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
No, I don't I'm afraid.
Boy, I really need to upgrady my Rig. Just ran Outrider42's scene and it took 23 minutes and 34.81 seconds. The GTX 960 4GB is really asthmatic. Getting an almost free EVGA 08G-P4-6571-KR GTX 1070 to upgrade my GPU. Hopefully it will help.
Current System Specs:
Intel i5 4670K
Gigabyte Z97X-Gaming 7
EVGA GTX 960 SSC 4GB
16GB (2x8GB) Patriot Viper 3 DDR-3 1866 memory
Samsung 860 EVO 1TB Sata III SSD
Hitachi HDT721010SLA360 1TB SATA II HDD
Western Digital Blue WD20EZRZ 2TB SATA III HDD
Corsair HX1000W PSU
Windows 10 Professional 64
DS 4.10.0.123 Pro 64 bit
Thanks for posting. Is the 960 throttling?
Are you going to add that 1070 as a second GPU or swap them? You have more than enough headroom with a 1000 Watt PSU to run both. The 960 and 1070 together would be a decent combo if the scene fits them both.
Decided to post what would be more of a hobbyist machine. The 960 was running at speed but not overclocked. OptiX was off and Optimization was set to speed (thought OptiX was checked but it was not).
I will be swapping cards so I can use all 8GB of the 1070. I may pick up a second 1070 since they still show avaialble on the EVGA website. I bought the Corsair HW1000 PSU about 12 years ago (when it was $330.00) but it has been a great PSU. Before I trimmed down GPUs and HHDs, I had it running in a gaming rig back when the E-pin contests at Quakecon was who could show up with the most storage in their gaming rigs. I topped out at about 6.8 GB (primary raid 0 array with 2 WD 660 MB HDDS, 2 750 GB HDDs, and 4 1TB HDDs with two Radeon 4670s). Thought I had it won until a friend showed up with a DC++ server with 26TB of storage that was full of data. After I got my CoolerMaster CM 690 II Case with a HDD tock, I didn't need to have all the HDDs on the case any more. The GTX 960 is 3 years old so it is time for a new GPU. it is also about time to upgrade m yCPU, starting to see games (Fallout 76) that the CPU dies not meet minimum specs. I should have the new (to me) card this weekend and will rerun the bench.
I don't know how you feel about buying used, but there are tons of 1070's on ebay, and 1070ti's too. I've been thinking of picking up a 1070ti as a 2nd card so I can have 2 cards rendering more often. I have a 1080ti and 970 right now. Sometimes the 970 doesn't have enough VRAM. It certainly not urgent, but if I find one at a good price I'll buy it.
I always like seeing as broad a spectrum as possible, and weird build are always welcome, too. We even got people running 1070s with Core 2 Quads, and I love it.
Actually, looking at your build, your PC is strikingly similar to mine! I have the same i5-4690k, the same Samsung 860 1TB, an EVGA 1000 Watt PSU, 16GB RAM, and several drives connected. Its almost freaky how similar they are, LOL.
If you are running 1080p, the i5 wont be much of an issue. It will bottleneck, but not enough to drop you under 60. At higher resolutions it will be an issue. I have not played Fallout 76, but I have played games that recommend more than the i5 without issues. When I jump to 4K I plan on rebuilding. I haven't decide on what because I want to balance Daz and gaming. I'm looking forward to Ryzen 3000. Most games really aren't dependent on CPU these days.
Ran the test with the GTX 960 again last night with OptiX on and got the following time: Total Rendering Time: 20 minutes 31.24 seconds so it ran just over 3 minhtes faster with OptiX on.
Now just ran with my freshly installed GTX 1070 and geuss what: 9 minutes 14.69 seconds! So jumping up on level and advancing one generation cut my render time in half.
I am now a happy camper and trying to figure out how to get anotehr one to REALLY up my Cuda Cores.
Getting ready to go back and try to render some of the problem images from the past that kept going to CPU and will see how it works on my D-Force hair products. They kept blowing up before.
