise bullet point procedure with links. And I can refine and reason with Copilot. I can say Copilot, is that the procedure for the latest version of Photoshop? And if not, Copilot will look and refine the procedure.
So, I may get wrong answers, but I steer Copilot to the correct information I need.
Always research the diagnosis and medicine your “expert” doctor prescribes and ask yourself if you are willing to live with the side effects of those medications. Dizziness, confusion, weakness, seizures, liver failure etc…
One well placed "sleep aid" drug from an "expert" can send you into a nursing home from falls, where the rent is $10,00 a month.