Pascal was a great jump over the 900 series, so that is not so surprising. That's also why when people stress getting the best GPU you can afford for Iray, they really mean it. There is no substitute for GPU power for Iray as long as it has the VRAM to run the scene.
You'll also find CUDA core are not equal across generations, generally the new generations are faster. The best example of this is comparing the 6 year old GTX 680 to the 2 year old GTX 1050ti. The 680 was for its time, the most powerful GPU you could buy, and was a very popular card along with the 670 (one of which I happily owned.) The 680 has 1536 CUDA cores. The newer 1050ti has exactly half that amount, 768. So at face value you might think the former GPU king would destroy the meager 1050ti, but that is not the case. In fact, the 1050ti is actually slightly faster. That shows just how much tech advanced in a few short generations. What was once the fastest single GPU you could buy is not barely enough for mid tier in the span of 4 years. Along with advancements to CUDA, the 1000 series has much higher clockspeeds across the board, and that combination makes them very fast. So now you have a 1070 that has a lot more CUDA, these CUDA are not only more efficient, they are clocked much faster. And the icing the cake is 8GB VRAM, much more than previous generations.
BTW, the 960 is neck and neck with the 1050ti, they trade blows in gaming benchmarks.
Rendering Outrider's scene I scored==> Total Rendering Time: 6 minutes 11.20 seconds
Running (2) GTX 970's, I7 7700K, MSI Z270 Gaming M7 Mother Board, Overclocked to 4.7 GHz
i7 5930K, Asus X99 Deluxe, 2 x GTX 970
at T=6 min, render was only at 92%, got impatient and just canceled it
Gigabyte 2080 TI Gaming OC
GPU rendering time = 1 min 27 sec
weird
GPU and CPU rendering time = 1 min 29 sec
I was planning to get 2 GTX 1080 TI's, but a lot of the venders in Canada jacked up the prices to the point where they were almost the same price as a 2080 TI. So I figured I'd get a 2080 Ti now, and a 2nd when the prices lower a bit. IF the 2080 TI wasn't so fat, I could have installed a 3rd GTX 970 I have.
i7 5930K, Asus X99 Deluxe, 2 x GTX 970, Gigabyte 2080 TI Gaming OC edition
Optix prime acceleration on
GPU's only rendering time = 49 sec
i7 5930K, Asus X99 Deluxe, 1 x GTX 970, Gigabyte 2080 TI Gaming OC
Optix prime acceleration on
GPU's only rendering time = 54 sec
i7 5930K, Asus X99 Deluxe, Gigabyte 2080 TI Gaming OC
Optix prime acceleration on
GPU only rendering time = 1 min 5 sec
Intel I5 8500
Asus Z370 A
2 x 8GB 3000mhz ram
Samsung Evo970
MSI Geforce 2080 TI Ventus OC
SickeYield test scene:
Total Rendering Time: 1 minutes 8.55 seconds
5000 iterations
No iray viewport. I just opened the scene, disabled CPU render, enable optix prime and hit render.
Latest public beta, 4.11
GPU temperature hit 55 celcius during the render. I think the clock speed of this card is capped at 1920mhz. It was operating at 1920mhz during the render, most of the time. 2, 3 times I saw it dropped to 1880Mhz, but it was mostly stable at 1920mhz.
Intel i7-9700
Asus Z390-A
Corsair 2x8GB DDR4-3600
EVGA - GeForce RTX 2080 8 GB XC ULTRA GAMING Video Card
Samsung Evo 860 1 TB
Public Beta 4.11
Sickleyield's Scene:
CPU+/GPU+/Optix Prime- : 2:00 mins
CPU-/GPU+/Optix Prime-: 1:50
CPU-/GPU+/Optix Prime+: 1:24
CPU+/GPU+/Optix Prime+: 1:23
Did someone realize a graphic or something that compares the average performance per GPU model? :D
ASRock Z290 Gaming K6
i7-7700K @ 4.2GHz
32GB RAM
GTX 1080 + GTX 1080Ti in Thunderbolt 3 eGPU case
w/o Optix Prime: 2:04
with Optix Prime: 1:14
I’m running a long batch of renders right now, but I’ll give this a shot once I’m done.
EDIT: Details
To be clear, the 1080Ti is the only one in the eGPU case. Presently, there isn't an off-the-shelf case that will do more than a single eGPU. I'm using the Akitio Node, which I've had for a while. It used to have a 980Ti in it and I used it with my Mac laptop. Today's configuration is on my Windows Desktop with a newly added ASRock Thunderbold 3 AIC (Add In Card). The GTX 1080 is in PCIe slot 2. I have a 1050Ti in slot 1 for handling the display(s)
EDIT: More Stats
1080-Optix: 4 minutes 46.65 seconds
1080+Optix: 2 minutes 52.42 seconds
1080Ti-Optix: 3 minutes 27.98 seconds
1080Ti+Optix: 2 minutes 3.71 seconds
Thanks for that. It looks like the egpu is not slowing it down at all.
Just one more question, which version of DS, is 4.10 or the 4.11 beta?
I'm running 4.10. If I do any experimenting with 4.11, I'll be sure to post the results. When the 2080 Ti becomes affordable, I'll probably switch to an open frame case so I can use all three cards without buying another eGPU case.
Hi. I did the test and I'm wondering - everything was pre-set in the scene file, correct? Nothing we'd have to adjust?
This is the first I've even tried Iray. I had deleted everything from My Library a long time ago, so I simply installed G2F Starter Essentials and ran the test in DS4.10 Pro. I made no changes to the setup, but I wouldn't know what to change, anyway. Heck, I can barely remember how to use 3DL.
It took 11 minutes, 23 seconds to reach 100%. Still lots of fireflies, especially in the lower right.
I have an i5-2500K with a GTX 750 Ti, and the MB has PCI-e 2.0 x 16 slots. So I guess I can't expect the 9 minutes with a 750 Ti that I see commonly in these tests. I'd planned to get a GTX 1070 to reach the average 3 minutes I see here. If my math is correct, a 1070 might render it in just under 4 minutes on my system.
But this scene was only 400x520! So a normal-sized scene, yipes. And a scene with actual cool stuff, forget it. And, uh, ... my goal was animation!
Anyway, I'm still looking for a bank to rob in order to get a system with an RTX 2080 Ti. Then, watch me roar!
You don't need to change anything in the scene, its all set up and render to render. However, there are some settings in the Iray Advanced Settings that can change times and help. I bet what you need is to turn Optix ON and you will get near the times posted by other 750ti users. Optix makes a big difference in many situations. The other option is under the Iray Editor tab, there is an option under Optimization to choose 'memory' or 'speed'. Choosing speed can help a little bit. The other tip is to turn on the Iray preview in the viewport before rendering, this will shave some time that it takes to load the scene, and we are not interested in benchmarking that as there are too many factors that effect it like your harddrive speed. So with those things you should be able to hit better times in line with other 750ti's.
However, you are indeed being held back by your hardware. I will point out that pcei 2.0 does NOT make a difference at all, so no, that is not your issue. Your issue is simply your 750ti. That card is just not ideal for Iray. Not only is it slower, but it is also heavily restricted by only having 2GB of VRAM. If you run out of VRAM, Iray will stop using the GPU completely and fall back to CPU, which is terribly slow even with most new CPUs. If you by chance made the scene resolution very large, you would run out of 2GB VRAM very quickly. Trust me, I used to have a GTX 670 2GB. It was OK in speed for its time, but that 2GB was the killer. Even 4 is tough. But how much you need depends a lot on what you want to do.
If you want to do animation, I am afraid your machine is not going o handle it for Iray. Iray would be a brutal choice for animation regardless. It is possible, but you would have to cut corners. The new denoiser in 4.11 beta may help with animation, but you would still need better hardware to render animation, or even stills for that matter. The good news is that if your budget is really tight, you can render pretty darn well with old PCs like yours if you get a fast GPU in it. There are still caveats, of course. You need to be sure your power supply, case, and motherboard can all handle what you buy. But you can buy a decent GPU and get rendering faster today. Still you will want to be saving up for a proper new machine to install that GPU into later, but for the short term this can help you get started.
As for what can be better, that is very tough and up to what you want to do. I personally would look at a 1070 or 1070ti at this time. Prices have come down on these cards, especially if you are not afraid of buying them used. The 1070ti is actually just as fast at Iray as the 1080, so it is a fantastic value. All the 1070s, 1070ti's, and 1080s have 8GB of VRAM, which would be a big bump over the 2GB you have and should fit reasonable scenes. It would be quite hard to get more VRAM than that, as only the 1080ti and 2080ti offer more. And of course, in case you were not aware, you can only use Nvidia cards for Iray, no AMD.
For animation, you may want to consider a different render engine. I don't animate, so I am not one to offer any suggestions.
My 750Ti has 4GB RAM.
Thanks outrider42. I just checked and Optix was on. I may have set it back when I once tried following a tutorial. I'll study the rest of your points very soonly, because I'm really tired of waiting for the "right time" to begin honestly learning this stuff.
I do have the 2GB version of 750Ti. I'm sure the test scene didn't overshoot that. Today I'm going to see about getting a small, single-fan 1060 with 6GB VRAM, and just use it for the CUDAs. I have a 750W PS, so I'm sure that'll be OK. I'm not sure if I want to add my 750Ti's CUDAs, for fear of generating too much heat blowing up across the CPU area. My system is old and it's all I have.
I'm sorry I can't study your suggestions in detail immediately. I have some doctoring to be done (on me) also today. But this is looking very 'doable!' For more speed, I could create most of the scenes in 3DL and composite close-up props in Iray over them. Use backgrounds in the backgrounds. Thanks again for your response and info!
If you play with the 4.11 beta, one of the best new features is the denoiser. The denoiser uses AI to 'guess' what the unrendered pixels will look like. It also nearly eliminates all noise in final renders. It can cause a loss of fine detail, but it can also help a scene look finished long before it is done converging. So it may not be great for people's skin in close ups, but otherwise it can be a big difference maker. It is really good for pumping out backgrounds to use later for compositing, for example. And for animation the denoiser might make the biggest difference of all. You can place a limit on how many iterations Iray runs, and you play around with it to see how many iterations (or convergence %, either can work) it takes on average for a scene to look "good enough". Once you do that, you can try rendering animation using these limits. The denoiser can potentially make images look decent even when they are very low converged. Since animation is in motion, most people wont notice minor flaws in individual frames, so you can get away with more. So if you can render each frame to say 10% instead of 100% you can shave a ton of time and potentially make some kind of animation in a reasonable amount of time.
The 1060 6GB is not bad. Its certainly going to be better than the 750ti, several times faster. I will point out that the 2060 or 1160 are due out soon, so you may want to hold on just a bit longer. Rumors suggest the 2060 is as fast or faster than the 1070ti in gaming. As it stands, there appear to be a lot of different versions, including different VRAM specs. The 2060 will have some Turing features (ray tracing and tensor cores), while the 1160 will only have CUDA cores. People are speculating the 2060 to be about $350, but the 1160 might be as cheap as $200. So if the 1160 is real it could be a great deal. However while the 2060 is confirmed to be true, the 1160 has not been officially listed by Nvidia at all. So it may be a ways off from production. The only clue about the 1160 comes from a laptop listing showing the 1160 as an option.
So, it may eventually be possible to incorporate more Iray than 3DL in an animation, with denoiser! There's also Resource Saver, which I just bought. And Scene Optimizer. My plans now are to create very short videos, almost like animated gifs, so this is really very doable. Your information is absolutely valuable, showing me a clearer path ahead.
I won't be waiting for a 2060 etc, though. I want a 2080 Ti in a new machine which I'll put together bit by bit, so to speak. Today I ordered the 1060 6GB for my current machine for the reasons stated earlier. Got it for $250. A 2060,though, could some day be added to my 2080 Ti machine!! Those decisions will necessarily need to wait until we see more info on the cards, but also, what sort of changes are to come with Iray.
I don't know how interesting my comments are to many people here. I'm struggling because I'm not that tech-savvy. But these are the factors that people in my situation, retired with quite low income, need to understand if we want to join the game. So, thank you again for your information, and thanks to SickleYield for the scene test! It helped me to start thinking about alternatives and compromises. Very exciting!
As it is, the denoiser still serves its purpose, as it can remove most grain and make the render more acceptable at an earlier stage of convergence. Like on the newer bench you can probably get away with stopping the render sooner than without and get a respectable result. To take advantage of Tensor and RT, you have to write code to do that. While the beta can use Turing, the beta predates the Turing launch. The fact is we are very lucky that Turing JUST HAPPENS to work in Daz as the drivers are apparently similar to that of Volta. Support for the Titan V was added with the Daz beta.
However, there is something very curious about the render times that Turing is getting, because they are faster than what you would expect them to be if they were only using the CUDA cores. So I do wonder if the beta is in fact using Tensor to aid in the actual rendering process. Again, since the Titan V does have Tensor and it is supported, it is possible that this may be the case.
The reason why: The 2080ti is about 100% faster than the 1080ti. However for every other task the 2080ti is only about 30% faster, maybe 50% if you cherry pick. But not 100%. And while the 2080ti has more CUDA, it doesn't have double the CUDA, it is not clocked faster, and the IPC gains of Turing are not that huge. So this leads me to believe that Tensor is being used. Unfortunately, we don't have any data to help verify this, and this only my speculation.
Also, note that Iray uses the OptiX Prime SDK, which is a little different than full OptiX.
Lol oops.
Long time lurker, first time poster here. Been lusting after all things 3dcg (and especially DAZ related) for a good 9+ years, but only recently came into a life situation where I could start pursuing things properly.
All this is to say that I had my first dedicated gpu upgrade in about ten years arrive in the mail last week in the form of a Titan RTX, and I plan to give it the full DAZ Studio deep dive treatment asap.
because 2080ti CUDA is bigger O(∩_∩)O
Yeah, outrider42, keep in mind that Turing cuda cores have been fundamentally re-designed from their Pascal predecessors both in terms of cache size/bandwidth and support for independent floating-point and integer data paths in each streaming multiprocessor. The purported result of which is a code-transparent performance uplift of somewhere in the realm of 50-100% per cuda core for compute-heavy workloads. So the prospect of a 2080ti being 100% faster than a 1080ti for Iray workloads based on cuda core performance increases alone really isn't all that far-fetched.
For what it's worth, to the best of my understanding OptiX Prime is just a subset api of plain old OptiX which software optimizes hardware rendering platforms for "simplistic" ray-traced workloads ("simplistic" as in triangles-based, which is all that a rendering platform like DS/Iray ever needs.) This explains why the net effect of enabling OptiX Prime on Turing GPUs so far observed seems to be a 1-3 second increase in total render times rather than the significant decreases seen with previous GPU generations like Pascal. Remember - the Turing platform is already hardware optimized for ray-traced workloads. Whatever hardware-level processing optimizations are being accomplished through enabling OptiX Prime on Pascal are already being done automatically at the hardware level on Turing. Meaning that the only thing to be gained by enabling OptiX Prime on Turing would be an initial several second increase in rendering prep time while the software needlessly pre-optimizes code.
Btw this also explains why OptiX Prime is slated to not be getting hardware-based ray-tracing support. On a platform like Turing with hardware-based ray-tracing support, OptiX Prime is effectively already baked-in and always turned on. So "giving it support" for those features would make no sense.
Also, for those stressing out about OptiX slated to be getting hardware RT support while OptiX Prime not getting it and how that's going to effect RT support in DS/Iray since it only seems to support OptiX Prime: OptiX Prime is a subset api of OptiX. This means that any program which lists OptiX Prime as an optional rendering pipeline (eg. DS/Iray) is implicitly using the full OptiX api any time that option isn't selected. So just as long as Daz continues to get fresh updates of Iray once full hardware RT support appears in the OptiX api and Optix Prime is turned off in advanced render settings, Daz Studio should be able to benefit from hardware-level ray-tracing acceleration with Turing GPUs. At this point it should just be a matter of waiting for api/software updates...